arXiv:gr-qc/9303025 v8 5 Oct 2001
RadiationfromaUniformlyAcceleratedCharge
andtheEquivalencePrinciple
StephenParrott
DepartmentofMathematics
UniversityofMassachusettsatBoston
100MorrisseyBlvd.
Boston,MA02125
USA
March1,2002
Abstract
Wearguethatpurelylocalexperimentscandistinguishastationary
chargedparticleinastaticgravitational eldfromanacceleratedpar-
ticlein(gravity-free)Minkowskispace.Somecommonargumentstothe
contraryareanalyzedandfoundtorestonamisidenti cationof\energy".
1Introduction
ItisgenerallyacceptedthatanyacceleratedchargeinMinkowskispaceradiates
energy.Itisalsoacceptedthatastationarychargeinastaticgravitational
eld(suchasaSchwarzschild eld)doesnotradiateenergy.Itwouldseemthat
these two facts imply that some forms of Einstein’s Equivalence Principle do
not apply to charged particles.
To put the matter in an easily visualized physical framework, imagine that
the acceleration of a charged particle in Minkowski space is produced by a tiny
rocket engine attached to the particle. Since the particle is radiating energy
which can be detected and used, conservation of energy suggests that the ra-
diated energy must be furnished by the rocket | we must burn more fuel to
produce a given accelerating worldline than we would to produce the same world-
line for a neutral particle of the same mass. Now consider a stationary charge
in Schwarzschild space-time, and suppose a rocket holds it stationary relative
to the coordinate frame (accelerating with respect to local inertial frames). In
this case, since no radiation is produced, the rocket should use the same amount
of fuel as would be required to hold stationary a similar neutral particle. This
gives an experimental test by which we can determine locally whether we are
1
2
accelerating in Minkowski space or stationary in a gravitational eld | sim-
ply observe the rocket’s fuel consumption. (Further discussion and replies to
anticipated objections are given in Appendix 1.)
Some authors (cf. [3]) explain this by viewing a charged particle as inextri-
cably associated with its electromagnetic eld. They maintain that since the
eld extends throughout all spacetime, no measurements on the particle can
be considered truly local. To the present author, such assertions seem to di er
only in language from the more straightforward: \The Equivalence Principle
does not apply to charged particles".
Other authors maintain that the Equivalence Principle does apply to charged
particles. Perhaps the most in uential paper advocating a similar view is one
of Boulware [2], an early version of which formed the basis for the treatment
of the problem in Peierls’ book [11]. This paper claims to resolve \the equiv-
alence principle paradox" by establishing that \all the radiation [measured by
a freely falling observer] goes into the region of space time inaccessible to the
co-accelerating observer."
A recent paper of Singal [8] claims that there is no radiation at all. Singal’s
argument, which we believe awed, is analyzed in [7].
The present work analyzes the problem within Boulware’s framework but
reaches di erent conclusions. He shows that the Poynting vector vanishes in the
rest frames of certain co-accelerating observers and concludes from this that
\in the accelerated frame, there is no energy ux, ... , and no radi-
ation".
Singal [8] rederives a special case of this result (his equation (7) on page 962),
and concludes that \there are no radiation elds for a charge supported in a a
gravitational eld, in conformity with the strong principle of equivalence.
We obtain a similar result by other means in Appendix 3, but interpret
it di erently. We believe that the above quote of [2] incorrectly identi es the
\radiated energy in the accelerated frame", and therefore does not resolve what
he characterizes as a \paradox".
Also, we do not think there is any \paradox" remaining, unless one regards
the inapplicability of the Equivalence Principle to charged particles as a \para-
dox". Even if the Equivalence Principle does not apply to charged particles, no
known mathematical result or physical observation is contradicted.
2 What is \energy"?
The identi cation of \energy" in Minkowski or Schwarzschild spacetime may
seem obvious, but there is a subtlety hidden in Boulware’s formulation. This
section examines this issue with the goal of clearly exposing the subtlety.
To deserve the name \energy", a quantity should be \conserved". The follow-
ing is a well-known way to construct a conserved quantity from a zero-divergence
3
time
space
space
3-dim volume at a xed time
? 2-dim boundary of the volume at a xed time
2-dim boundary evolving through time
Figure 1: One space dimension is suppressed. The \top" and \bottom" of the
box represent three-dimensional spacelike volumes; the \sides" represent two-
dimensional surfaces moving through time; the interior is four-dimensional.
symmetric tensor T = Tij and a Killing vector eld K = Ki on spacetime. Form
the vector vi := Ti K (repeated indices are summed and usually emphasized
by Greek and \:=" means \equals by de nition"), and note that its covariant
divergence v j vanishes ([12], p. 96).
By Gauss’s theorem, the integral of the normal component of v over the
three-dimensional boundary of any four-dimensional region vanishes.1 Such a
region is pictured in Figure 1, in which one space dimension is suppressed. The
particular region pictured is a rectangular \box" with spacelike \ends" lying in
the constant-time hyperplanes t = t1 and t = t2 and time-like \sides". (We
use t as a time coordinate and assume that it is, in fact, timelike.) The \end"
corresponding to time ti, i = 1;2, represents a three-dimensional region of
space at that time. The integral of the normal component of v over the end
corresponding to t = t2 is interpreted as the amount of a \substance" (such as
energy) in this region of space at time t2. The integral of the normal component
over the sides is interpreted as the amount of the substance which leaves the
region of space between times t1 and t2. Thus the vanishing of the integral
over the boundary expresses a law of conservation of the substance. Similar
interpretations hold even if the boundary of the region is \curved" and does not
necessarily lie in constant coordinate surfaces.
We shall take as Tij the energy-momentum tensor of the retarded electro-
magnetic eld produced by a charged particle whose worldline is given. That
is, if F = Fij is the electromagnetic eld tensor, then
Tij := Fi F j (1=4)F F gij ; (1)
where gij is the spacetime metric tensor. Given T, to every Killing vector eld
1When there are points at which the boundary has a lightlike tangent vector, this must
be interpreted sympathetically; see [5], Section 2.8 for the necessary de nitions. However,
we shall only need to integrate over timelike and spacelike surfaces, on which the concept of
\normal component" is unambiguous.
4
x
t t = x
:5 1 1:5
X = :5 X = 1 X = 1:5
D
DD
BB
AA
TT
J
JJ
D
D
DD
B
B
A
AA
T
T
%
%
%
DD
L
AA
S
S
SS
e
e
e
Figure 2: The orbits for the ow of the one-parameter family of boosts (3).
K corresponds a conserved scalar quantity as described above. We have to
decide which such quantity deserves the name \energy".
In Minkowski space, the metric is
ds2 = dt2 dx2 dy2 dz2 ; (2)
and there seems no question that the energy is correctly identi ed as the con-
served quantity corresponding to the Killing vector @t generating time trans-
lations. (We use the di erential-geometric convention of identifying tangent
vectors with directional derivatives.) If this were not true, we would have to
rethink the physical interpretation of most of the mathematics of contemporary
relativistic physics. Translations in spacelike directions similarly give Killing
vectors whose corresponding conserved quantities are interpreted as momenta
in the given directions.
There are other Killing vector elds which are not as immediately obvious.
For example, consider the Killing eld corresponding to the ow of the one-
parameter family 7! ( ; ; ; ) of Lorentz boosts
(t;x;y;z) := (tcosh +xsinh ;tsinh + xcosh ;y;z) : (3)
The relevant timelike orbits of this ow (curves obtained by xing t;x >
0;y;z and letting vary) are pictured in Figure 2. For xed y;z, they are
5
hyperbolas with timelike tangent vectors. Any such hyperbola is the worldline
of a uniformly accelerated particle.
On any orbit, the positive quantity X satisfying
X2 = (tsinh + xcosh )2 (tcosh +xsinh )2 = x2 t2
is constant, and its value is the orbit’s x-coordinate at time t = 0. Thus an orbit
is the worldline of a uniformly accelerated particle which had position x = X
at time t = 0.
Such an orbit can conveniently be described in terms of X as the locus of all
points (X sinh ;X cosh ;y;z), as varies over all real numbers. The tangent
vector of such an orbit is
@ := (X cosh ;X sinh ;0;0) :
This is the Killing vector eld, expressed in terms of X and . Its length is X,
so that a particle with this orbit has its proper time given by
= X ; (4)
its four-velocity @ is
@ = 1X@ ; (5)
and its proper acceleration is 1=X.
The conserved quantity corresponding to the Killing vector @ has no rec-
ognized name, but it does have a simple physical interpretation which will be
given below. We then argue that it is this quantity which [2] (p. 185) identi es
(mistakenly, in our view) as the relevant \energy ux" in the accelerated frame.
3 Energy in static space-times
Consider a static spacetime whose metric tensor is
ds2 = g00(x1;x2;x3)(dx0)2 +
3X
I;J=1
gIJ(x1;x2;x3)dxIxJ : (6)
The important feature is that the metric coe cients gij do not depend on the
timelike coordinate x0, so that @x0 is a Killing eld.
Another way to say this is that the spacetime is symmetric under time trans-
lation. In general, the ow of a Killing eld can be regarded as a space-time
symmetry. The symmetry of time translation was obvious from looking at the
metric, but for some metrics there may exist less obvious, \hidden" symme-
tries. An example is the Minkowski metric (2), which possesses symmetries
corresponding to one-parameter families of boosts which might not be obvious
at rst inspection.
6
Consider now the most important spacetime after Minkowski space, the
Schwarzschild space-time with metric tensor
ds2 = (1 2M=r)dt2 (1 2M=r) 1dr2 r2(d 2 + sin2 d 2) : (7)
It can be shown ([12], Exercise 3.6.8) that the only Killing vector elds K
are linear combinations of @t and an \angular momentum" Killing eld A =
K (r; ; )@ + K (r; ; )@ , where K and K satisfy some additional condi-
tions which are unimportant for our purposes. The elds @t and A commute,
as do their ows. In other words, the only Killing symmetries of Schwarzschild
spacetime are the expected ones arising from rotational and time invariance:
there are no hidden Killing symmetries.
In this situation, the only natural mathematical candidate for an \energy"
is the conserved quantity corresponding to the Killing eld @t; for one thing, it
is the only rotationally invariant choice. It is also physically reasonable in our
context of analyzing the motion and elds of charged particles. If we surround
a stationary charged particle2 by a stationary sphere which generates a three-
dimensional \tube" as it progresses through time, the integral of the normal
component of T0i over the tube between times t1 and t2 physically represents
the out ow of the conserved quantity corresponding to @t between these times.
When the calculation is carried out, it is seen to be the same as integrating
the normal component of the Poynting vector E B=4 over the sphere and
multiplying by a factor proportional to t2 t1. It is usually assumed that
the eld produced by a stationary charged particle may be taken to be a pure
electric eld, and Appendix 3 proves this under certain auxiliary hypotheses.
In other words, B = 0, so the integral vanishes, and there is no \radiation" of
our conserved quantity. We expect no energy radiation; otherwise we would be
able to garner an unlimited amount of \free" energy, since it takes no energy to
hold a particle stationary in a gravitational eld.
Thus it seems eminently reasonable in this situation to identify the conserved
quantity associated with @t with the energy. We expect a conserved \energy",
this is the only natural mathematical candidate, and its physical properties turn
out to be reasonable.
However, these arguments lose force when hidden symmetries exist. Consider
a metric
ds2 = c(x)2dt2 dx2 dy2 dz2 : (8)
Here c(x) represents the x-dependent speed of light as observed from the co-
ordinate frame. Such a metric corresponds to a pseudo-gravitational eld in
the x-direction. By a \pseudo" gravitational eld we mean that a stationary
particle has a worldline which is accelerated in the x-direction, but the Riemann
tensor may happen to vanish for some functions c( ), in which case there is no
2By a \stationary" particle we mean one whose worldline is x0 7!(x0;c1;c2;c3) relative to
the static coordinate frame with respect to which the metric is (7), where the ci are constants
independent of x0.
7
curvature of space-time and no true gravitational eld. It is well known that
when the Riemann tensor vanishes, spacetime may be metrically identi ed with
a piece of Minkowski space.
Routine calculation shows that the only nonvanishing connection coe cients
are, in an obvious notation,
ttx = txt = c
0
c ;
x
tt = c
0c :
The four-velocity u of a stationary particle is u = c 1@t , so a stationary particle
has acceleration (Duu)k = u @ uk + k u u given by
Duu = c
0
c@x ; (9)
That is, the acceleration is in the x-direction with a magnitude given by the
relative rate of change of c in the x-direction. This acceleration Duu is what we
mean by \acceleration with respect to local inertial frames".
It might seem reasonable, even natural, to identify the conserved quantity
associated with @t with energy, in analogy with Schwarzschild spacetime. How-
ever, the reasonableness of such an identi cation must ultimately be justi ed
by its mathematical and physical consequences. We shall argue that such an
identi cation is sometimes inappropriate.
The \obvious" Killing symmetries of (8) are those associated with time trans-
lation, translations in spatial directions perpendicular to the x-axis, and rota-
tions about the x-axis. Only for very special choices of c( ) will there exist other,
\hidden" symmetries. One such choice yields the following metric, in which for
later purposes we replace the coordinate symbol x by X and t by :
ds2 = X2d 2 dX2 dy2 dz2 : (10)
The Riemann tensor vanishes for this spacetime, and it can be identi ed with
a piece of Minkowski space. If t;x;y;z denote the usual Minkowski coordinates
with metric given by (2), then this identi cation is:
t = X sinh
x = X cosh :
Moreover, the present Killing eld @ is the same as the Minkowski space Killing
eld @ discussed in Section 2.
The part of Minkowski space covered by the map ;X;y;z7!t;x;y;z con-
sists of the region jxj > jtj, but we will only be concerned with the smaller
region x>jtj, which is called the \Rindler wedge".
The coordinates ;X;y;z for this portion of Minkowski space are known
as Rindler coordinates ([13], Section 8.6). They are also sometimes known as
elevator coordinates because we shall see below that X;y;z may be regarded as
8
space coordinates as seen by occupants of a rigidly accelerated elevator. Boul-
ware [2] uses in place of for the timelike coordinate. We prefer because it
seems more natural to reserve = X for the proper time on the worldlines of
points of the elevator.
For constant y and z, a curve X = constant is the orbit oft = 0;x = X under
the ow (3). This curve is also the worldline of a uniformly accelerated particle
with proper acceleration 1=X. The set of all such curves for all X;y;z may be
regarded as the worldlines of a collection of uniformly accelerated observers all
of whom are at rest in the Minkowski frame at time t = 0.
The Rindler coordinates X;y;z specify the particular worldline in the col-
lection. The spatial distance between two points with the same Rindler \time"
coordinates say ;X1;y1;z1 and ;X2;y2;z2, is just the ordinary Euclidean dis-
tance [(X2 X1)2 + (y2 y1)2 + (z2 z1)2]1=2 . Moreover, the corresponding
spatial displacement vector is orthogonal to the worldlines of constant X;y;z.
This says that an observer following such a worldline sees at any given moment
other such worldlines at a constant distance in his rest frame at that moment.
Thus we may take a collection of such worldlines and imagine connecting them
with rigid rods (the rods can be rigid because the proper distances are constant),
obtaining a rigid accelerating structure which we might call an \elevator".
However, it would be misleading to call it a uniformly accelerating elevator.
Though every point on it is uniformly accelerating, the magnitude 1=X of the
uniform acceleration is di erent for di erent points. Because of this, the every-
day notion of a uniformly accelerating elevator gives a potentially misleading
physical picture. A more nearly accurate picture is obtained by thinking of
each point of the elevator as separately driven on its orbit through Minkowski
space by a tiny rocket engine. Observers moving with the elevator experience
a pseudo-gravitational force which increases without limit as the \ oor" of the
elevator at x = 0 is approached; observers nearer the oor need more powerful
rockets than those farther up.
We have two ways to view the physics of such an elevator. On the one hand,
since the elevator is a subset of Minkowski space, we can transform the well-
understood physics of Minkowski space into elevator coordinates to derive what
residents of the elevator should observe. In particular, if a particle of charge q
is situated at X = 1, say, its motion being driven by a tiny rocket attached to
it, then the energy required by the rocket per unit proper time would be the
energy required for an uncharged particle of the same mass plus the radiated
energy, the proper-time rate of radiated energy being (2=3)q2 as required by the
Larmor Law for proper acceleration 1=X = 1.
A second approach would be to emphasize the analogy of the metric (10) with
the Schwarzschild metric (7), interpreting the conserved quantity corresponding
to @ as the \energy". We want to emphasize that these two approaches are
essentially di erent and yield di erent physical predictions.
We’ll see below that the second approach (which seems similar to that of [2])
yields a conserved quantity whose integral over the \walls" of (say) a spherical
9
elevator surrounding the particle is zero. That is, there is no radiation of this
conserved quantity, which we’ll call the \pseudo-energy" to distinguish it from
the above Minkowski energy. If we interpreted this pseudo-energy as energy
radiation as seen by observers in the elevator (such as the pilot of the rocket
accelerating the charge), then by conservation of energy we should conclude
that no additional energy is required by the rocket beyond that which would be
required to accelerate an uncharged particle of the same mass.
This is a di erent physical prediction than the corresponding prediction
based on Minkowski physics, and the di erence between the two predictions is
in principle experimentally testable. It is precisely at this point that we di er
from [2]. That reference does distinguish between the Minkowski energy and
the pseudo-energy, but it gives the impression that they are somehow the same
\energy" measured in di erent coordinate systems. We think it is worth empha-
sizing that they are not the same energy measured in di erent systems; instead,
they are di erent \energies" derived from di erent Killing elds. The observa-
tion that the pseudo-energy radiation is zero does not validate the equivalence
principle.
4 Discussion of calculation of radiation
We want to brie y discuss what we think is the physically correct way to calcu-
late the energy radiated by an accelerated charge in Minkowski space. Almost
everything we shall say is well known, but we want to present it in a way which
will make manifest its applicability to the present problem. The analysis to be
given does not apply to non at spacetimes for reasons which will be mentioned
later. It applies to any simply connected subset of Minkowski space. In par-
ticular, it applies to the Rindler \elevator" described in Rindler coordinates by
the metric (10) with X > 0, and alternately as the Rindler wedge x >jtj in
Minkowski space.
Suppose we are given the worldline of a (not necessarily uniformly) acceler-
ated particle and a proper time . Surround the particle by a two-dimensional
surface S . It may be useful to think of S as a sphere, but we don’t assume any
metrical properties for S, such as rigidity. All we assume is that S surrounds
the particle.
As the particle progresses on its worldline, let S move with it in such a
way that the particle is always surrounded. As proper time progresses from an
initial value 1 to a later value 2, the surface S generates a three-dimensional
manifold S( 1; 2) in Minkowski space which is customarily called a \tube",
because it looks like a tube surrounding the worldline in a picture of Minkowski
space in which one space dimension is suppressed.
The integral of the energy-momentum tensor T = Tij over this three-
10
-space
6
time
worldline = 2 -
tube S( 1; 2) - S( 1; 2)
tube S( 1; 2) - S( 1; 2)
spacelike 3-volume inside sphere = 1 *
:
D
D
D
D
DD
B
B
BB
A
AA
D
D
D
D
DD
B
B
BB
%
%%
DD
L
AA
S
S
SS
e
ee
D
D
D
D
DD
B
B
BB
AA
E
E
E
E
E
E
EE
((((((
(
hhhhhh
h
Figure 3: Two three-dimensional tubes which coincide at their ends.
dimensional manifold will be denotedZ
S( 1; 2)
Ti dS : (11)
The precise mathematical de nition of (11) is discussed in detail in [5]. Since
the de nition entails summing vectors in di erent tangent spaces, it does not
make sense in general spacetimes, in which there is no natural identi cation of
tangent spaces at di erent points.
The intuitive meaning is that for xed i, we integrate the normal component
of the vector Ti over the tube, the integration being with respect to the natural
volume element on the tube induced from Minkowski space. Physically, (11) is
interpreted as the energy-momentum radiated through S for 1 2. The
energy radiated is (11) with i = 0.
Suppose we have two tubes, say S and S , which coincide at the initial
and nal proper times 1 and 2: S 1 = S 1 and S 2 = S 2 . Such a situation
is pictured in Figure 3, in which two space dimensions are suppressed. Taken
together, they form the boundary of a four-dimensional region, and since T has
vanishing divergence o the worldline,
Z
S( 1; 2)
Ti dS =
Z
S( 1; 2)
Ti dS : (12)
In other words, the calculated radiation is independent of the tube, so long as
the tubes coincide at their ends. Put another way, no matter how the sphere
11
distorts on its journey, (11) always produces the same numerical results for the
radiated energy-momentum.
This leads to considerable conceptual and mathematical simpli cation in
the important special case in which the particle is unaccelerated in the distant
past, put into accelerated motion for a while, and re-enters an unaccelerated
state in the distant future. We can take the ends of the tube as any convenient
geometrical shape in the rest frame of the unaccelerated particle in the distant
past (or future), say a sphere of given radius. The eld energy-momentum inside
the sphere is the in nite energy of the Coulomb eld, which is discarded in a
mass renormalization. It is unfortunate that the energy is in nite, but at least
it is well understood and can be unambiguously calculated in this special case;
this is the reason for insisting that the particle be unaccelerated in distant past
and future.
Picture the two-dimensional surrounding surface as the walls of an elevator
with the charged particle at its center. Suppose the elevator is initially at rest,
and then both elevator and particle are gently nudged into uniformly accelerated
motion, with both particle and elevator at rest in the Rindler frame (10).3 The
state of rigidly accelerated motion is then maintained for an arbitrarily long
period, after which the acceleration is gently removed and the elevator enters
a state of uniform motion thereafter. The result of the integral (11) for a
spherical elevator of initial radius is well-known. Denoting the particle’s four-
velocity at proper time as u( ), the proper acceleration as a( ) := du=d , and
a2 := a a 0, it is ([5], p. 160):
Z
S( 1; 2)
Ti dS = 23q2
Z 2
1
a2( )ui( )d + q
2
2 [u
i( 2) ui( 1)] ; (13)
where q is the particle’s charge. The last term on the right is traditionally
discarded in a mass renormalization. The energy component of the rst term is
always positive. We conclude that there is energy radiation through the walls
of the elevator.
This energy radiation can be detected in several ways in an arbitrarily small
elevator. First of all, if we believe in conservation of the usual Minkowski en-
ergy, the pilot of the rocket driving the charge will observe an additional fuel
consumption when his payload is a charged particle, relative to the correspond-
ing identical motion of an uncharged particle, the additional fuel consumption
being exactly the amount necessary to \pay" for the radiated energy. (However,
the details of how this \borrowed" energy must be repaid may be controversial,
as discussed in Appendices 1 and 2.)
A more fundamental way to meaure it, at least in principle, is to divide the
elevator walls into a large number of small coordinate patches with an observer
3It is not essential that the elevator be at rest in the Rindler frame, but this case is
particularly easy to visualize and calculate. Our argument requires only that the elevator be
initially and nally in uniform motion, and that it always surround the particle.
12
stationed on each patch. Instruct the observers to measure the elds, calcu-
late the corresponding energy-momentum tensor, and approximate to arbitrary
accuracy the energy component of the integral (11).
We want to emphasize that this is not the same as having each observer
calculate his local energy out ow n (E B=4 ) S (where n is the outward
unit normal vector to the wall in the observer’s rest frame, S the area of
his patch, the increment in his proper time, and E and B his electric and
magnetic elds, respectively), and nally adding up the total energy out ow of
all the observers. For arbitrary motion (i.e. elevator allowed to distort), this last
procedure would have no invariant meaning because each observer has his own
private rest frame at each instant of his proper time. The \energy" obtained as
the nal result of this procedure would in general depend on the construction of
the elevator. For instance, if on the same trip we had a small elevator surrounded
by a larger one, there is no reason to suppose that the observers on the larger
elevator would obtain the same number for \energy" radiation as those on the
smaller. Neither number would be expected to be related in any simple way
to the additional energy required by the rocket for a charged versus uncharged
payload.
In the procedure just described, the observers are not measuring \energy";
they are measuring something else. It may seem tempting to call it something
like \energy as measured in the (curvilinear) elevator frame", but it is conceptu-
ally and experimentally distinct from the usual Minkowski energy. For arbitrary
motion, it is not a conserved quantity and therefore probably does not deserve
the name \energy". For the special case of an elevator with constant spatial
Rindler coordinates, it does happen to be independent of the elevator’s shape (in
fact, it’s zero for all!), but it is still not \energy" as the term is normally used.
We’ll show below that it is the conserved quantity corresponding to the Killing
vector for @ ; i.e. the quantity which we previously named the \pseudo-energy".
The pseudo-energy as physically measured by the procedure just described
for a spherical elevator S of radius R in Rindler coordinates is mathematically
given by the following integral in spherical Rindler coordinates R; ; (which
bear the same relation to rectangular Rindler coordinates X;y;z that ordi-
nary spherical coordinates r; ; bear to Euclidean coordinates x;y;z). In the
integral, u = u( ;R; ; ) denotes the four-velocity of the point of the elevator
located at Rindler spherical coordinates R; ; at its proper time (i.e. Rindler
time coordinate = =X ), and n = n(R; ; ) is the spatial unit normal vector
to the sphere at the indicated point (i.e., n is orthogonal to u and normal to
the sphere, so that in Rindler coordinates, n = (0;n)):
pseudo-energy radiation =R
2
1 d
R
0 d
R2
0 d R
2 sin u T ( n ) : (14)
(The minus sign is because the spatial inner product is negative de nite.)
Recall from (5) that u = @ = @ =X, so that in Rindler coordinates in which
13
@ =: @0 is associated with the zero’th tensor index, u0 = 1=X, so u0 = X.
Hence u T n = XT0 n = XP3J=1T0JnJ. Recalling also from (4) that
= X and that K0 = X2, we may rewrite (14) in Rindler coordinates as:
pseudo-energy radiation =R
2
1 d
R
0 d
R2
0 d R
2 sin P3
J=1 K T
JnJ ; (15)
with i := i=X, i = 1;2. Equation (15) demonstrates that (14) is actually
computing the radiation of the conserved quantity corresponding to the Killing
vector K = @ .
A su cient condition for (15) to vanish is for T0J = 0 for all spatial indices
J in Rindler coordinates. Equation (IV.3), p. 185 of [2] establishes that T0J = 0
and from this draws the conclusion that:
\in the accelerated frame there is no energy ux, ... , and no radia-
tion".
That T0J = 0 is essentially the well-known \fact"4 that a stationary charged
particle in a static spacetime does not radiate energy, where \energy" is de-
ned as the conserved quantity corresponding to translation by the formal time
coordinate (in this case, ) in this spacetime.
We agree with [2] that there is no radiation of the conserved quantity cor-
responding to the Killing vector @ , but we believe that this fact is irrelevant
to questions concerning physically observed radiation and to questions about
the applicability of the Equivalence Principle. Whether it is Minkowski energy
radiation or pseudo-energy radiation which corresponds to energy that must be
furnished by the driving forces is an experimental question. In principle, it could
be settled by uniformly accelerating a large charge in a rocket and observing if
more fuel were required than for a neutral payload of the same mass. We would
bet that more fuel would be required, which would mean that Minkowski energy
is the physically relevant \energy".
On the other hand, in the Schwarzschild spacetime (7), the energy corre-
sponding to the analog @t of @ is universally accepted as the physically relevant
\energy". The spacetime (10) provides an interface between a Schwarzschild-
type spacetime and Minkowski space within which questions about the Equiva-
lence Principle can be conveniently addressed. If our hypothesis that Minkowski
energy is the physically relevant \energy" in (10) is correct, then the vanish-
ing of (15) not only does not validate the Equivalence Principle, but strongly
suggests that it does not apply to charged particles. If we treat questions of
radiation in the spacetime (10) in exactly the same way that such questions are
treated in Schwarzschild space (7), then we are led to the probably incorrect
4We put \fact" in quotes because although this assertion is often made, we know of no
proof in the literature, and in fact, it seems unlikely that it has been proved. Appendix 3
discusses this problem and furnishes a proof under certain auxiliary hypotheses.
14
conclusion that the rocket accelerating the charge does not require any extra
fuel, since there is no radiation.
The assertion that T0J = 0 for spatial indices J implies that \in the acceler-
ated frame there is ... no [energy] radiation" merits further discussion because
similar arguments are used by other authors (cf. [18]), and we believe that lan-
guage such as \energy radiation in the accelerated frame" encourages a subtle
error. The 3-vector T0J is the Poynting vector: (T01;T02;T03) = (E B)=4 ,
so that T0J = 0 says that every elevator observer sees a zero Poynting vector. If
we identify seeing a zero Poynting vector with seeing no energy radiation, then
this says that no elevator observer sees any energy radiation, which seems to
lead to the conclusion that there is no energy radiation \in the elevator frame".
Of course, one could obtain this conclusion by taking the vanishing of the
Poynting vector in the elevator frame to be the de nition of \no energy radiation
in the elevator frame", but we argue that such a de nition would be physically
inappropriate. This is the main point of this section:
Although each observer in a rigidly accelerating elevator surrounding
the particle measures a vanishing Poynting vector in his own private
rest frame, nevertheless, taken as a whole there is radiation through
the elevator walls. Adding the (zero) energy uxes measured by each
observer on the wall in his private rest frame to (incorrectly) con-
clude zero total energy radiation is an illegitimate operation because
these energy uxes refer to di erent rest frames.
5 Remarks on detecting energy radiation near
a particle
Reference [2] (unlike [8]) does recognize that Minkowski energy radiation is
nonzero but concludes that it cannot be detected within the elevator (and thence
that there is no violation of the Equivalence Principle). It discusses and dis-
cards several possible methods to observe Minkowski radiation within the eleva-
tor. For example, \if one identi es the radiation by the 1=r dependence of the
eld along the light cone, one cannot ... remain within [the region covered by
the elevator coordinates] and let r become large enough for the radiation eld
to dominate." This overlooks the fact that the eld components are analytic
functions o the worldline, and an analytic function is uniquely determined by
its values on any open set, however small. To pick o the radiation terms that
go to zero like 1=r as r!1, we need only evaluate the eld at a few points,
which can be as close to the worldline as we want, and perform a few algebraic
calculations to nd the coe cients of the 1=r terms. For example, if we write
the elds in terms of the retarded distance rret, the eld components in a given
direction from the retarded (emission) point are simple quadratic polynomials
in 1=rret, whose coe cients can be easily determined.
15
6 Conclusions
Does Einstein’s Equivalence Principle hold for charged particles? We can-
not de nitively answer this because a mathematically precise statement of the
\equivalence principle" seems elusive | most statements in the literature are
not su ciently de nite to be susceptible of proof or disproof. However, we
do conclude that most usual formulations seem not to hold in any direct and
obvious way for charged particles.
We believe that [2], which is widely cited in contexts suggesting that its
analysis supports the validity of the Equivalence Principle for charged particles,
does not in fact validate any form of the Equivalence Principle. We argue that
its conclusion that \in the accelerated frame, there is no energy ux, ... and
no radiation", is correct only if \energy" is misidenti ed (in our view) as the
conserved quantity associated with a one-parameter family of Lorentz boosts in
Minkowski space, instead of with the one-parameter family of time translations.
Appendix 1: The relation of the Lorentz-Dirac
equation to this problem
We anticipate that some readers may be uneasy about our assertion that a
uniformly accelerated charge in gravity-free (i.e., Minkowski) space may be lo-
cally distinguished from a stationary charge in (say) Schwarzschild space-time
by observing how much energy an external force, such as our fanciful rocket,
must supply to maintain the worldline. Some may observe that in Minkowski
space, the radiation reaction term in the Lorentz-Dirac equation vanishes, so one
might think that no more energy would be required in either case than would
be needed for a neutral particle of the same mass. This appendix discusses this
point, which is important but peripheral to the main text.
The Lorentz-Dirac equation[4] for a particle of mass m and charge q in an
external eld F = Fij is:
mdud = qF(u) + 23q2
da
d +a
2u
; (16)
where is proper time, u = ui the particle’s four-velocity, a := du=d its proper
acceleration, a2 := a a and F(u)i := Fi u .
The left side is the rate of change of mechanical energy-momentum, the term
qF(u) is the Lorentz force, and the remaining term (2=3)q2[da=d +a2u] is the
\radiation reaction" term which describes the e ect of the particle’s radiation
on its motion.
For a uniformly accelerated particle (i.e., a2 is constant) moving in one space
dimension, the radiation reaction term (2=3)q2[da=d + a2u] vanishes identi-
16
cally.5 It is tempting to interpet this as implying that there is no physical
radiation reaction for a uniformly accelerated charged particle, by which we
mean that a rocket-driven uniformly accelerated charge requires no more en-
ergy from the rocket than an otherwise identical neutral charge. However, we
believe such an interpretation is unlikely to be correct.
An obvious aw in the argument just given is that it is inconsistent with
usual ideas of conservation of energy. If we grant that the uniformly accelerated
charge does radiate energy into Minkowski space which can be collected and
used, as nearly all modern authors ([8] and [21] excepted) seem to agree, then
this radiated energy must be furnished by some decrease in energy of other
parts of the system. Fulton and Rohrlich [15] suggest that it may somehow
come from the eld energy but give no proof. (Since the eld energy in a
spacelike hyperplane is in nite for a point electron, it’s not clear what would
constitute a proof.)
We look at the matter di erently. All derivations of the Lorentz-Dirac equa-
tion are motivated by conservation of energy-momentum: the change of energy-
momentum of the particle over a given proper-time interval should equal the
energy-momentum furnished by the external forces driving the particle minus
the radiated energy-momentum, assuming that it is legitimate to absorb in nite
terms of a certain structure into a mass renormalization. Although this principle
motivates the derivation, the nal equation unfortunately does not guarantee
such conservation of energy-momentum in general, but only in certain special
cases. One such special case is when the particle is asymptotically free, mean-
ing that its proper acceleration vanishes asymptotically in the in nite past and
future.
Thus it’s not clear that the equation should apply to a particle which is
not asymptotically free, such as a particle which is uniformly accelerated for all
time.6 Since the equation doesn’t guarantee conservation of energy-momentum
for uniform acceleration for all time, the fact that the radiation reaction term
vanishes implies nothing about the additional force which the rocket must fur-
nish for perpetually uniformly accelerated motion.
But we should at least try to understand the case of a particle which is
unaccelerated in the distant past, nudged into uniform acceleration, uniformly
accelerated for a long time, and nally nudged back into an unaccelerated state.
For this case, the Lorentz-Dirac equation does imply conservation of energy-
momentum. However, since the radiation reaction term vanishes for the period
5 To see this, write u = ( ;v ;0;0) with v the velocity and := (1 v2) 1=2, and let
w := (v ; ;0;0) be an orthogonal unit vector associated with the same spatial direction. By
general principles, the proper accelerationa is orthogonal tou, so thata = Aw for some scalar
function A, and A2 = a2 is constant. Since w is a unit vector, dw=d is orthogonal to w,
and hence da=d = Adw=d is a multiple of u. That the multiple is a2 can be determined
by taking the inner product u da =d = d(u a )=d a du =d = a2.
6 Actually, the equation is controversial even for asymptotically free particles, but that
brings up issues outside the scope of this article. The reader can nd more information in
[16], [5], [6], [17].
17
of uniform acceleration, the equation implies that all the radiation energy must
be furnished at the beginning and ending of the trip, while the particle is nudged
into or out of its uniformly accelerated state.
In other words, if we believe in the Lorentz-Dirac equation, we need to add
a bit of energy to start the uniform acceleration, and thereafter the radiation,
which can persist for an arbitrarily long time and add up to an arbitrarily
large amount, is \free" until the end of the trip. In e ect, we can \borrow" an
arbitrarily large amount of radiated energy (which in principle can meanwhile
be collected and used by other observers in Minkowski space), so long as we
pay it back at the end of the trip. Although there is no logical contradiction
here, this is hard to accept physically, and seems one of many good reasons to
question the Lorentz-Dirac equation.
Most of the above issues are only peripherally relevant to the present work,
and we present them only to dispel potential confusion. The point is that
the vanishing of the radiation reaction term does not imply that the rocket
accelerating the charged particle in Minkowski space does not have to furnish the
radiation energy. The rocket almost certainly does have to supply this energy,
and this gives a local experiment which distinguishes certain accelerated motion
in Minkowski space from similar motion in Schwarzschild space.
Appendix 2: The equation of motion of a charged
rocket
A noted expert in the eld raised the following interesting objection to the dis-
cussion of Appendix 1 in an earlier version of this paper. Consider a charged
rocket which undergoes a modest uniform acceleration g (one gravity, say) from
just after an inital time i to just before a nal time f. More precisely, the
rocket is at rest in some Lorentz frame (the initial frame) up to some initial
proper time i, nudged into uniform acceleration over a small proper time inter-
val [ i; i+ ], uniformly accelerated up to proper time = f , nudged back
into an unaccelerated state over the interval [ f ; f], to remain unaccelerated
for > f.
He presented a simple estimate showing that the energy required to accom-
plish the nal deceleration, as measured in the nal rest frame at = f, is
modest and independent of the period of uniform acceleration. This can be an-
ticipated without calculation, since from the viewpoint of the nal rest frame,
going backwards in time from = f to = f only requires nudging the
rocket back up to a modest uniform acceleration, and this cannot not require
an unbounded energy change. Thus it would seem that from the point of view
of the rocket’s pilot, only a modest amount of fuel must be burned to start the
acceleration at the beginning of the trip and stop it at the end, with no excess
fuel (relative to an uncharged rocket) required during the period of uniform
18
x
t
ti = 0
tf - - - - - - - - - -
- - - - - - - - - - - - -2tf
Figure 4: The worldline of a particle at rest up to time ti and uniformly accel-
erated from time ti + to tf , where is the length of a small time interval
during which the particle is nudged into or out of uniform acceleration. At
time tf the acceleration is reversed in a time-symmetric way so as to bring the
particle back to rest at time 2tf.
acceleration (which can be arbitrarily long).7
The modest amount of energy used in the nal frame (along with its asso-
ciated momentum) can Lorentz-transform into a large amount of energy in the
initial rest frame at = i, so there is no apparent violation of conservation of
energy from the standpoint of the initial frame. However, we can obtain what
might appear to be a violation if we imagine reversing the proper acceleration
a at the nal time f in a time-symmetric way (i.e., a( f + ) = a( f ))
to eventually bring the rocket back to rest in the initial frame at = 2 f, as
depicted in Figure 4. The expert’s estimate shows that the excess fuel used over
the entire trip from = i to = 2 f is modest and independent of the duration
of the uniform acceleration. At the beginning and end of the trip the rocket is at
rest in the initial frame, so the energy of the radiation plus the exhaust should
7This was produced in evidence for the widely held belief (which we think incorrect) that
there is no radiation reaction for a uniformly accelerated charge in Minkowski space. This
line of reasoning suggests that we could allow the uniform acceleration to continue inde nitely
without using any more fuel. (By extension, perpetual uniform acceleration would presumably
require no fuel at all.) That would violate conservation of energy, assuming that the radiation
energy is physically accessible, but proponents of this view sometimes use arguments similar
to those criticized in Section 5 to assert that radiation cannot be observed within a Rindler
elevator.
19
equal the rest-mass loss of the rocket (fuel used). If the loss of rest mass is nite
and independent of the duration of the uniform acceleration (hence independent
of the arbitrarily large energy radiation), we have a violation of conservation of
energy in the initial frame.
The situation was clari ed by actually solving the equation of motion for
the radiating rocket, using the Lorentz-Dirac radiation reaction expression. It
turns out that with a xed amount of initial fuel, one cannot obtain an arbi-
trarily large period of uniformly accelerated motion (i.e., arbitrarily large f)
unless one allows the rocket mass to go negative. Put another way, a charged
rocket will run out of fuel if it uniformly accelerates long enough, so our time-
symmetric motion is impossible with xed initial fuel and arbitrarily large f.
This is in contradistinction to a uniformly accelerated uncharged rocket which
can accelerate forever, assuming that all of its mass can be used as fuel.
In retrospect, this conclusion seems natural and the analysis leading to it
elementary, but I found the expert’s objection su ciently troubling to feel it
necessary to actually work it out. Having done so, perhaps including it here
may save readers with similar questions some work.
The rocket will always move in the positive x-direction, and the other two
constant space coordinates will be suppressed. If its initial-frame velocity is v,
its initial-frame \rapidity" is de ned by := tanh 1(v). Then its four-velocity
u is given in initial-frame coordinates by
u = (cosh ;sinh ) :
The scalar proper acceleration A is de ned by du=d = Aw where w is the unit
vector
w := (sinh ;cosh )
orthogonal to u. The (four-vector) proper acceleration is a := Aw.
The scalar proper acceleration is related to the rapidity by A = d =d . In
particular for constant scalar proper acceleration A( ) g, we have ( ) =
g + (0).
The Lorentz-Dirac expression for the proper-time rate of energy-momentum
radiation of a charge q is:
rate of energy-momentum radiation = (2q2=3)(da=d + a2u)
= (2q2=3)(dA=d )w : (17)
The second line follows from the rst in a fashion similar to that of the rst
footnote in Appendix 1. We eliminate the constant factor by choosing units so
that 2q2=3 = 1.
Let m( ) denote the rocket’s rest mass, so that dm=d is the rate of ejection
of rest mass into the exhaust. (This is not the same as the rate at which the
exhaust acquires rest mass, as will be apparent from the expressions to follow.
Rest mass is not conserved in general.)
20
There are two parameters which can be used to control the rocket’s worldline:
the exhaust velocity and the rate dm=d of ejection of rest mass into the
exhaust. The analysis to follow assumes that the exhaust velocity as seen from
the rocket is always constant and that dm=d is varied so as to produce the
desired worldline. The rocket is always moving to the right in the initial frame,
so the exhaust is always moving left. We allow dm=d to have either sign. A
positive dm=d means that the rocket is taking on mass. This could physically
be accomplished by shooting bullets into it from the right, with the bullets
constituting the \exhaust".
Let denote the exhaust rapidity in the rocket’s instantaneous rest frame.
That is, is positive, and an exhaust particle has velocity tanh( ) in the this
rest frame. Then the exhaust’s four-velocity is:
four-velocity of exhaust = ucosh wsinh : (18)
Let = ( ) denote the proper-time rate at which the exhaust rest mass
is increasing. Let Rw = R( )w( ( )) denote the proper-time rate at which
electromagnetic energy-momentum is being emitted. It is assumed that this rate
is a multiple of w because the above Lorentz-Dirac expression is of this form.
It is convenient to allow arbitrary R because this enables us to simultaneously
treat the case of an uncharged rocket by setting R 0. The uncharged case is
also worked out in [19], in a similar fashion with identical results.
The equation of energy-momentum balance is:
0 = d(mu)d + (ucosh wsinh ) + Rw
= (dmd + cosh )u+ (mA sinh + R)w : (19)
The rst term in the rst line is the proper-time rate of change of energy-
momentum of the rocket, the second term the proper-time rate at which the
exhaust is acquiring energy-momentum, and the third the proper-time rate of
energy-momentum radiation.
Since u and w are orthogonal, the last line separates into two independent
equations:
= 1cosh dmd ; (20)
and
mA sinh + R = 0 : (21)
Equation (20) may be regarded as de ning , and then (21) becomes, setting
:= 1=tanh :
dm
d + mA + R = 0 : (22)
21
Recalling that A = d =d , we can immediately write down the solution with
zero initial rapidity in terms of :
m( ) = e ( )m( i) e ( )
Z
i
e (s)R(s)ds : (23)
For R 0, corresponding to an uncharged rocket, we see that m( ) decreases
exponentially with ( ). It also decreases exponentially with during the rst
period of uniformly accelerated motion, since in that period, ( ) = ( i
)g + ( i + ). In particular, m can never vanish for an uncharged rocket. An
uncharged rocket can uniformly accelerate forever, assuming that all of its rest
mass can be used as fuel.
Now consider a charged rocket with R given by the Lorentz-Dirac expression
R( ) := dA=d . Then (23) becomes:
m( ) = e ( )m( i) e ( )
Z
i
e (s)
dAds
ds : (24)
To dispel the notion that the charged rocket can uniformly accelerate for an
arbitrarily long period without using any more fuel than would an uncharged
rocket, we want to show that if A decreases monotonically from a constant value
g down to 0 over a nal proper-time interval [ f ; f] of xed length , then
m( f) must become negative for large f. That is, for such an A and for a xed
initial mass m( i), we cannot nd positive-mass solutions de ned for arbitrarily
large proper times f. This can be seen from the following simple estimates, in
which it is helpful to remember that both and dA=ds are positive.8
First observe that from the Mean Value Theorem, for f s f,
( f) (s)
( f) (s)
f s
= d d (^ ) for some ^ with s ^ f
= A(^ )
g :
Using this, we have:
e ( f)
Z f
f
e (s)
dAds
ds =
Z f
f
e ( ( f) (s))
dAds
ds
e g
Z f
f
dAds
ds ;
8That dA=ds is positive follows from the previous assumption, made for simplicity, that
A decreases monotonically from g to 0. If we agree to eject mass at a positive rate (i.e.,
dm=d < 0) until A = 0, and if we de ne f to be the rst time after deceleration that A = 0,
then this assumption follows from (22) with R := dA=d .
22
= e g ( A( f) + A( f )) ;
= e g g :
(25)
Substituting (25) in (24), we see that to obtain a positive mass solution for
arbitrarily large f (and arbitrarily large radiated energy), we need arbitrarily
great rocket mass (i.e., fuel) m( i) to start with.
In other words, a charged rocket in Minkowski space which starts with a nite
amount of fuel cannot uniformly accelerate for an arbitrarily long time, after
which the acceleration is removed. Unlike a corresponding uncharged rocket,
it must eventually run out of fuel. What is peculiar is that if it has su cient
fuel to get into the uniformly accelerated state, it will not run out of fuel until
after the uniform acceleration is removed! It can uniformly accelerate for an
arbitrarily long period, radiating all the while, but the physical contradiction
of running out of fuel followed by the mass going negative will not be revealed
until after the uniform acceleration is removed.
We emphasize that this is a rigorous mathematical conclusion from the given
assumptions|there are no approximations in the analysis which led to it. Phys-
ically, it is very hard to believe. The most questionable assumption seems to be
the Lorentz-Dirac expression (17) for the radiated energy.
It is enlightening to follow the solution further to the nal resting state at
= 2 f, but before doing this let’s think about what we would expect for
an uncharged rocket. Since our formulation assumes that the exhaust velocity
cannot be varied, the deceleration after = f is accomplished by taking in
mass (and momentum), so we will have dm=d > 0 for f < < 2 f . In
e ect, deceleration is accomplished by returning some of the previous exhaust
energy-momentum to the rocket. For an uncharged rocket, the symmetry of the
situation suggests that this energy-momentum return will be accomplished in
time-symmetric fashion, and we can anticipate without calculation that all the
exhaust energy-momentum will have been returned to the rocket at the nal
resting time = 2 f. In particular, the nal rest mass should be the same as
the initial rest mass. Indeed, this is what equation (24) does give if the radiation
term dA=d is omitted.
However, the result is quite di erent for the charged rocket. In this case,
m(2 f) di ers from m( i) by the amount of the second term containing the
integral. We have (2 f) = 0, so the exponential factor in front of the integral
doesn’t contribute. The mass de cit at the end is
m( i) m(2 f) =
Z i+
i
e (s)dAd ds
Z f+
f
e (s)dAd ds
Z 2 f
2 f
e (s)dAd ds:
The rst and third integrals are of moderate size, while the second integral over
the interval [ f ; f + ] is large for large f because e is large on this
23
interval. In e ect, the large initial-frame energy furnished over [ f ; f +
] (corresponding to a small loss of rest mass at f with high initial-
frame velocity) has been transferred to the same large energy loss caused by a
correspondingly large initial-frame rest mass loss.
To put it more physically, by observing his fuel gauge, the charged rocket
pilot sees only a modest excess fuel loss over [ f ; f + ] (relative to an
uncharged rocket), but he does observe this loss, and he can gure out that
because he is going very fast in the initial frame, it corresponds to a large
initial-frame energy loss. Moreover, as he decelerates back to rest at = 2 f,
this modest rest mass loss grows exponentially to an initial-frame excess rest
mass loss large enough to pay for the radiated energy.
This last observation may sound strange, but properly viewed it is to be
expected. The rest mass of an uncharged rocket, will increase exponentially
during the period [ f + ;2 f ] of uniform deceleration, and the same is true
of the charged rocket. Over this period, the charged rocket behaves identically
to an uncharged rocket with the same rest mass at = f + . However, the
uncharged rocket which started with initial rest mass mi at = 0 does not
have exactly the same rest mass at f + as the charged rocket with the same
worldline and initial mass. There is a di erence due to the radiation in the time
interval [0; f + ]. This di erence is modest even when the radiation is large. If
f is large enough to give large radiation, this modest di erence in rest masses
at = f + is ampli ed by the exponential growth to a correspondingly large
di erence in rest masses at = 2 f .
This analysis provides additional insight into the discussion of Appendix 1. It
demonstrates by explicit calculation that contrary to widely held beliefs, there is
indeed physical radiation reaction for a particle which is uniformly accelerated
for a nite time even though the Lorentz-Dirac radiation reaction expression
vanishes identically during the period of uniform acceleration.9 However, if
we believe in the Lorentz-Dirac equation (and many experts don’t), we must
accept the very strange conclusion that all of this radiation reaction occurs at
the beginning (t ti) and end (t tf) of the trip while the particle is being
nudged into or out of its uniformly accelerated state.
9Whether there is radiation reaction for a perpetually uniformly accelerated particle de-
pends on one’s de nition of \radiationreaction". The \radiation reaction" term in the Lorentz-
Dirac equation does vanish identically, but there is no good physical reason to identify this
term with physically observed radiation reaction. Instead, it seems more reasonable to obtain
the answer for uniform acceleration for all time as a limit of whatever answer is eventually
generally accepted for uniform acceleration for nite times. There is probably no reasonable
way to do this without rejecting the Lorentz-Dirac equation, since the above answer for uni-
form acceleration for nite times (which is a consequence of the Lorentz-Dirac equation) is so
strange.
24
Appendix 3: The eld of a stationary particle in
a static spacetime
It is often stated in the literature (e.g., [2]) that a charged particle which is
stationary with respect to the coordinate frame in a static spacetime generates
a pure electric eld in that frame; since the Poynting vector vanishes, there is
no radiation. However, we know of no proof in the literature, and the matter
seems to us not as simple as it apparently does to the authors who make this
assertion.
Implicit in such statements is that the eld generated by the particle is
the \retarded eld" for its worldline. The problem is that there is no gener-
ally accepted, mathematically rigorous de nition of \retarded eld" in general
spacetimes. In Minkowski space one can de ne the retarded eld via the usual
explicit formula, but no similar closed-form expressions are known for general
spacetimes.
A retarded- eld construction should be a rule which assigns to each charged
particle worldline 7! z( ) (de ned as a curve in spacetime with unit-norm
tangent u( ) := dz=d ) a 2-form F = F(x) satisfying Maxwell’s equations with
source the distribution current associated with the worldline. Symbolically,
these equations are
dF = 0
( d F)(x) =
Z
(x z( ))qu[( )d ;
where d is the di erential operator on alternating forms, the Hodge duality op-
eration, the four-dimensional Dirac delta distribution, q the particle’s charge,
and u[ the 1-form corresponding to u (see below).
To make the eld \retarded", it is also required that the value of F(x) at
any spacetime point x o the worldline should depend only on the part of the
worldline on or within the backward light cone with vertex x. In other words,
any two worldlines which are identical inside this cone should yield the same
F(x).
Other assumptions might also reasonably be imposed. For example, one
expects that for x o the worldline, the components of F(x) would be an or-
dinary in nitely di erentiable 2-form (a priori it is only a distribution). This
assumption is not necessary for our purposes, but it does no harm and simpli es
thought. One very plausible assumption which we shall need is that in a static
spacetime, the retarded eld for a stationary particle is time-independent.
Unfortunately, no mathematically rigorous retarded- eld construction seems
to be known for general spacetimes or even for static spacetimes. The discussion
of Section 5.6 of [20], p. 220 gives the avor of the mathematical di culties.
Despite the lack of rigorous mathematical proof, most physicists seem pre-
pared to believe that in any given spacetime, a unique retarded- eld construc-
25
tion with the above properties ought to exist. Under this meta-mathematical
assumption, we can show that the retarded eld of a stationary particle in a
static spacetime (6) is a pure electric eld and that consequently the particle
does not radiate. More precisely, there is no radiation through a stationary
closed surface (stationary with respect to the \static" coordinates of (6)) sur-
rounding the particle.
The idea is very simple. Given a retarded eld, we can project out the
electric part of it (relative to the static coordinates), and this projected electric
part will still be \retarded". It is not obvious that it will satisfy Maxwell’s
equations (with the particle’s worldline as source as above), but we shall show
that it does. It follows that the electric part is also a retarded eld.
If we believe in the uniqueness of the retarded eld construction for the
given spacetime, then this shows that the original retarded eld was already a
pure electric eld. If we are not willing to make the uniqueness assumption,
then at least we have shown that there exists a retarded eld construction for
static spacetimes for which the retarded eld is pure electric and the particle
does not radiate. If the retarded eld construction is not unique, then we need
additional physics to select the physically relevant retarded eld in order to
answer the question of whether a stationary charge radiates.
Now we prove the above assertion that the pure electric part of the retarded
eld for a stationary particle in a static spacetime is itself a solution of the above
Maxwell’s equations. As mentioned above, we assume that the retarded eld is
time-independent, and this is the only use of the \retarded eld " assumptions.
Thus we are really proving that the pure electric part of a time-independent
solution is itself a solution.
Consider a particle stationary at the origin in a spacetime with the static
metric (6). The four-velocity of the particle will be denoted u(= ui), and the
corresponding index-lowered one-form as u[(= ui := gi u ). Explicitly, u =
g 1=200 @x0, and u[ = g1=200 dx0. Suppose we have a time-independent distribution
2-form F = Fij satisfying the Maxwell equations
dF = 0
d F = 3u[ ; (26)
where denotes the Hodge duality operation, d the di erential operator on al-
ternating forms, and 3(x;y;z) := (x) (y) (z) is the three-dimensional Dirac
delta distribution.10 Time-independence means that the coe cients Fij =
Fij(x1;x2;x3) do not depend on the coordinate time x0.
We may uniquely write
F = E[^u[ + ; (27)
10De nitions of the di erential-geometric quantities such as the Hodge dual can be found in
[5], Chapter 2. The proof can be given within the rigorous framework of distribution theory,
but we write it in the traditional physics language of Dirac delta \functions".
26
where E = P3I=1EI@xI is a purely spatial vector eld, E[i := gi E the cor-
responding index-lowered 1-form, and is a purely spatial 2-form. (We use
bold-face for vectors in 4-space which are purely spatial with respect to the
coordinate system used in (6), and we generally use capital Roman letters for
space indices. All index lowering and raising is with respect to the spacetime
metric rather than the Euclidean 3-space metric.) To say that is purely spatial
means that
=
3X
I;J=1
IJdxIdxJ :
Physically, is the 3-space Hodge dual of the 1-form corresponding to the
magnetic eld vector B. The proof of (27) follows routinely from expanding F
as a linear combination of dx ^dx , noting that u is proportional to dx0, and
collecting terms involving u.
We shall now show that if F satis es (26), then the electric part E[^u[ of F
also satis es (26).
(a) Consider the rst Maxwell equation 0 = dF = d(E[^u) + d . We want
to show that d(E[^u) = 0. By routine calculation (directly, or cf. [5] ,
Section 5.4),
du[ = u[^a[;
where a := du=d is the acceleration of a stationary observer. Hence
d(E[^u[) = d(u[^E[) = du[^E[ + u[^dE[
= u[^(E[^a[ + dE[) : (28)
The point is that d(E[^u[) = u[^(something) and hence is orthogonal
to any purely spatial 3-form (with respect to the inner product on two-
forms induced by the spacetime metric). On the other hand, d is a purely
spatial 3-form because its coe cients are time-independent by assumption.
Hence d(E[^u[) and d must separately vanish.
(b) Now we consider the other Maxwell equation d F = 3u[ and try to
prove that this can happen only if d = 0. The 2-form is purely
spatial, so its Hodge dual = u[^S[ for some purely spatial vector S.
Apply the argument of part (a) with S in place of E to conclude that
d = u[^(something). Now take a Hodge dual to see that d is a
purely spatial 1-form; i.e. d is the 1-form corresponding (under index-
raising) to a vector orthogonal to u.
On the other hand, (E[^u[) is purely spatial with time-independent co-
e cients, hence d (E[^u[) is a purely spatial 3-form, hence d (E[^u[) is
a multiple of u[. Thus we have
3u[ = d (E[^u[) + d
27
with the rst term on the right a multiple of u[ and the second term
orthogonal to u[; this can happen only if d (E[^u[) = 3u[ and d = 0.
This completes the proof that E[^u[ satis es the Maxwell equations (26).
However, the eld E is not usually a Coulomb eld, contrary to impressions
given by [2] and other authors.11 To see that E is not necessarily a Coulomb
eld, consider a metric of the special form (8), for which (9) gives the acceleration
as a = (c0=c)@x6= 0. A Coulomb eld
C := (x@x + y@y +z@z)=(x2 + y2 +z2)3=2
would satisfy dC[ = 0 except at the spatial origin (i.e. r C = 0 in 3-space),
but this is inconsistent with the vanishing of (28) because
u[^C[^a[ = (c0=c2)dt^dx^C[6= 0 : (29)
Finally, we note that with F := E[^u[, the energy-momentum tensor (1) has
T0J = 0 for J = 1;2;3, which says that the Poynting vector vanishes and there
is no radiation through any stationary closed surface surrounding the particle.
This was worked out in [2] for the metric (8), and [8] obtains a special case of
the same result in di erent language.
Appendix 4: Later references
This body of this paper was originally posted in the Internet archive
www.arXiv.org/abs/gr-qc/9303025 in March, 1993, in response to a discussion
in the Internet newsgroup sci.physics concerning the applicability of the Equiv-
alence Principle to charged particles. Appendices 2 and 3 were added later in
response to questions from readers.
Appendix 2 was posted in July, 1994. In the interim, minor errors in the
body of the paper have been corrected, and minor stylistic changes made. A
minor revision with the addition of Appendix 3 and references to Singal’s 1995
paper [8] was posted in January, 1996. The present Appendix 4 comments on
work published since 1994 and bearing on the substance of this paper.
It was noted above that Singal’s 1995 paper [8] in General Relativity and
Gravitation (GRG) expresses a view opposite to that of the present work: he
believes that a perpetually uniformly accelerated charged particle would not
radiate. In response, our 1997 GRG paper [7] pointed out that Singal’s un-
usual method implies that a charged particle uniformly accelerated for only a
11 This is probably more a question of language than of substance. For instance, although
[2] states on page 172 that for the metric (10), \the accelerated observer ... only detects
a Coulomb eld", the expressions derived for the eld are not precisely Coulomb elds in
either the accelerated or Minkowski frames. Probably what was meant was something like
\Coulomb-type" eld.
28
nite time (as in Appendix 2) does radiate, in quantitative accordance with the
Larmor law.
Thus the answer given by Singal’s method for uniform acceleration for all
time is di erent than the answer which would be obtained from the same method
by calculating radiation from uniform acceleration over a long but nite time ,
and then taking a limit as !1. This suggests that the question of radiation
for a perpetually accelerated particle may be too singular for traditional math-
ematical analysis. This is because two reasonable methods, whose mathematics
are unchallenged, lead to di erent conclusions. For details, see [7].
In the same issue of GRG in which [7] appeared, Singal published a sequel
[9] entitled \The Equivalence Principle and an Electric Charge in a Gravita-
tional Field II. A Uniformly Accelerated Charge Does Not Radiate". Although
[7] and [9] happened to appear simultaneously in GRG, the two authors were
corresponding and were familiar with each others’ work during the acceptance
process. As its title suggests, Singal was unconvinced by [7], though he has not
challenged its mathematics. Singal’s sequel [9] comments on various aspects of
the problem, but does not address the analysis of the comment [7] on his original
paper [8].
It should be emphasized that our disagreement with Singal’s work, is solely
a matter of de nition. Our [7] does not question the mathematics of [8] (which
indeed we have checked carefully and believe correct), and [9] does not question
[7].
In 1999, Shariati and Khorrami published [21]. This work comes to conclu-
sions opposite to those of the present paper (which is cited, but not discussed
in detail). They de ne a \supported" observer as one whose worldline is that
of a point on a Rindler \elevator" in Minkowski space (i.e., a point stationary
in the Rindler frame (10)), while an unaccelerated observer is \freely falling".
They conclude:
\A supported charge does not radiate according to another sup-
ported observer."
They identify the truth of this statement with the mathematical fact that any
supported observer sees a vanishing Poynting vector12 (except in the singular
case when the observer coincides with the charge). This is precisely Boulware’s
argument quoted above in Section 4, that
\in the accelerated frame there is no energy ux, ... , and no radia-
tion".
Since regarding this point, Shariati and Khorrami’s analysis is essentially that of
Boulware, our concluding remarks for Section 4 apply verbatim to their analysis
also:
12Their argument that the Poynting vector vanishes for stationary observers in a static
spacetime (of which \supported" observers in a Rindler elevator is an instance) is basically
that of Appendix 3, expressed in more traditional notation.
29
\Of course, one could obtain this conclusion [of no energy radiation]
by taking the vanishing of the Poynting vector in the elevator frame
to be the de nition of ‘no energy radiation in the elevator frame’, but
we argue that such a de nition would be physically inappropriate."
We feel that the main contribution of the present paper was pointing out
that Boulware’s conclusion rested on the hidden assumption of this de nition,
which we regard as unlikely. It is not clear whether Shariati and Khorrami
recognize this assumption; if they do, evidently they disagree that it is unlikely.
Thus mathematically, the issue reduces to a question of de nition.
The physically correct de nition could be determined by doing experiments
discused above, such as comparing the fuel consumption of a uniformly acceler-
ating charged rocket with that of an uncharged rocket. Shariati and Khorrami
state that they believe the fuel consumptions would be identical:
\In the previous section, it was shown that a uniformly acceler-
ated charge in a Minkowski spacetime does not radiate, in the sense
that for the Rindler observer the Poynting vector vanishes, and an
energy-like quantity for the electromagnetic eld is constant. This
means that, according to Rindler observers, no extra force is needed
to maintain the uniform acceleration of such a charged particle (of
course no extra force beside the force needed for a neutral parti-
cle of the same mass to have that acceleration). In other words,
the world-line of the charged particle will be the same as that of a
neutral particle."
No one can prove or disprove this assertion without doing the experiment. Their
\energy-like quantity" is what we called the \pseudo-energy".
Shariati and Khorrami [21] also maintain that
\ ::: in a static spacetime, :::, a freely falling charge do[es] not
radiate, in the sense that no extra force is needed to maintain [its]
world-line the same as that of a neutral particle."
This is an issue not addressed in the present work, but those interested in the
analysis of Shariati and Khorrami should be aware that it may be inconsistent
with the DeWitt/Brehme/Hobbs (DBH) equation.
The DBH equation is a generalization of the Lorentz-Dirac equation to ar-
bitrary spacetimes. A similar equation was originally derived by DeWitt and
Brehme [3], but an error in the very complicated derivation eliminated impor-
tant terms. The error was corrected eight years later by Hobbs [14].
Shariati and Khorrami assume without proof a di erent equation of motion
(their equation (20)) for charged particles. It di ers from the DBH equation by
the omission of terms involving the Ricci tensor, and also omission of a so-called
nonlocal "tail" term. They observe that their equation implies that a freely
30
falling charge (i.e., a charge with zero acceleration) experiences no radiation
reaction.
This conclusion does not seem to follow from the DBH equation. A particle
satisfying the DBH equation can fall freely only if some mathematical miracle
causes the terms omitted by the Shariati/Khorrami equation to vanish. Shariati
and Khorrami do not discuss this issue; indeed, they do not mention the DBH
equation.13
Acknowledgment
Because [2] is so widely cited (always favorably, to my knowledge), I chose to
focus on it as representative of a point of view with which I have come to
disagree. The scienti c criticisms presented are in no way intended to denigrate
its important contributions to our understanding of these issues. One such
contribution is recognizing the role of the metric (10) as an interface between
the physics of Minkowski space and the physics of more general spacetimes
within which they can conveniently be compared. The present paper is deeply
indebted to this advance.
References
[1] W. B. Bonnor, A new equation of motion for a radiating charged particle,
Proc. Roy. Soc. London A337, 591-598 (1974)
[2] D. Boulware, \Radiation from a uniformly accelerated charge", Annals of
Physics 124, 169-187 (1980)
[3] B. DeWitt, and R. Brehme, \Radiation damping in a gravitational eld",
Annals of Physics 9, 220-259 (1960)
[4] P. A. M. Dirac, \Classical theories of radiating electrons", Proc. Royal Soc.
London A167, 148 . (1938)
[5] S. Parrott, Relativistic Electrodynamics and Di erential Geometry,
Springer, New York, 1987
13This should not be interpretedas our endorsement of the DBH equation. Like the proposed
equation of Shariati and Khorrami, it reduces to the Lorentz-Dirac equation in at spacetime,
and hence implies the usual zoo of bizarre predictions of that equation, such as those worked
out in Appendix 2. But those who do believe in the Lorentz-Dirac equation typically also
believe in the DBH equation. It seems strange that Shariati and Khorrami base their analysis
on yet another variant of the Lorentz-Dirac equation, which may not have been proposed
before, and certainly is not widely accepted.
31
[6] S. Parrott, \Unphysical and physical(?) solutions of the Lorentz-Dirac
equation", Foundations of Physics 23, 1093-1121 (1993)
[7] S. Parrott, \Radiation from a particle uniformly accelerated for all time",
General Relativity and Gravitation 27 1463-1472 (1995), gr-qc/9711027
[8] A. K. Singal, \The equivalence principle and an electric charge in a gravi-
tational eld", General Relativity and Gravitation 27 953-967 (1995)
[9] A. K. Singal, \The Equivalence Principle and an Electric Charge in a Grav-
itational Field II. A Uniformly Accelerated Charge Does Not Radiate",
General Relativity and Gravitation 27 1371-1390 (1997)
[10] J. D. Jackson, Classical Electrodynamics, Second Edition, Wiley, New York,
1975
[11] R. Peierls, Surprises in Theoretical Physics, Princeton University Press,
Princeton, NJ, 1979
[12] R. Sachs, and H. Wu, General Relativity for Mathematicians, Springer,
New York, 1977
[13] W. Rindler, Essential Relativity, 2nd Edition, Springer, New York, 1977
[14] J. M. Hobbs, \A vierbein formalism of radiation damping", Annals of
Physics 47, 141-165 (1968)
[15] T. Fulton, and F. Rohrlich, \Classical radiation from a uniformly acceler-
ated charge", Annals of Physics 9, 499-517 (1960)
[16] C. J. Eliezer, \The hydrogen atom and the classical theory of radiation",
Proc. Camb. Phil. Soc. 39, 173 (1943)
[17] S. F. Gull, \Charged particles at potential steps", in The Electron: new
theory and experiment, eds. D. Hestenes, and A. Weingartshofer, Kl uwer,
Dordrecht, 1991
[18] A. Kovetz, and G. Tauber, \Radiation from an accelerated charge and the
principle of equivalence", Amer. J. Phys. 37, 382-385 (1969)
[19] E. Taylor, and J. Wheeler, Spacetime Physics, Freeman, New York, 1966,
pp. 141-143
[20] F. G. Friedlander, The Wave Equation on Curved Spacetime, Cambridge
University Press, Cambridge, 1976
[21] A. Shariati, and M. Khorrami, \Equivalence Principle and Radiation by a
Uniformly Accelerated Charge", Found. Phys. Lett. 12 427-439 (1999)