Classical and Quantum Radiation Reaction For Linear Acceleration

Foundations of Physics, Vol. 35, No.
7, July 2005 ( 2005)

DOI: 10.1007/s10701-005-6405-0
Classical and Quantum Radiation Reaction

for Linear Acceleration
Atsushi Higuchi1, 2 and Giles D. R. Martin1

Received April 11, 2005
We investigate the effect of radiation reaction on the motion of a wave packet

of a charged scalar particle linearly accelerated in quantum electrodynamics
(QED). We give the details of the calculations for the case where the particle
is accelerated by a static potential that were outlined in Higuchi and Martin
Phys. Rev. D 70 (2004) 081701(R) and present similar results in the case of
a time-dependent but space-independent potential. In particular, we calculate the
expectation value of the position of the charged particle after the acceleration,
to rst-order in the ne structure constant in the
h 0 limit, and nd that the
change in the expectation value of the position (the position shift) due to radi-
ation reaction agrees exactly with the result obtained using the LorentzDirac
force in classical electrodynamics for both potentials.
KEY WORDS: radiation reaction; quantum electrodynamics; LorentzDirac;

semiclassical approximation.
1. INTRODUCTION
A charged particle radiates when it is accelerated. This simple discovery

has been the basis of a multitude of technological advances since. Most
work, however, has concentrated on the radiation itself. The effect of the
radiation, which carries energy and momentum, on the particle itself is
very small but still present. The overall effect is a change in the equa-
tion of motion for the particle. The nature of this modied equation has
been subject to much controversy since the initial work of Abraham and
Lorentz. The standard equation for the change in the energy and momen-
tum of the particle in classical electrodynamics is the relativistic gener-
alization by Dirac of the work of Abraham and Lorentz and called the
1 Department of Mathematics, University of York, Heslington, York, YO10 5DD, U.K.

2 To whom correspondence should be addressed; e-mail: ah28@york.ac.uk
1149
0015-9018/05/0700-1149/0 2005 Springer Science+Business Media, Inc.
1150 Higuchi and Martin
LorentzDirac (or AbrahamLorentzDirac) equation.(13) (See, e.g., Ref.

4 for a modern review.) Thus, if a charge e with mass m is accelerated by

an external 4-force Fext , then its coordinates x ( ) at the proper time
obey the following equation:
d 2x
m = Fext + FLD , (1)
d 2

where the LorentzDirac 4-force FLD is given by

2c d 3 x dx d 2 x d 2 x
FLD + . (2)
3 d 3 d d 2 d 2
We have let c = 1 and dened c e2 /4. Our metric is g = diag

(+1, 1, 1, 1). This equation of motion is fundamentally different in
form from those which one usually encounters in mechanics as it is
third-order. This implies that a third initial condition is needed in addition
to the position and velocity. Herein lies the cause of much of the debate
as to the physical correctness of this theory. It should be noted that there
are many derivations of the LorentzDirac force using a number of differ-
ent methods besides the original work of Dirac(3) (see, e.g., Ref. 5). How-
ever, many unphysical and problematic effects occur in the implementation.
One example is the existence of run-away solutions for which a particle will
continue to accelerate under its own radiation reaction. This is certainly
an unphysical effect and thus a problem for the theory. Alternative forms
of implementing the LorentzDirac force, including the integro-differential
form, solve the problem of run-aways but create another problem in the form
of pre-acceleration. Thus, in this implementation of the LorentzDirac force
the particle accelerates before the external force causing the acceleration is
applied; this is an acausal effect which should not be present in classical
mechanics. This effect is represented in Fig. 1, which shows the acceleration
of the particle, a, against time, t, where a constant force is applied after time
t0 , i.e., in the shaded region.
Additional conditions and procedures are therefore necessary in order
to treat the LorentzDirac theory as a normal causal classical theory, and
there are such procedures (see, e.g., the review(4) or the discussions in Refs.
6 and 7). One such procedure is the reduction of order,(8) in which the
LorentzDirac force is treated order by order as a perturbation. Despite
this, the debate over the controversial aspects of Eq. (2) has continued, its
impact muted for a couple of reasons. First, the effect due to the third-
derivative term, often called the Schott term, is tiny for most physical sit-
uations; in fact the time scale for the pre-acceleration is too small for
Classical and Quantum Radiation Reaction for Linear Acceleration 1151
Fig. 1. Preaccelaration of a charged particle.
any known classical interactions to have a measurable effect. Second, the

advent of quantum electrodynamics (QED), which is the more fundamen-
tal theory of electrodynamics, has rendered the problems concerning the
classical LorentzDirac force less urgent. The second point, however, nat-
urally leads to the following question: If the perturbative treatment of the
LorentzDirac force is causal and satisfactory, can it be derived in QED in
the h 0 limit in perturbation theory? This question is the main moti-
vation of this paper.
In fact it was found in Refs. 9 and 10 (after an initial claim to the
contrary) that the change of the position, called the position shift, due
to the LorentzDirac force of a charged particle linearly accelerated by a
static external potential is reproduced by the h 0 limit of one-photon
emission process in QED in the non-relativistic approximation. (See, e.g.,
Refs. 1214 for other approaches to arrive at the LorentzDirac theory
from QED.) Here we present in detail the generalization of this work to a
fully relativistic particle, which was outlined in Ref. 11. At the end of the
paper, we shall also show similar results with a time-dependent but space-
independent potential. The paper is organized as follows. In Secs. 2 and 3,
we outline the classical and quantum models that are used in the paper.
In Sec. 4, we calculate expressions for the position expectation value for
both the hypothetical non-radiating particle (Sec. 4.1) and the radiating
particle (Sec. 4.2) before identifying the expression for the position shift
in terms of the emission amplitude. This amplitude is calculated in Sec. 5.
The result is used in Sec. 6 to show that the energy emitted in the h 0
limit in QED is given by the classical Larmor formula and to compute
the quantum position shift in Sec. 7. The comparison between the clas-
sical and quantum position shifts is presented in detail in Sec. 8 and the
time-dependent case is presented in Sec. 9 before we conclude the paper in

Sec. 10.
2. CLASSICAL MODEL
We start by describing the model to be investigated in this paper.

Consider a charged particle with charge e and mass m moving in one
space dimension under a potential. Let this motion be in the positive
z-direction and let the potential be a function of the spatial coordinate z,
viz V = V (z). We wish to analyze the change in the position of the par-
ticle due to radiation reaction. Thus we consider a model in which there
has been a period of acceleration, i.e. non-constant potential, at some time
in the particles history. We assume that V (z) = V0 = const. for z < Z1
and V (z) = 0 for Z2 < z for some Z1 and Z2 , both positive constants
with Z1 > Z2 . Thus, there is non-zero acceleration only in the interval
(Z1 , Z2 ) as represented in Fig. 2 for the case with V0 > 0.
The external 4-force in Eq. (1) representing the electric force
corresponding to this potential is given as Fextt = V (z)dz/d , Fext
z
=
y
V (z)dt/d with Fext = Fext = 0. The LorentzDirac 4-force can be given
x
t =F y
LD dz/d , FLD = FLD dt/d , FLD = FLD =
z x
in the following form: FLD
0 with
2c d
FLD ( 3 z ) . (3)
3 dt
A dot indicates the derivative with respect to t. We have dened

(1 z 2 )1/2 as usual.
Fig. 2. The potential V (z) and period of accelaration.

Suppose that this particle would be at z=0 at time t=0 if the

LorentzDirac force was absent. The true position at t=0, which we denote
z and call the position shift, can readily be found to lowest non-triv-
ial order in FLD by treating the LorentzDirac force as a perturba-
tion. The calculation can be facilitated by using the fact that the total
energy, m dt/d + V (z), changes by the amount of work done by the
LorentzDirac force. Thus we nd
t
d m
FLD (t )z(t )dt = z + V (z)z
d z 1 z 2

d z
= m 3 z 2 , (4)
dt z
where we have used (d/dt)(m z ) = m 3 z = V (z) to zeroth order, i.e. in

the absence of the LorentzDirac force. Rearranging and integrating, we
obtain the position shift due to the LorentzDirac force as
t
v0 0 dz 1
zLD = FLD dt dt, (5)
m dt (t)[z(t)]2
3
where v0 = z |t=0 is the nal velocity in the absence of radiation reaction.

We shall see that this result is reproduced in QED in the h 0 limit.
It is useful for later purposes to nd the change in the position shift
(5) caused by letting the particle, in the absence of the LorentzDirac
force, be at z = z0 = 0 at t = 0 as opposed to z = 0, with the same
nal velocity v0 . (We assume that z0 > Z2 .) The time this particle spends
after the acceleration and before t = 0 (when the position shift is mea-
sured) is lengthened by t0 =z0 /v0 . This has the effect of shifting the entire
trajectory earlier in time by t0 . Hence the calculation of the new position
shift may be made using the original trajectory, but measuring the shift at
t = t0 instead of t = 0. The extra contribution is therefore
t
v0 t0 dz 1
zextra = FLD dt dt. (6)
m 0 dt (t)[z(t)]2
3
Noting that for values of t [0, t0 ], z (t) = v0 and (t) = 0 (1 v02 )1/2
are constant and FLD = 0, we can interchange the order of integration to
obtain

v0 0 t0 1 dz
zextra = dt FLD dt
m 0 03 v02 dt
z0
= Eem , (7)
m03 v02
where Eem is the energy emitted as radiation given by
0
2c
Eem = ( 3 z )2 dt, (8)
3
which is the relativistic Larmor formula for one-dimensional motion. We

shall see that the extra contribution (7) to the position shift is also repro-
duced by QED in the h 0 limit. The total classical position shift can
now be written
zclass = zLD + zextra . (9)
3. THE QED MODEL
We now turn our attention to the quantum eld theoretic model.

We shall nd the position shift in this model to order e2 to compare it
with the classical counterpart in Sec. 2. The corresponding Lagrangian
density is
L = [(D + ieA )] [(D + ieA )](m/h)2 41 F F 21 ( A )2 ,

(10)
where F A A and D + iV /h. The eld describes

a charged scalar particle with mass m and charge e, the eld A is the
electromagnetic eld, and the function V is the external potential which
accelerates the charged scalar particle and is given as V = V (z)0 for the
static potential. The non-interacting quantum electromagnetic eld A (x)
is expanded as

d 3k
A (x) = 3
a (k)eikx + a (k)eikx , (11)
2k(2)

where k = k and the annihilation and creation operators, a (k) and a (k),
respectively, for the photons with momenta hk in the Feynman gauge satisfy
a (k), a (k ) = g (2)3 2hk(k k ). (12)

The Fourier expansion of the non-interacting quantum scalar eld is not

as straightforward because its eld equation involves the potential V (z) as
follows:

h2 [it V (z)]2 + h2 2 m2 = 0. (13)
Positive-frequency mode functions can be found in the form

i
p (t, x) = p (z) exp (p x p0 t) , p0 > 0, (14)
h
where x (x, y) and p (p x , py ). We dene the z-component of the

momentum p pz by the relation p02 = p2 +p 2 +m2 . We let p be positive
(negative) if the mode function corresponds to a wave coming in from z =

(z = +). The function p (z) with p = 0 satises
d2
h2 p (z) = [p (z)]2 p (z), (15)
dz2
where

p (z) [p0 V (z)]2 m2 . (16)
If p = 0, then the mass term m2 has to be replaced by m2 + p

2 . How-
ever, we need only the cases with p = 0 and p > 0 in our calculations.
The WKB approximation for the function p (z) with p > 0, which will be
useful later, reads
z
p
p (z) = exp p ( ) d , (17)
p (z) 0
where we have required p (0) = 1.

The non-interacting quantum charged scalar eld is expanded in
terms of these mode functions as

d 3p
(x) = h A(p) p (x) + B
(p) (x) . (18)
2p0 (2 h)3 p
(This expansion needs to be modied if there are modes which are totally
reected by the potentialthis is the case if V0 = 0or if there are modes
which fail to reach |z| = . However, such modications will not be rel-
evant since the affected modes will not enter our calculations.) The oper-
ators A(p) and B(p) satisfy

A(p), A (p ) = B(p), B (p ) = (2 h)3 2p0 (p p ), (19)
if one normalizes the mode functions appropriately, with all other commu-
tators vanishing. The WKB mode functions p (t, x) with p (z) given by
Eq. (17) are correctly normalized since it satises the usual normalization
condition | p (t, x)| = 1 in the region (Z2 , +) where V (z) = 0.
Now, in the interaction picture, if there is one scalar particle with
momentum p in the initial state, then the nal state to lowest non-trivial
order in e is a superposition of a state proportional to the initial state and
states with one scalar particle and one photon. Thus, the state A (p)|0
evolves as follows:
A (p)|0 [1 + iF(p)]A (p)|0

i d 3k
+ A (p, k)a (k)A (P)|0. (20)
h 2k(2)3
The F(p) is the forward-scattering amplitude of order e2 coming from

the one-loop diagram, which we do not evaluate explicitly. The emission
amplitude A (p, k) will play a central role in our calculations. Note that
the momentum hk of the photon is of order h because the wave number
k rather than the momentum has the classical limit. The momentum P of
the charged particle in the nal state is determined using conservation of
the transverse momentum and energy: p = P + hk , p0 = P0 + hk.
The z-component P P z is determined by the on-shell condition, P02 =
P 2 + P
2 + m2 .
We shall consider the initial state |i given by

d 3p
|i = f (p)A (p)|0, (21)
2p0 (2 h)3
where the function f (p) is sharply peaked about a momentum in the pos-
itive z-direction with width of order h. We also require that the p-deriv-
ative of f (p) has width of order h. This wave packet corresponds to the
classical particle considered in Sec. 3. We make a further assumption that
the WKB approximation (17) is valid for the momenta p where the func-
tion f (p) is not negligibly small. The WKB approximation is known to
be applicable if the wavelength stays approximately constant over many
periods (see, e.g., Ref. 17). Noting that the time-dependent wavelength
is 2 h[p (z)]1 , where p (z) is dened by Eq. (16), we nd that this con-
dition is satised as long as p (z) and p (z) can be regarded as quantities
of order h0 .
The normalization condition i | i = 1 implies

d 3p
|f (p)|2 = 1 . (22)
(2 h)3
This equation shows that the function f (p) can heuristically be regarded
as the one-particle wave function in the momentum representation. In
Ref. 11 it was assumed that f (p) was real for simplicity. This assumption
will be dropped in this paper.
4. POSITION EXPECTATION VALUE
In this section we derive a formula for the position shift in terms

of the emission amplitude A (p, k). We closely follow Ref. 10, correct-
ing a few errors. In particular, the fourth term in Eq. (47) was missing in
Ref. 10.
4.1. Non-radiating Particle
Dene the charge density by
i
(x) : t t :, (23)
h
where : . . . : denotes normal ordering. We let the potential satisfy |V0 | <
2m, thus precluding the possibility of scalar-particle pair creation. Then,
the charge density (x) coincides with the probability density for the par-
ticle if there is only one charged particle in the state. Hence the expecta-
tion value of z is

z = d 3 x z (t, x). (24)
We shall rst need to consider the expectation value of z for the hypotheti-
cal scalar particle that interact with the external potential but not with the
quantum electromagnetic eld. This will be the benchmark against which
we can dene the position shift due to the radiation. Hence, we must
evaluate the expectation value of the probability density in the absence
of radiation and use Eq. (24) to nd z. The initial and nal states in
this case are the same. Thus, using the commutation relations (19) for the
expectation value of the charge density (23) in the state |i given by (21),
we nd

d 3 p d 3p
(t, x) = i h f (p )f (p)
3 2p (2
h) 3
2p0 (2 h) 0

p (t, x)t p (t, x) t p (t, x) p (t, x) . (25)

The t-dependence of the mode functions is given by p (t, x) eip0 t/ h .
Hence, we nd

d 3 p d 3p
(t, x) = 1
f (p )f (p)
2 (2 h) 3 (2 h)3

p0 p0
+ (t, x) p (t, x). (26)
p0 p0 p
We arrange the wave packet so that at time t = 0 it is located far into the
region where the potential vanishes. Then we may approximate the mode
functions as follows:

p (0, x) ei(pz+p x )/ h . (27)
After substituting Eq. (26) in the position expectation formula (24) we use
the approximation (27) to write z p (0, x) i hp p (0, x), where p
/p. We can now use integration by parts for the variable p in Eq. (24)
to obtain the expectation value of z for the hypothetical non-radiating
particle, which we denote by z0 , as

i h d 3 p d 3 p p 0 p0
z0 = + f (p )f (p)
2 (2 h)3 (2 h)3 p p0 p0

d 3 x p (0, x) p (0, x). (28)
Substitution of the approximation (27) for p gives (2 h)3 3 (p p )

through the x-integration, and we obtain in the end

i h d 3p
z0 = 3
f (p) p f (p), (29)
2 (2 h)

where p = p p , after integrating by parts. This result can be inter-
preted as the expectation value of the position operator i hp in the
p-representation of the one-particle wave function. An example of a func-

tion f (p) satisfying Eq. (29) is f (p) = fR (p)eiz0 p/ h with the function
fR (p) being real though it is not necessary to make this assumption.
Note that we did not need to dene a position operator for the
one-particle Klein-Gordon wave function though our denition agrees
with the expectation value of the so-called NewtonWigner operator.15
(See, e.g., Ref. 16 for a discussion of problems associated with the New-
tonWigner operator in the context of the main theme of IARD 2004.)
4.2. Radiating Particle
Now we consider the case of the radiating particle. The nal state
resulting from the initial state |i can be found from Eq. (20) as

d 3p
|f = F (p)A (p)|0

2p0 (2 h) 3

i d 3k d 3p
+ G (p, k)a (k)A (P)|0, (30)
h 2k(2)3 2p0 (2 h)3
where P = p hk and P0 = p0 hk as before. We have dened
F (p) [1 + iF(p)] f (p), (31)

G (p, k) A (p, k)f (p). (32)
One can heuristically regard the function F (p) as the one-particle wave
function in the zero-photon sector in the p-representation and the function
G (p, k) as that in the one-photon sector with a photon with momentum
hk in the P-representation. By introducing the denition

i d 3p
C
(k) = A (p, k)f (p)A (P), (33)
h 2p0 (2 h)3
the nal state can be written

d 3p d 3k
|f = F (p)A (p)|0 + C (k)a (k)|0. (34)
2p0 (2 h)3 2k(2 )3
We rst derive the relation between the imaginary part of F(p) and
the scattering probability which results from unitarity. (Note that there is
no particle creation because of the condition V0 < 2m). Recalling the de-
nition (31) and using the unitarity of time evolution and noting f | f , we
nd

d 3p
(1 2 Im F(p))|f (p)|2 + Pem = 1 (35)
(2)3
to rst-order in e2 , where the emission probability Pem is given by

d 3k
Pem = h 0|C (k )C (k)|0
2k(2)3

1 d 3k d 3p
= f (p)A (p, k)
h 2k(2)3 2p0 (2 h)3

d 3 p
f (p )A (p , k) 0|A(P)A (P )|0. (36)
2p0 (2 h)3
To perform the p -integration recall rst that 0|A(P)A (P )|0 = (2 h)3
2P0 (P P ). We need to integrate this delta function with respect to p
) = (p p ) because
by relating it to (p p ). We note that (P P
P = p hk . Thus, we nd
P0 dp
0|A(P)A (P )|0 = (2 h)3 2p0 (p p ). (37)
p0 dP
By substituting this formula in Eq. (36), we obtain

1 d 3p d 3k P0 dp
Pem = |f (p)|2 A (p, k)A (p , k) . (38)
h (2 h)3 2k(2) 3 p0 dP
Equation (35) must hold for any function f (p), so we must have

1 d 3k P0 dp
2 Im F(p) = 3
A (p, k)A (p, k) . (39)
h 2k(2) p0 dP
The contribution of the one-photon-emission term to (t, x) is

d 3k
(t, x)em = h 0|C (k)(t, x)C (k)|0. (40)
2k(2)3
The integrand here is obtained from (t, x) for the non-radiating particle
if we replace the state |i in Eq. (21) by

h1/2 C (k)|0 = i d 3P
g (P)A (P)|0, (41)
2P0 (2 h)3
where

P0 dp
g (P) h1/2 A (p, k)f (p) . (42)
p0 dP
Hence we conclude from Eq. (29) that

d 3P
h d 3 x z 0|C (k)(t, x)C (k)|0 = i h g (P) P g (P). (43)

2 (2 h) 3
This implies that the one-photon-emission contribution to z is

i d 3k d 3P
zem = 3 3
f (p)A (p, k) P
2 2k(2) (2 h)

P0 dp 2
f (p)A (p, k) . (44)
p0 dP
We convert the integration variable from P to p by using d 3 P =

(P /p)d 3 p and the P -derivative to a p-derivative by using P =
(p/P )p to nd

i d 3p
d 3k P0 dp 2
zem = f (p) p f (p) A
(p, k)A (p, k)
2 (2 h)3 2k(2)3 p0 dP

i d 3p d 3k
|f (p)|2 A (p, k) p A (p, k). (45)
2 (2 h)3 2k(2)3
We have dropped the factor (P0 /p0 )(dp/dP )2 from the second term for
the following reason: it will turn out later that the emission amplitude A
and its p-derivative are both of order h0 . Thus the second term in the
above expression for zem is of order h0 . Since P0 = p0 and dp/dP = 1
to order h0 , we can drop the factors of P0 /p0 and dp/dP from the second
term. However, one needs to keep these factors in the rst term, which is
of order h1 .
From the expressions for the nal state (34) it is clear that the contri-
bution to z from the process without photon emission is obtained from
Eq. (29) by replacing f (p) with F (p) [1 + iF(p)]f (p). Thus

i h d 3p
zfor = 3
2i|f (p)|2 p Re F(p)+[1 2 Im F(p)] f (p) p f (p)
2 (2 h)
(46)
to rst-order in e2 . (Recall that the forward-scattering amplitude F(p)

is of order e2 .) We can now add the two contributions zem and zfor
together to nd the position expectation value:

i h d 3p

d 3p
z = f (p) p f (p) h |f (p)|2 p Re F(p)
2 (2 h)3 (2 h)3

i d 3p 2 d 3k
3
|f (p)| 3
A (p, k) p A (p, k)
2 (2 h) 2k(2)

i d 3p
f (p) p f (p)
2 (2 h)3

d 3k P0 dp dp
A (p, k)A (p, k)

1 . (47)
2k(2)3 p0 dP dP
The unitarity relation (39) has been used to eliminate 2 Im F(p). Each
term in Eq. (47) can be interpreted as follows: the rst term is the position
expectation value z0 for the non-radiating particle; the second term is the
contribution from what can be regarded as the one-loop quantum correc-
tion to the potential; therefore, we identify the sum of the third and forth
terms as the position shift to be compared with the classical position shift
zclass given by Eq. (9). Using Eq. (29) and the assumption that the func-
tion f (p) is sharply peaked about a momentum in the positive z-direction
with the width of order h, we nd these two terms in the h 0 limit as
zq = zq1 + zq2 (48)
with

i d 3k
zq1 = 3
A (p, k) p A (p, k), (49)
2 2k(2)

z0 d 3k dp
zq2 = A
(p, k)A (p, k) 1 , (50)
h 2k(2)3 dP
where the momentum at which the function f (p) is peaked is now denoted
simply by p. We have used the fact that dp/dP 1 is of order h to drop
the factor (P0 /p0 )(dp/dP ) in zq2 . We shall demonstrate that the quan-
tum position shift zq in the h 0 limit is identical with the classical
counterpart zclass by showing zq1 = zLD and zq2 = zextra .
Let us rst examine the latter equality. To this end we need to nd an
expression for dp/dP in terms of p and k. The energy conservation equa-
tion p0 P0 = hk gives a one-to-one relation between p and P for a given
2 = p2 = 0 because these are of order
k after letting P h2 . Then we nd

dp m2
= 1 2 hk. (51)
dP p p0
By using this formula in Eq. (50) we obtain
m 2 z0
zq2 = Eem , (52)
p 2 p0
where

d 3k
Eem kA (p, k)A (p, k) (53)
2k(2)3
is the expectation value of the energy emitted as radiation. By comparing

Eq. (52) with Eq. (7), it can be seen that the equality zq2 = zextra will
hold if
0
2c
Eem = ( 3 z )2 dt, (54)
3
which is identical to the relativistic generalization of the classical Larmor

formula. To show this equality (and also zq1 = zLD ) we need to nd the
h 0 limit of the one-photon-emission amplitude A (p, k) to which we
now turn.
5. EMISSION AMPLITUDE
The one-photon-emission part of the evolution of the state was rep-

resented earlier in Eq. (20) by

i d 3 k
A (p)|0 + A (p, k )a (k )A (P)|0. (55)
h 2k (2)3
The evolution in perturbation theory in the interaction picture is generated

by the interaction Lagrangian density as follows:

i
A (p)|0 + d 4 xLI (x)A (p)|0. (56)
h
We can take the inner product of the two expressions (55) and (56) with
the state 0|a (k)A(p ) and equate them in order to nd the amplitude
A (p, k). This procedure leads to

2hp0 (2 h)3 (p P)A (p, k) = d 4 x0|a (k)A(p )LI (x)A (p)|0.
(57)
Integrating over the variable p , with the appropriate measure, and rear-
ranging, we nd that A (p, k) is given by

1 d 3 p
A (p, k) = d 4 x0|a (k)A(p )LI (x)A (p)|0. (58)
h 2p0 (2 h)3
The LI (x) is obtained from the Lagrangian density (10) as follows:
ie
LI (x) = A : D (D ) : . (59)
h
Recall that D = + iV , where V = 0 V (z). By substituting Eq. (59)

in Eq. (58) and using the commutation relations (12) and (19) we nd

d 3 p
A (p, k) = ieh
2p0 (2 h)3

d 4 x eikx p (x)D p (x) D p (x) p (x) , (60)
where the mode functions p (x) are given by Eq. (14). As we mentioned
before, we assume that the WKB approximation (17) is valid for the initial
mode function. Since the momentum hk of the emitted photon is of order
h, the WKB approximation can also be used for the mode functions for
the nal state.
Since the transverse momentum p is assumed to be of order h, the
nal transverse momentum p hk is also of order h. This means that
we do not need to consider the x- and y-components of A (p, k) because
they are smaller by a factor of h compared to the t- and z-components.

Using D0 = t + iV (z)/h we obtain

d 3 p
At (p, k) = e d 4x
2p0 (2 h)3

p (z)p (z) p0 + p0 2V (z) eikx ei [(p0 p0 )t(p p )x ]/ h .
(61)
Integration with respect to t, x and y gives

dp
dz p0 +p0 2V (z) p (z)p (z)eik z (p0 +hkp0 ).
z
At (p, k)=e
2p0
(62)
We convert the p -integration to p0 -integration by noting that dp /p0 =

dp0 /p and nd

P0 + p0 V (z)
P (z)p (z)eik z ,
z
At (p, k) = e dz (63)
2P

where P0 = p0 hk with P P02 m2 as before. We let P0 = p0 and
P = p in Eq. (63) because the differences P0 p0 and P p are of order
h. Thus, we obtain

p0 V (z)
P (z)p (z)eik z .
z
At (p, k) = e dz (64)
p
Proceeding similarly for the z-component Az (p, k) we have

ieh dp (z) dP (z)
p (z) eik z .
z
Az (p, k) = dz P (z) (65)
2p dz dz
To simplify the expression for the emission amplitude A further, we need

to use the explicit form of p (z) given by Eq. (17). We note rst
z
Pp i
P (z)p (z) = exp p () P () d . (66)
P (z)p (z) h 0
Now, to lowest order in h we have

p (z)
p (z) P (z) = (p0 P0 )
p0
p0 V (z)
= hk. (67)
p (z)
Note that p (z) and p0 V (z) are the z- and t-components of m dx /d ,
respectively, for the classical particle with nal momentum p and vanish-
ing transverse momentum. Hence, vp (z) p (z)/[p0 V (z)] is the velocity
of this classical particle. Thus, p (z) P (z) = hk/vp (z). Using this for-
mula in Eq. (67) and substituting the result in Eq. (66) and letting P = p
and P (z) = p (z), because the differences P p and P (z) p (z) are of
order h, we nd
z
p dz
p (z)p (z) = exp ik
p (z) 0 vp (z)
p ikt
= e , (68)
p (z)
where the time t is dened by dz/dt = vp (z) with the condition t = 0 at
z = 0. Hence
+
z
At (p, k) = e dt eiktik z . (69)

It is convenient to dene the variable by t z cos with cos

k z /k. Then
+
dt ik
At (p, k) = e d e . (70)
d
Next we turn our attention to the z-component given by Eq. (65), We note
that, to lowest order in h, we can let h(d/dz)p (z) = ip (z)p (z). Hence
+
p + P
P (z)p (z)eik z
z
Az (p, k) = e dz
2p
+
dz ik
=e d e , (71)
d
where we have let P (z) = p (z) as before and used Eq. (68). This formula
and Eq. (70) can be combined as
+
dx ik
A (p, k) = e d e , (72)
d
where x is the classical trajectory with nal momentum p which passes

through (t, z) = (0, 0). (The minus sign for the z-component is due to the
raised index.) This emission amplitude is identical with that for a classical
point charge passing through (0,0).(11)
The expression (72) is ill-dened since dx /d remains nite as
. Therefore, we introduce a smooth cut-off function ( ) which
takes the value one while the acceleration is non-zero with the property
lim ( ) = 0. Thus,
+ dx
A (p, k) = e

d ( ) eik . (73)
d
In the end we take the limit ( ) 1 at any given in such a way that
+
d [ ( )] 0.
2
6. DERIVATION OF THE LARMOR FORMULA
In this section, we derive the relativistic generalization of the Larmor

formula for one-dimensional motion, Eq. (54), thus completing the dem-
onstration of the equality zq2 = zextra . It is convenient to use the fol-
lowing form of the emission amplitude obtained by integration by parts:

ie + d 2x dx
A (p, k) =

d + ( ) eik , (74)
k d 2 d
where we have used the condition that ( ) = 1 if d 2 x /d 2 = 0. By

substituting this formula in Eq. (53) we obtain
+ +
dk
Eem = e2 d d d
0 16 3

d 2x dx d 2 x dx
+ ( ) + ( ) eik( ) , (75)
d 2 d d 2 d
where d is the solid angle in the k-space. We extend the integration

range for k from [0, +) to (, +) and divide by two. Then using
+ ik( )
e dk = 2( ), we nd
! "
e2 + d 2 x d 2 x dx dx 2
Eem = d d + ( ) . (76)
16 2 d 2 d 2 d d
The second term tends to zero in the limit ( ) 1 due to the require-
+
ment [ ( )]2 d 0. Hence, we have in this limit

c + d 2 x d 2 x
Eem = d d , (77)
4 d 2 d 2
where c e2 /4 as before. Now, one can readily show, by using d/dt =
1 z cos , that
d 2z z
= , (78)
d 2 (1 z cos )3
d 2t d 2z
= cos . (79)
d 2 d 2
By substituting these formulae in Eq. (77) we obtain Eq. (54), thus dem-
onstrating the equality zq2 = zextra .
7. QUANTUM POSITION SHIFT
Our next task is to show that zq1 = zLD . This will establish that
the classical position shift zclass is equal to the quantum one zq in the
h 0 limit.
In the product A (p, k)p A (p, k) in Eq. (49) we use Eq. (74) for
A and Eq. (73) for p A . Proceeding in a way similar to that which led
to Eq. (77) we nd
! "
c d 2 x dx 1 dx dx d 2
zq1 = d d + [ ( )] . (80)
4 d 2 p d 4 p d d d
Next we demonstrate that the second term, which still contains the cut-off
function, vanishes when integrated over the solid angle and . Noting that
(dx /d )(dx /d ) = (d/d )2 , we nd by integration by parts that the
contribution of this term is proportional to the following integral:
+ ! 2 "
d d
I d d [ ( )]2 . (81)
d p d
Since /p is taken with xed, the - and p-derivatives commute. Hence

this integral is equal to
2 =+
d
I= d , (82)
p d
=
where we have used the fact that the -derivative of (d/d )2 is non-zero
only if the acceleration is non-zero and, therefore, only when the cut-off
function ( ) equals one. It can readily be shown that d(d/d )2 =
4. Hence, we have I = 0. Thus, the contribution from the second term
in Eq. (80) vanishes. The remaining term gives the main contribution to
the position shift due to radiation reaction in QED in the h 0 limit as

c d 2x dx
zq1 = d d . (83)
4 d 2 p d
To compare zq1 given by this equation with zLD given by Eq. (5) we
need to nd an expression of zq1 in terms of t rather than . Thus, we
need to eliminate the variable using its denition = t z cos . The
two components of the second derivative d 2 x /d 2 are given by Eqs. (78)
and (79). Let us consider the second factor of the integrand of (83), i.e. the
p-derivative of dx /d . By interchanging the order of integration we obtain

dx dt d x
= , (84)
p d d dt p
where the subscript indicates that the partial derivative with respect to p
is taken with xed. Then by differentiating the equation t = z cos +
with respect to p with xed, we nd

t z
= cos . (85)
p p
Combining this equation with Eqs. (78) and (79), which give d 2 x /d 2 , we
can write the integrand in Eq. (83) as

d 2x dx z d z
= sin2 . (86)
d 2 p d (1 z cos )4 dt p
We substitute this expression in the quantum position shift (83) and

change the integration variable from to t and integrate by parts. The
result is
0
c d z z
zq1 = d dt sin2 , (87)
4 dt (1 z cos ) 3 p
where we have also changed the integration range from (, +) to

(, 0] because z = 0 for [0, +). Finally, we need to relate (z/p)
to (z/p)t . By regarding z as a function of t and p we have dz = z dt +

(z/p)t dp. By substituting dt = d + cos dz in this equation and rear-
ranging, we obtain

z 1 z
dz = d + dp. (88)
1 z cos 1 z cos p t
Thus,

z 1 z
= . (89)
p 1 z cos p t
Using this expression in the position shift (87) we have

0
c z d z sin2 sin 2
cos
zq1 = d dt +3z2 .
4 p t dt (1 z cos )4 (1 z cos )5
(90)
We can now perform the integration over the solid angle with the
following result:

2c 0 d z z
zq1 = dt 4 + 3 6 z 2 z
3 dt p t
0
z
= dt FLD , (91)
p t
where we have used the expression of FLD given by Eq. (3). This formula
will be used in the next section to show that zq1 = zLD . We note here
that Eq. (91) is valid for an external force of any form provided that the
photon emission amplitude is given by Eq. (72).
8. COMPARISON OF CLASSICAL AND QUANTUM POSITION

SHIFTS
Interchanging the order of integration for Eq. (5) gives

t
v0 0 1 dz
zLD = 3 2
dt FLD dt. (92)
m 0 (t )[z(t )] dt
Comparison between Eqs. (91) and (92) shows that the equality zq1 =
zLD follows if
t
z v0 1
= z (t) dt . (93)
p t m 3 z(t )]2
0 (t ) [
This equation can be demonstrated as follows. By using energy conserva-

tion we obtain
2 1/2
dz
= 1 m2 p 2 + m2 V (z) . (94)
dt
Bydifferentiating both
sides with respect to p with t xed, and noting that
p/ p 2 + m2 = v0 and p 2 + m2 V (z) = m , we obtain

d z 1 z
= v0 V (z) . (95)
dt p t m 3 z p t
By substituting the formula V (z) = m 3 z (see the equation given after

Eq. (4)) in Eq. (95) we nd

d 1 z v0
= . (96)
dt z p t m 3 z 2
Then by integrating this formula, remembering that z = 0 at t = 0 for all

p, we arrive at Eq. (93).
The derivation of zq1 = zLD above uses an explicit relation, Eq.
(94), between the the trajectory z(t) and the momentum p. However, the
simple form for the quantum position shift zq1 given by Eq. (91) suggests
that there should be a derivation which does not rely on the relation (94).
We show next that there is indeed such a derivation which is valid in a
more general setting.
We consider a charged particle moving in the z-direction accelerated
by an external force which may depend on both z and t. Thus, we con-
sider the following equation of motion:
d 2z dt
m 2
= [Fext (t, z) + FLD ] . (97)
d d
This equation describes a classical particle under the inuence of an exter-

nal force Fext (t, z) and the LorentzDirac force. The model analyzed so far
in this paper is the special case given by Fext (t, z) = V (z).
We rst consider a solution to Eq. (97) in the absence of radiation
reaction, i.e. with FLD set to 0, with the momentum and the position
being p and 0, respectively, at t = 0. We dene P m dz/d and let this
solution be denoted (z, P ) = (z(t), P (t)). We have (z(0), P (0)) = (0, p).
Let a linearized solution about (z, P ), also with FLD = 0, be (z, P ) =
(z + z, P + P ). From the equation of motion (with FLD = 0) the quan-

tities z and P are found to satisfy the equations
d
z = m1 (1 z )3/2 P A(t)P ,
2
(98)
dt
d Fext ##
P = # z B(t)z. (99)
dt z z=z
Let (zs (t), Ps (t)) be a set of solutions labeled by s of these equations
satisfying (zs (s), Ps (s)) = (0, 1). The linearized solutions represent a
particle whose position at t = 0 coincides with z (0) (see Fig 3). The quan-
tity (z/p)t in Eq. (91) is the rate of change in the position z at time t
as the momentum P at t = 0 is changed while the position z at t = 0 is
xed. Thus,

z z0 (t)
= = z0 (t). (100)
p t P0 (0)
By substituting this formula in Eq. (91) we obtain

0
zq1 = dt FLD (t)z0 (t). (101)

Next a similar expression will be derived for zLD . We again let

(z(t), P (t)) be a solution of (97) with FLD = 0, and now let (z+z, P +P )
be the approximate solution obtained by perturbing (z, P ) to rst-order in
Fig. 3. The world lines for the solutions

(z(t), P (t)) and (z(t) + z0 (t), P (t)+P0
(t)).
FLD with the initial conditions (z, P ) = (0, 0) before the acceleration
starts. The quantities z and P satisfy
d
z = A(t)P , (102)
dt
d
P = B(t)z + FLD (t), (103)
dt
where A(t) and B(t) are dened by Eqs. (98) and (99), respectively. The
quantity z (P ) is the difference in position (momentum) between the
hypothetical non-radiating particle and the radiating particle. Thus, the
quantity z at t = 0 is equal to zLD .
The solutions to Eqs. (102) and (103) with the initial conditions
specied above can be expressed in terms of another set of solutions
(zs (t), Ps (t)) of the non-radiating equations, Eqs. (98) and (99), as
t
z(t) = ds FLD (s)zs (t), (104)

t
P (t) = ds FLD (s)Ps (t). (105)

These can readily be shown to satisfy Eqs. (102) and (103) by direct differ-
entiation using the conditions zt (t) = 0 and Pt (t) = 1. The classical
position shift due to the LorentzDirac force, at time t = 0, is therefore
given by
0
zLD = dt FLD (t)zt (0). (106)

This expression for the classical position shift zLD is similar to that for
the quantum position shift zq1 in Eq. (101), the only difference being that
the quantity z0 (t) is replaced by zt (0). In Fig. 4, we show the world
lines corresponding to the solution (z(s), P (s)) and the approximate solu-
tion (z(s) + zt (s), P (s) + Pt (s)) to the equations for the hypothetical
non-radiation particle motion.
Now, we only need to establish that z0 (t) = zt (0) to show that
the quantum and classical position shifts (101) and (106) are equal. In fact
one can show in general that zs (t) = zt (s) for any values of s as fol-
lows. Given a pair of solutions (z(1) , P (1) ) and (z(2) , P (2) ) to Eqs.
(98) and (99) we dene the symplectic product by
z(1) , P (1) |z(2) , P (2) z(1) P (2) P (1) z(2) . (107)

Fig. 4. The world lines for the solutions

(z(s), P (s)) and (z(s) + zt (s), P (s) + Pt
(S)).
It can readily be shown that the symplectic product is time-independent by

differentiating it with respect to t and using Eqs. (98) and (99). With the
identication (z(1) , P (1) ) = (zs , Ps ) and (z(2) , P (2) ) = (zt , Pt )
we obtain
zs (t)Pt (t) Ps (t)zt (t) = zs (s)Pt (s) Ps (s)zt (s). (108)
Since zs (s) = zt (t) = 0 and Ps (s) = Pt (t) = 1, we have zs (t) =
zt (s). Hence the equality zq1 = zLD holds. This equality and the
equality zq2 = zextra establish zq = zclass (see Eqs. (9) and (48)). Thus,
the position shift due to QED to order e2 in the limit h 0 is equal to
the position shift due to the LorentzDirac force for linear motion if the
acceleration is caused by a static potential.
We have shown that the classical position shift zLD equals zq1 in
Eq. (101) for any external force which depends on t and z. We have also
shown that Eq. (101) is valid provided that the photon emission amplitude
is given by Eq. (72). Thus, the equality zq1 = zLD holds if the ampli-
tude A (p, k) is given by Eq. (72). We have demonstrated Eq. (72) for the
case where the external force Fext is t-independent. The extension to the
case where the external force depends on both t and z does not appear to
be straightforward. However, our method can be applied to the case with
an external force that depends only on t because the WKB approximation
is very similar to that for the t-independent case as we demonstrate in the
next section.
9. TIME-DEPENDENT POTENTIAL
Let the potential be given by Vt = 0, Vz = V (t) in the QED

Lagrangian density (10). We show below that the emission amplitude in
this case is also given by Eq. (72).
Since the calculation is quite similar to the time-independent case, we
present only a brief description. The WKB approximation to the mode
functions is given by

i

p (t, x) p (t) exp (p x + p z) ,
z
(109)
h

p0 i t

p (t) exp p ( )d , (110)
p (t) h 0

p (t) [p V (t)]2 + m2 if p = 0. (The mass term m2 needs to be
replaced by m2 + p 2 if p
= 0 as before.) The function p (t) is the
t-component of the kinetic energy-momentum m dx /d of the non-radi-
ating classical particle at t. Note that the momentum p including the z-
component p z = p is conserved, but the energy is not. The calculation of
the emission amplitude follows the same pattern as before with the roles
of t and z (and those of p and p0 ) reversed.
Let A (p, k) be, again, the emission amplitude for the process in
which the charged particle with momentum p emits a photon with
momentum hk so that the nal momentum of the charged particle is given
by P = p hk. We again assume that the transverse momenta are of order
h and can be neglected. Then the z-component to order h0 (to be com-
pared with Eq. (64)) is

p V (t)
Az = e dt P (t) p (t) eikt . (111)
p0
For the t-component (to be compared with Eq. (65)) we obtain

$ %
ieh
At = dt P t p t P p eikt . (112)
2p0
The product of mode functions in Eq. (111) is

P0 p 0 i t
P (t) p (t) = exp (p ( ) P ( ))d . (113)
P (t)p (t) h 0
To lowest order in h we can change P0 to p0 and P (t) to p (t) in

the pre-factor as the differences are of order h. Recalling the classical
momentum conservation equation, m dz/d + V (t) = p, and the formula
m dt/d = p (t), where z(t) is the classical trajectory with z(0) = 0, we
obtain to lowest order in h
p dz z
p (t) P (t) = (p P ) = hk . (114)
p dt
Thus the product of the mode functions in question can be written

p0 i t dz z

P (t)p (t) = exp dt hk
p h 0 dt
p0
= exp(ik z z). (115)
p
This gives the z-component of the emission amplitude as

p V (t) iktik z z
Az (p, k) = e dt e
p (t)

dz ik
= e d e , (116)
d
where = t z cos and cos k z /k as before. To calculate the t-com-

ponent of the amplitude, note rst that to lowest-order in h we have
h p (t)t p (t) = ip0 exp(ik z z), (117)
where Eq. (115) has been used. By using this equation in Eq. (112) we
obtain

dt ik
At (p, k) = e d e . (118)
d
Equations (116) and (118) can be written A (p, k) = e d (dx /d )eik .

As we mentioned before this equality is sufcient to conclude that zq1 =
zLD .
The equality zq2 = zextra , which is necessary to establish the equal-
ity of the classical and quantum position shift, is trivial in this case.
Since the system is translationally invariant, the classical position shift is
invariant under the shift in the position, z = 0z0 , of the non-radiating
hypothetical particle at t = 0. Hence zextra = 0. Furthermore, since P =
p hk z , we have dp/dP = 1 and hence zq2 = 0 in Eq. (50). Thus, we have
zq2 = zextra = 0. Hence, zq = zclass , i.e., the position shift from the Lo-
rentzDirac force agrees with that in QED in the limit h 0 to order e2
for a space-independent potential as well.
The equality zq1 = zLD can be veried by a more explicit calcula-
tion as in the case with a static potential. Note that the momentum con-
servation equation reads
d
[m z + V (t)] = FLD (119)
dt
because the momentum m z +V (t) would be constant if there was no radi-

ation reaction. Hence
t
d
m 3
(z) = FLD (t ) dt . (120)
dt
Thus, the classical position shift zclass = LD is given by
0 t
1
zclass = dt FLD dt, (121)
0 m 3
where we have interchanged the order of integration. Thus, what we need

to establish is
t
z 1
= dt (122)
p t 0 m 3
for the hypothetical non-radiating particle. By using the momentum con-

servation for this particle in the z-direction and the condition z = 0 at
t = 0 we nd
t! "1/2
m2
z= 1+ dt. (123)
0 [p V (t)]2
this expression with respect to p and using p V (t) =

By differentiating
m dz/d and [p V (t)]2 + m2 = m dt/d , we indeed obtain Eq. (122).
10. CONCLUSION
In this paper, we showed that the change in position due to the radi-
ation reaction of a particle, which we call the position shift, according
to the LorentzDirac theory in classical electrodynamics is reproduced by
the h 0 limit of QED with a scalar charged particle. The calculation
was performed to lowest non-trivial order in e2 for wave packets linearly
accelerated for a nite time by either a purely space-dependent or time-
dependent potential. The quantum wave functions constituting the wave
packets were approximated using the WKB functions for each potential.
The agreement between QED and the LorentzDirac theory was demon-
strated by concrete calculations which gave the position shifts explicitly.
We also showed by a general argument that this agreement holds as long
as the emission amplitude coincides in the h 0 limit with that for a
classical point charge.
This work extended the corresponding work in the non-relativistic
approximation(9,10) and provided the details omitted in Ref. 11 as well
as the analysis for the time-dependent but space-independent potential.
The results presented here can be generalized to the case with a parti-
cle moving in three space dimensions if the potential depends only on
one variable, e.g. only on time. This generalization will be presented else-
where together with the demonstration that the one-loop correction to
the potential can be neglected at order h0 . [In the (unrealistic) model in
which the acceleration is achieved by a space-dependent mass term(10) , this
correction is of order h1 , thus overwhelming the contribution from the
LorentzDirac force.] It will also be interesting to see whether our results
can be justied when more than one photons are emitted: our results, as
they stand, are logically consistent only if the emission probability Pem
given by Eq. (38) is much smaller than one so that the process is domi-
nated by one-photon emission. These issues are currently under investiga-
tion.
REFERENCES
1. M. Abraham and R. Becker, Theorie der Elektrizitat (Springer, Leipzig, 1933), Vol. II.
2. H. A. Lorentz, Theory of Electrons (Dover, New York, 1952).
3. P. A. M. Dirac, Classical theory of radiating electrons, Proc. R. Soc. Lond. A167,
148 (1938).
4. E. Poisson, An introduction to the LorentzDirac equation, ArXiv:gr-qc/9912045.
5. C. Teitelboim, Splitting of the Maxwell tensorradiation reaction without advanced
elds, Phys. Rev. D 1, 1572 (1970); Splitting of the Maxwell tensor. II. Sources,
3, 297 (1971); Radiation reaction as a retarded self-interaction, 4, 345 (1971).
6. J. D. Jackson, Classical Electrodynamics (Wiley, New York, 1975).

7. F. Rohrlich, Classical Charged Particles (Addison-Wesley Publishing, Massachusetts,
1965).
8. L. D. Landau and E. M. Lifshitz, The Classical Theory of Fields (Pergammon, Oxford,
1962).
9. A. Higuchi, Radiation reaction in quantum mechanics, arXiv:quant-ph/9812036.
10. A. Higuchi, Radiation reaction in quantum eld theory, Phys. Rev. D 66, 105004
(2002); 69, 129903(E) (2004), arXiv:quant-ph/0208017.
11. A. Higuchi and G. D. R. Martin, LorentzDirac force from QED in linear acceler-
ation, arXiv:quant-ph/0407162, Phys. Rev. D 70, 081701(R) (2004).
12. E. J. Moniz and D. H. Sharp, Absence of runaways and divergent self-mass in nonrel-
ativistic quantum electrodynamics, Phys. Rev. D 10, 1133, (1974); Radiation reaction
in non-relativistic quantum electrodynamics, 15, 2850, (1977).
13. P. R. Johnson and B. L. Hu, Stochastic theory of relativistic particles moving in a
quantum eld: scalar Abraham-LorentzDirac-Langevin equation, radiation reaction,
and vacuum uctuations, Phys. Rev. D 65, 065015 (2002), ArXiv: quant-ph/0101001.
14. V. S. Krivitski and V. N. Tsytovich, Average radiation-reaction force in quantum elec-
trodynamics, Sov. Phys. Usp. 34, 250 (1991).
15. T. D. Newton and E. P. Wigner, Localized states for elementary systems, Rev. Mod.
Phys. 21, 400 (1949).
16. L. P. Horwitz, Time and evolution of states in relativistic classical and quantum
mechanics, ArXiv: hep-ph/9606330, IASSNS-HEP-96/59.
17. R. Shankar, Principles of Quantum Mechanics, (Plenum, New York, 1994), pp. 435438.

Classical and Quantum Radiation Reaction For Linear Acceleration

Încărcat de

Informații document

Titlu original

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

Classical and Quantum Radiation Reaction For Linear Acceleration

Încărcat de

Drepturi de autor:

Formate disponibile

Foundations of Physics, Vol. 35, No.

7, July 2005 ( 2005)

Classical and Quantum Radiation Reaction

Atsushi Higuchi1, 2 and Giles D. R. Martin1

We investigate the effect of radiation reaction on the motion of a wave packet

KEY WORDS: radiation reaction; quantum electrodynamics; LorentzDirac;

A charged particle radiates when it is accelerated. This simple discovery

1 Department of Mathematics, University of York, Heslington, York, YO10 5DD, U.K.

LorentzDirac (or AbrahamLorentzDirac) equation.(13) (See, e.g., Ref.

We have let c = 1 and dened c e2 /4. Our metric is g = diag

Fig. 1. Preaccelaration of a charged particle.

any known classical interactions to have a measurable effect. Second, the

time-dependent case is presented in Sec. 9 before we conclude the paper in

We start by describing the model to be investigated in this paper.

A dot indicates the derivative with respect to t. We have dened

Fig. 2. The potential V (z) and period of accelaration.

Suppose that this particle would be at z=0 at time t=0 if the

where we have used (d/dt)(m z ) = m 3 z = V (z) to zeroth order, i.e. in

where v0 = z |t=0 is the nal velocity in the absence of radiation reaction.

where Eem is the energy emitted as radiation given by

which is the relativistic Larmor formula for one-dimensional motion. We

zclass = zLD + zextra . (9)

3. THE QED MODEL

We now turn our attention to the quantum eld theoretic model.

L = [(D + ieA )] [(D + ieA )](m/h)2 41 F F 21 ( A )2 ,

where F A A and D + iV /h. The eld describes

a (k), a (k ) = g (2)3 2hk(k k ). (12)

The Fourier expansion of the non-interacting quantum scalar eld is not

Positive-frequency mode functions can be found in the form

where x (x, y) and p (p x , py ). We dene the z-component of the

(negative) if the mode function corresponds to a wave coming in from z =

If p = 0, then the mass term m2 has to be replaced by m2 + p

where we have required p (0) = 1.

A(p), A (p ) = B(p), B (p ) = (2 h)3 2p0 (p p ), (19)

A (p)|0 [1 + iF(p)]A (p)|0

The F(p) is the forward-scattering amplitude of order e2 coming from

We shall consider the initial state |i given by

4. POSITION EXPECTATION VALUE

In this section we derive a formula for the position shift in terms

4.1. Non-radiating Particle

Dene the charge density by

p (t, x)t p (t, x) t p (t, x) p (t, x) . (25)

Substitution of the approximation (27) for p gives (2 h)3 3 (p p )

4.2. Radiating Particle

where P = p hk and P0 = p0 hk as before. We have dened

F (p) [1 + iF(p)] f (p), (31)

the nal state can be written

to rst-order in e2 , where the emission probability Pem is given by

By substituting this formula in Eq. (36), we obtain

The contribution of the one-photon-emission term to (t, x) is

Hence we conclude from Eq. (29) that

This implies that the one-photon-emission contribution to z is

We convert the integration variable from P to p by using d 3 P =

to rst-order in e2 . (Recall that the forward-scattering amplitude F(p)

zq = zq1 + zq2 (48)

By using this formula in Eq. (50) we obtain

is the expectation value of the energy emitted as radiation. By comparing

which is identical to the relativistic generalization of the classical Larmor

The one-photon-emission part of the evolution of the state was rep-

The evolution in perturbation theory in the interaction picture is generated

The LI (x) is obtained from the Lagrangian density (10) as follows:

Recall that D = + iV , where V = 0 V (z). By substituting Eq. (59)

they are smaller by a factor of h compared to the t- and z-components.

If p = 0, then the mass term m2 has to be replaced by m2 + p

where we have used the condition that ( ) = 1 if d 2 x /d 2 = 0. By

where d is the solid angle in the k-space. We extend the integration

(z + z, P + P ). From the equation of motion (with FLD = 0) the quan-

z(1) , P (1) |z(2) , P (2) z(1) P (2) P (1) z(2) . (107)

this expression with respect to p and using p V (t) =