a

r
X
i
v
:
1
0
1
0
.
0
2
2
1
v
1


[
h
e
p
-
t
h
]


1

O
c
t

2
0
1
0
On The Relativistic Classical Motion of a
Radiating Spinning Particle in a Magnetic
Field
Arnab Kar

and S. G. Rajeev

October 4, 2010
Department of Physics and Astronomy
University of Rochester
Rochester NY 14627
USA
Abstract
We propose classical equations of motion for a charged particle with
magnetic moment, taking radiation reaction into account. This generalizes
the Landau-Lifshitz equations for the spinless case. In the special case of
spin-polarized motion in a constant magnetic eld (synchrotron motion) we
verify that the particle does lose energy. Previous proposals did not predict
dissipation of energy and also suffered from runaway solutions analogous to
those of the Lorentz-Dirac equations of motion.
1 Introduction
What is the classical motion of an electron in a constant magnetic eld? Surpris-
ingly, this question does not have a denitive theoretical answer even now. The
complications arise when the radiation from the particle is considered. Without

arnabkar@pas.rochester.edu

Also at the Department of Mathematics

1
it, the answer is quite elementary: a helical path around the magnetic eld[1]
obtained by solving the Lorentz equation of motion.
1
du

d
=
q
m
F

(1)
But we know that matters are not so simple in reality: an accelerated charge radi-
ates, and the radiation emitted by it exerts a force back on it. If the energy is not
replenished the path should be a spiral, the radius of the circle decreasing with the
energy.
Determining the force exerted on a point particle by its own eld is fraught
with conceptual difculties: the force is innite. Dirac[2] made a major step
forward by showing that this divergence can be removed by a renormalization of
the mass. The resulting equation (the Lorentz-Dirac equation of motion)
2
du

d
=
q
m
F

+
_

d
2
u

d
2
, =
2
3
q
2
m
(2)
is still not the correct answer. It predicts runaway solutions: even in the absence
of any external elds, a particle can accelerate away to innity with exponentially
growing energy. Part of the problem is that the radiative terms persist even in
the absence of external forces. Also, the equation involves the third derivative of
position; so it needs a new initial condition. Dirac suggested that this be chosen
such that the energy remains nite forever. However, such solutions suffer from
another problem: a violation of causality; the particle would accelerate before the
external eld is turned on.
The textbook of Landau and Lifshitz[3] proposed (without proof) a way out of
this dilemna. Replace the derivative of acceleration by its leading approximation
in an expansion in powers of :
du

d
=
q
m
F

+
_

d
d
_
q
m
F

_
, =
2
3
q
2
m
(3)
The radiative term now vanishes when the extrenal force is zero; and the equa-
tion now involves only the second order derivative of position. The free particle
does not run away. Further analysis shows that the orbits in a constant magnetic
1
In units with c = 1. q and m are the charge and mass respectively. Also, u

=
dx

d
is the four-
velocity and is proper time. Note the constraint

= 1 and its derivative

u
du

d
= 0.
2
The last term includes a projection operator that ensures that the constraint

u
du

d
= 0 is
satised.
2
eld [4] and in a Coulomb eld [5] decay as expected. But why should we stop at
rst order in the iteration in powers of ? Although it works, the ad hoc prescrip-
tion of Landau and Lifshitz is unsatisfactory.
To get sensible solutions to the Lorentz-Dirac equation, its new degree of free-
dom must be continuously ne-tuned so that the trajectory stays on a submanifold
of the phase space (the critical submanifold). Spohn used modern ideas in the
theory of dynamical systems to determine the equation of motion projected to this
critical submanifold: it is precisely the above Landau-Lifshitz equation! Thus
we consider the century-old problem of determining the equation of motion of a
radiating point charge to be settled nally.
However, the electron has a magnetic moment as well as electric charge; ra-
diation arises not only from the acceleration of the charge but also the precession
of the magnetic moment. In the relativistic case, these are of comparable magni-
tude. Therefore, it is necessary to extend the Lorentz-Dirac and Landau-Lifshitz
equations to the case of a spinning charged particle.
Bargman, Michel and Telegdi[6] proposed, more than a half century ago, an
equation for a (non-radiating) classical particle with charge and magnetic moment.
However, it does not seem to be the classical limit of the Dirac wave equation of
the electron. The correct equation must have a canonical form whose quantization
leads to the Dirac wave equation. The BMT equation does not appear to satisfy
this condition.
A better approach was proposed more recently by Barut and Zanghi[7]. We
will review their ideas, which show that a canonical quantization does lead to the
Dirac equation. We will also nd solutions of the Barut-Zanghi equation for a
non-radiating charged particle in a constant magnetic eld: itself a new result.
Barut and Unal[8] also derived the analogue of the Lorentz-Dirac equation,
after a mass renormalization. However, it suffers from the same runaway solutions
of the free particle, as the spinless case.
In this paper, we propose an equation of motion for a radiating spinning charged
particle with gyromagnetic ratio 2 (e.g., the electron or the muon) by applying a
prescription analogous to that of Landau and Lifshitz to the Barut-Zanghi equa-
tion. We will show that, at least for spin-polarized orbits in a constant magnetic
eld, the orbits decay as expected physically. It should be possible to test our
equations using a small (few MeV) electron synchrotron or a muon storage ring,
such as the one used for the determination of g2. Also, it should not be difcult
to revise the predictions for an electron accelerated by a laser eld[9]. Current
predictions do not include the effect of the magnetic moment.
We do not attempt, in this paper, to derive our prescription from rst principles
3
as Spohn did.
2 Non-Radiating Spinning Charged Particle
Can we talk of the motion of a particle with spin in the classical limit? At rst this
would appear to be impossible. But we do it all the time. The motion of electrons
in synchrotron of a radius of several meters (if not kilometers) is surely classical:
even if the spin were polarized, the differences in the quantum energy levels of
the electron are much too small to be of signicance. If we can nd a lagrangian
whose canonical quantization is the Dirac wave equation, it can be the basis for
such a classical treatment.
The lagrangian proposed by Barut and Zanghi
3
is,
L =
i
2
[

zz z z] + p

H, H = z

z[p

qA

] , z = z

0
(4)
Here z is a Dirac spinor with four complex (not Grassmann) numbers as com-
ponents. The dimensions of z are that of the square root of action (=mass*length)
and that of is the inverse of mass, in units where c = 1 but h = 1. Then qF

has
dimensions of mass over length.
The equations of motion are, with

= p

qA

dx

d
= z

z
i
dz
d
=

z
d

d
= qF

z
(5)
Note that, is not proper time as x

is not necessarily of constant length. But

we have the orthogonality condition
x

= 0. (6)
These equations follow from the Hamiltonian
H = z

(7)
3
They use a slightly different system of units.
4
and Poisson Brackets (P.B.) are obtained using
{ f , g} = i{
f
z
g
z

g
z
f
z
}+g

{
f
p

g
x

f
x

g
p

} (8)
_

_
=qF

,
_

, x

_
=

{z, z} = i,
_

, z
_
= 0 ={z, x} (9)
All other pairs of P.B. are zero. We can introduce a constraint
zz = a (10)
consistent with the P.B. and the equations of motion. The quantity a, with
the dimensions of action, sets the scale for the intrinsic angular momentum of the
particle.
Upon quantization, the wave function can be thought of as a function of x,
p

=
h
i

; it is also a polynomial in z, with z = h

z
satisfying the constraint
h z

z
= a. (11)
Thus the parameter a is quantized in multiples of h. The smallest value is
a = h for which (x, z) = z(x) is a linear function.
The Dirac wave equation is the eigenvalue equation for H, with eigenvalue
am:
i h

= m (12)
Other quantizations, for which( z) is a higher degree polynomial, correspond
to the wave equations of Bhabha and Harish-Chandra. They could be interesting
as wave equations for composite particles like hadrons[10].
It is convenient to introduce the velocity and spin variables
v

= z

z, S

=
i
4
z[

]z (13)
so that
5
dx

d
= v

dv

d
= 4S

d
= qF

dS

d
= v

(14)
These equations describe a relativistic analogue of the isotropic precessing top
of Poinsot [11], as we explain in the Appendix. The hamiltonian
H = v

(15)
describes a coupling to the external eld which causes a precession of the
velocity and spin. For a constant electromagnetic eld, K = qS

is
conserved as well:
d
d
_
qS

= 2qF

2qF

= 0. (16)
This is the classical analogue of the square of the Dirac operator, correspond-
ing to the gyromagnetic ratio 2. We will see later that it is proportional to the
angular momentum in the plane of the electromagnetic eld.
3 A Non-Radiating Spinning Particle in a Constant
Magnetic Field
Now let us turn to solving the equations of motion in the case of a magnetic
eld pointed along the third direction: F
12
=F
21
> 0; and all other components
are zero. Before we solve the problem including radiation damping, we must
recover the solution without radiation. Even this appears to be new: Barut and
collaborators only solved the case of a free particle.
We consider only orbits that lie in the 12 plane. That is, we restrict to the
case where all the tensor indices , , are in the 012 subspace. It is useful to
introduce the dual variables
6
qF

=
1
2

(17)
For a constant magnetic eld, only the time component F
0
is non-zero. We
can assume it to be positive without loss of generality.
Then
d

d
=

dv

d
= 4

dS

d
=

(18)
If we dene

=
1
2
(S

+
1
2
v

),

=
1
2
(S

1
2
v

)
they reduce to
d

d
= 2

] F

d
= 2

d
= 2

(19)
3.1 Spin-Polarized Orbits
There is a special class of left-handed orbits for which

= 0. (Also, the op-

posite case with

= 0.) For these, the spin and velocity are synchronized in the
following way:
7
S

=
1
2

(20)
In the non-relativistic limit, this means the spin is pointed along the magnetic
eld (times q).
Since these solutions are periodic (not chaotic), orbits of this kind in a syn-
chrotron should be of special interest. If the gyromagnetic ratio is slightly differ-
ent from two, the spin would depart from this orientation, which is the principle
behind a famous experiment for the measurement of g2 for the muon.
The equations reduce to
d

d
=

dv

d
= 2

(21)
The tensor

form the structure constants of the three dimen-

sional Lorentz Lie algebra SO(1, 2). This is analogous to the way that the Levi-
Civita tensor gives the stucture constants of the Lie algebra of rotations in Eu-
clidean three dimensional space. Denote the Lie product of SO(1, 2) as a relativis-
tic cross product
(uw)

(22)
This differs from the familiar cross product in Euclidean geometry by some
signs:
(uw)
0
=u
1
w
2
u
2
w
1
, (uw)
1
=u
2
w
0
+u
0
w
2
, (uw)
2
=u
0
w
1
+u
1
w
0
(23)
Then our equations are a relativistic version of the equation for a precessing
top:
= v F, v =2v (24)
It follows that
v v = 0 (25)
8
so that v v = a
2
is conserved; a is some number (like Plancks constant) with
the dimensions of action; the spin is of magnitude
a
2
. Also, the parameter is
proportional to proper time in this special case.
The energy
0
= E and
K =

(26)
are conserved
4
. Now,
v
0
= 2(v
2

1
v
1

2
) (27)
and
v
2
0
= 4
_
_
v
2
1
+v
2
2
_

2
1
+
2
2

[v
1

1
+v
2

2
]
2
_
(28)
Re-expressing the r.h.s. in terms of conserved quantities,
v
0
= 2
_
_
v
2
0
a
2

[E
2
K+v
0
F
0
] [HEv
0
]
2
(29)
Weierstrass elliptic function satisfy this differential equation. The remaining
components of , v can be obtained by solving the other constraint equations. It
is also possible to understand the solution as an integrable hamiltonian system, a
relativistic analogue of the precessing top. (See the Appendix)
4 The Equation of Motion for Radiating Spinning
Particles
Barut and Unal followed the method of Dirac to derive the analogue of the Lorentz-
Dirac equation of motion including the self-force. Only the equation for

is
affected:
d

d
= qF

+q
2
_

v v
_
_
2
3
v

v v

9
4
v v v

(v v)
2
_
(30)
The last term is the contribution from self-interactions, after a renormaliza-
tion. It is a singular perturbation i.e. q
2
is small and it changes the order of the
equations. It has runaway solutions, just like the Lorentz-Dirac equation. Let us
4
K is proportional to angular momentum.
9
apply to it a prescription like that of Landau and Lifshitz, to get a second order
equation of motion.
Naively, the Landau-Lifshitz prescription amounts to using
v

= 4

+4S

(31)
and replacing by its zeroth order contribution:
v

+4S

qF

= 4[v

+4S

qF

(32)
But this cant be right: for a free particle (F

=0), the radiation reaction force

should be zero. Dropping the terms not proportional to F

,
d

d
= qF

+
8q
3
3v v
_

v v
_
S

(33)
Along with the previous equations for v and S, this system of ordinary differ-
ential equations is our proposal for the equation of motion of a spinning radiating
charged particle. We must verify that the solutions are physically sensible: there
should not be runaway solutions. We will study the case of polarized orbits in a
constant magnetic eld and show that the energy decreases with time. The ulti-
mate test of our proposal must be experimental, especially since we do not yet
have a derivation from rst principles.
5 Radiating Spin-Polarized Particles in a Magnetic
Field
Now let us turn to our proposed equations for a radiating spinning charged parti-
cle. Again, we will look for solutions in the special case where the spin is polar-
ized
S

=
1
2
v

(34)
Since the radiative terms do not change the equations for v

, S

this condition
is still preserved by time evolution i.e.

S

=
1
2
v

. Unpolarized orbits are chaotic

and the radiation damping is likely to be even larger. (See Appendix)
The equations then reduce to
10
d

d
= qF

+
4q
3
3v v
_

v v
_

dv

d
= 2

(35)
We know that, v v = 0 and v v = v
2
0
(v
2
1
+v
2
2
) = a
2
.
d

d
=

+
_
a
2

(36)
where
=
4q
2
3a
2
(37)
In terms of relativistic cross products
= v [F +F v], v =2v (38)
Thus, even the radiation damping terms have a simple meaning in terms of the
relativistic cross product of the Lie algebra SO(1, 2).
Note that
H = v (39)
is still conserved and m =
H
a
continues to have the physical meaning of mass.
But energy E =
0
and K are no longer conserved. The evolution of v
0
continues
to be determined by the equation we derived earlier. Thus we get the system
dE
d
= F
0
(v
2
0
a
2
)
dK
d
= 2F
0
_
a
2
E Hv
0

v
0
= 2
_
_
v
2
0
a
2

[E
2
K+v
0
F
0
] [HEv
0
]
2
(40)
Since E and K are no longer constants, the solution is no longer an elliptic
function. Also, noting that F
0
> 0 in our convention and v
2
0
a
2
= (v
2
1
+v
2
2
) > 0,
we see that energy is monotonically decreasing.
11
Acknowledgement
We thank Gregory Bentsen, V. Sreedhar and Tamar Friedman for discussions. This
work was supported in part by a grant from the US Department of Energy under
contract DE-FG02-91ER40685.
Appendix: Hamiltonian Formalism
The Barut-Zanghi equations for a spinning charges particle follow from the P. B.
forming the Lie algebra SO(2, 3):
{v

, v

} = 4S

{S

, v

} =

{S

, S

} =

(41)
Using previous relations we can also nd the following P.B. :
_
v

, v

_
= 4S

_
S

, S

_
= S

_
S

, v

_
=

(42)
which generalizes the rotation algebra. When the third axis is ignored, this
reduces to SO(2, 2) SO(1, 2) SO(1, 2). This splitting can be seen by in terms
of the linear combinations

:
_

_
=

_
=

_
= 0
(43)
We have also the P.B.,
12
_

_
= 0
_

_
= 0
_

_
=

(44)
The hamiltonian is then
H = 2[ ] (45)
Non-Radiating Spin-Polarized Orbits in Constant Magnetic Field
Since
v
0
=
_
a
2
+v
b
v
b
(46)
and
0
= E is conserved we only have four co-ordinates in the phase space:

a
, v
a
. The P.B. are
{
1
,
2
} =F
0
(47)
and
{v
1
, v
2
} = 2
_
a
2
+v
2
1
+v
2
2
(48)
The latter corresponds to the standard symplectic form on hyperbolic space. It
is not hard to nd canonical co-ordinates
{

1

F
0
,

F
0
} {
_
2pcos,
_
2psin}, {v
1
, v
2
} {a sinh cos, a sinh sin}
so that-
{p, } = 1, p =
1
2F
0
(
2
1
+
2
2
), = arctan

1
(49)
{v
0
, } = 1, = arctan
v
2
v
1
(50)
Note that J = p +
1
2
v
0
=
K
2F
0
is the generator of rotations; it is a conserved
quantity corresponding to angular momentum. The hamiltonian is then
13
H = Ev
0

_
2F
0
_
v
2
0
a
2

(J
1
2
v
0
)cos( ) (51)
Since
pd +
1
2
v
0
d = Jd +v
0
d, =
1
2
( ) (52)
we can use the pairs J, and v
0
, as canonical variables:
H = Ev
0

_
2F
0
_
v
2
0
a
2

(J
1
2
v
0
)cos2 (53)
The equation of motion is
dv
0
d
=
H

=2
_
F
0
_
v
2
0
a
2

(2J v
0
)sin2 (54)
Thus
_
dv
0
d
_
2
= 4F
0
_
v
2
0
a
2

(2J v
0
) 4(HEv
0
)
2
(55)
which is the Weierstrass equation. Because of its polarization, this also deter-
mines the spin.
Also
d
d
=
H
J
=

F
0
_
v
2
0
a
2

2J v
0
cos2 (56)
Thus
d
d
=
HEv
0
(2J v
0
)
(57)
which determines also in terms of elliptic integrals.
To determine the position variable, we note the that
d
d
_

= 0 (58)
This leads to the conserved quantity (corresponding to translation invariance)
X

(59)
14
Thus (in a co-ordinate system in which F
0
> 0, F
a
= 0 and X
c
= 0) we have
x
b
=
1
F
0

bc

c
(60)
In other words, knowing
1
,
2
is the same knowing the spatial orbit, except
for a rotation by 90

and a scaling by F
0
. In polar co-ordinates this amounts to
knowing p = J
1
2
v
0
and as functions of , a kind of elliptic curve:
_
dp
d
_
2
= 2F
0
_
4(J p)
2
a
2

p(H2EJ +2Ep)
2
,
d
d
=
HEJ
2p
+
E
2
(61)
It is not correct to interpret the x as the position of the electron: due to the
phenomenon of zitterbewegung. Even for a free particle, x

() is a spiral and not

a straightline. Barut et. al. have shown how to recover a center of mass variable
for the electron from this complicated solution.
A Symplectic Reduction of the General Planar Orbit
The general case of planar but not spin-polarized orbits does not appear to be
integrable. It corresponds to chaotic motion of an electron in a synchrotron. We
can use the conserved quantities

0
= E, a
2
, b
2
(62)
to reduce it to a conservative system with three degrees of freedom: the phase
space is a plane (
a
) times a pair of hyperboloids parametrized by (

). It is
easy to check that
{p, } = 1 (63)
with
p =
1
2F
0

a
, = arctan

1
(64)
Also,
15
{
0
, } = 1, = arctan

1
{
0
, } = 1, = arctan

1
(65)
In terms of these
H = 2E(
0

0
) 2(
a

a
)
a
= E(
0

0
)
_
2F
0
_

2
0
a
2

pcos[ ] +
_
2F
0
_

2
0
b
2

pcos[ ]
(66)
Another conserved quantity is the generator of rotations
5
J = p+
0
+
0
(67)
The reduced systemhas three degrees of freedomand two conserved quantities
H, J : just one conserved quantity short of an integrable system. When
0
= a
(or
0
= b ) one degree of freedom decouples and we get an integrable system.
Circular orbits correspond to the case = ,
0
+
0
= constant.
It may be more convenient to use = , = , as the independent
position variables. Then
p

d +p

d +p

d = pd +
0
d +
0
d = pd +
0
d[ +] +
0
d[ +]
(68)
p

=
0
p

=
0
p

= p+
0
+
0
= J
(69)
5
This is linearly related to the quantity K we introduced earlier:
K = 2F
0
J
16
Then we get a system with two degrees of freedom with hamiltonian
H =E(
0

0
)
_
2F
0
_

2
0
a
2

[J
0

0
] cos+
_
2F
0
_

2
0
b
2

[J
0

0
] cos
(70)
This appears to be a chaotic system. In general, the information about the
polarization of spin in the initial condition will be lost rapidly and the particle
beam will look unpolarized. For the special case
0
= 0 = b the spin remains
polarized and solutions can be obtained in terms of elliptic functions as we saw
above.
References
[1] J. D. Jackson, Classical Electrodynamics, Wiley(1998); F. Rohrlich, Classi-
cal Charged Particles, Westview Press (1994);
[2] P. A. M. Dirac, Proc. Roy. Soc. Lond A167 (1938) 148
[3] L. D. Landau and E. M. Lifshitz, The Classical Theory of Fields
Butterworth-Heinemann (1982); especially Section 76
[4] H. Spohn, Europhys. Lett. 50 287 (2000); H. Spohn, Dynamics of Charged
Particles And Their Radiation Field , Cambridge University Press (2004)
[5] S. G. Rajeev Ann. Phys. 323, 2654 (2008)
[6] V. Bargmann, L. Michel, and V. L. Telegdi, Phys. Rev. Lett. 2, 435 (1959)
[7] A. O. Barut and N. Zanghi, Phys. Rev. Lett. 52, 20092012 (1984)
[8] A. O. Barut and N. Unal, Phys. Rev. A 40, 54045406 (1989)
[9] Y. Hadad, L. Labun, J. Rafelski, N. Elkina, C. Klier and H. Ruhl,
arxiv/1005.3980
[10] S. G. Rajeev, Ann. Phys. 323, Pages 2873 (2008)
[11] L. D. Landau and E. M. Lifshitz, Mechanics, Butterworth-Heinemann
(1982)
17