B H. J. Bx ABHA, PH. D .
IT is well known that the heavy particles, namely the proton and neutron,
have an interaction with the meson field characterised by the " charge " g l
and an explicit spin moment g 2 . The latter is known to lead to a scattering of
neutral mesons which increases as the square of their energy, and it has
often been thought that in consequence this term would lead to the type of
explosions investigated by Heisenberg. It is clear, however, that in quantum
theory, the reaction of the emitted or scattered radiation has not been taken
into account, the effect of which would be to diminish the scattering. This
radiation damping cannot at the moment be included properly in a quantum
theory, but one can approach the problem if one can build up a complete
classical theory including radiation damping. The point at which radiation
damping begins to become important in the classical theory may then be
taken to indicate the point at which one might expect the quantum theory
to fail due to its neglect of this factor. Indeed, the well-known limitation
of the correctness of the quantum theory of the electron upto energies of
roughly 137 m (the velocity of light is put equal to unity in this paper)
is based on a comparison with the classical theory of the electron by
Lorentz, in which radiation damping became important at these energies.
An attempt at a non-relativistic treatment of damping has already been
made by Heisenberg (1939) for a dipole of finite extension by a method which
is analogous to that of Lorentz. His results do not agree with those of this
paper even in the limit of a point dipole and we believe that this discrepancy
is due to the method being inherently unsound. It consists in taking retarda-
tion into account for the field produced at one point of the dipole by a portion
of the dipole at another point, while the different parts of the dipole are
assumed to move together simultaneously. In the case where the field is
the Maxwell field which propagates itself with light velocity, this procedure
is inherently absurd, for it assumes that different parts of the dipole know
what the other parts are doing in a time far less than that required for light
to travel from one point to the other.
In this paper therefore we shall deal with a point dipole with no extension,
and it will be shown that completely consistent relativistic equations free
247
Al F
248 H. J. Bhabha
from singularities can be found for this case. For simplicity we shall make
the assumption that the dipole has only a moment g 2i but no charge, i.e., we
shall put g l = 0. The dipole will be characterised by an antisymmetric
six-vector S µy , and its velocity will be denoted by v,,. In this paper we shall
restrict ourselves to the case where the equations are so chosen that S = 0.
The dipole then always remains a pure magnetic dipole in the rest system.
The general case where g 2 * 0 and S is variable will be dealt with in
another paper with Corben. Equations for the case where S = 0 can be
derived at once from the general treatment by a procedure which is given
there, but the general theory is so much more complicated that there is point
in giving the derivation of these equations by a direct method.
We shall further restrict ourselves to a Maxwell field. It has been shown
in an earlier paper (Bhabha, 1939 a) that the finite rest mass of the meson
introduces no essential difference from electromagnetic theory for high ener-
gies, so that the results of this paper will also give a good indication of the
behaviour of neutrons in a meson field. The exact treatment of the meson
field will be carried out in the paper with Corben.
General Theory
We shall follow a method first used by Dirac (1938) in his classical theory
of electrons. Take the world-line of the dipole as given, and also the direction
of its spin at every point of it. The co-ordinates of the dipole denoted by
z,. (-r) are functions of the proper time T measured along the world-line from
some arbitrary point on it.
All calculations in this paper will be done assuming that for all time
Shy vy = 0. (1)
The equations derived by differentiating (1) are
Sµ ' + S µy vv = 0
SK," +
2S µ11 vv + Sp,, 7111 = 0 (2)
S / " +
3 S, 7J9 + 3S /, y vv + S =0
S / ,iii + 4SN,,,
71v + 6 7Jy + 4S -- SN, y vv = 0,
where we have written for brevity
S/hri _ Sv^v vy Sµviri _ S S/hiiir _ S (3)
Sµ - SEr,v vv,
' , vv
We shall then have to show that the equations of motion are such as to
preserve the equation (1) if it is valid initially.
Assume the Maxwell equations to be strictly valid in the whole of space
except on the world-line. The retarded field due to the dipole at any point x4
Classical Theory of Spinning Particles 249
is then determined by the Maxwell equations and is given by
ret ret ret
4
^µv 6xIA^v — ^ v ^p
sA, is the distance from the actual point to the retarded point determined by
Sp, s'L = {xµ — z,, (T0)} {xp' — z/4 (T0)}
0• _ (6)
In performing the differentiations in (4) and (5) account has to be taken of
the corresponding change of the retarded point. The method has been
given in the previous paper (Bhabha, 1939 a).
The actual field F, at any point is equal to the retarded field due to the
dipole plus the ingoing field.
ret in
Fµv = Fµv + F µv' " (7)
The energy momentum density tensor T,,, of the field is
T /AV ` F FP + 4 gµv I I p° ' (8)
The metric tensor ^ is taken to have the form goo = 1 , g 11 —g.22 =
933=— 1 , with the other component vanishing.
We now introduce as in a previous paper (Bhabha, 1939 b) a tensor to
represent the angular momentum density of the field, defined by
9 ( )
dM "
d V X ' = 0.
For the conservation of energy and momentum, this must only depend on
the conditions at the ends of the tube, so that the integrand must be a
perfect differential. Therefore we must have
T= A. (13)
A dot above a letter represents differentiation with respect to the proper
time r. For conservation of energy and momentum, the dipole must so
translate that (13) is satisfied; in other words, (13) determines the trans-
lational motion of the dipole.
The flow of angular momentum into the tube is given by
f M xlzy dS°. (14)
Using (9) and the definition of s, it can be written
f M, = f T 2 dT (M 1,+ z ?, A —
T1
For the conservation of angular momentum, this must only depend on the
conditions at the two ends of the tube, so that the integral must be a perfect
differential. Put it equal to B ay + dT (z A^ — z,,, AA ). This leads to
M y, µ — (v x A, — vµ A X) = B. (15)
Thus, for conservation of angular momentum, the dipole must so rotate that
(15) is satisfied. In other words this equation determines the rotation of the
dipole. To give the equations for the translation and rotation of the dipole
explicitly we must determine A. and By,µ.
Classical Theory of Spinning Particles 251
It is convenient from now on to adopt the following notation. The
invariant formed from two tensors A B will be denoted by (A B)
(AB) = A Bµ°
A 2 = (AA) = Apv AAP.
The invariant formed from any combination of tensors and vectors will be
written in the usual matrix notation. Thus, for example
(V S S Sv) 7Jµ S AU S QP S P ° v y
S AP SP ° S Q „ = 0 (22)
D'-' S AP S P A S,^ 4 (D IIµ S,) (S 2 )
Here D ?, N, is an arbitrary antisymmetric tensor. But these equations which
can be verified in this particular frame of reference are already in tensor form,
and hence are true in any co-ordinate system.
We now turn to T. Due to the quadratic form (8) of T and the split-
ting of Bµ „ into two parts by (7), it can be written as the sum of two parts
if we neglect terms that vanish when E —> 0.
ret mix
Tµ =T,,,+T µ .
ret
T^ consists of the terms which are multiplied by g 22 and does not contain
mix
the ingoing field while T contains the ingoing field and is proportional
to g 2 . Corresponding to this we may write AA, in three parts.
ret mix mech
Aµ =Al,,+A µ +A µ .
The first is proportional to g 2 2 , the second to g 2 and contains the ingoing
mech
field, while A w corresponds to the mechanical part of the energy and
does not contain g 2 .
Introducing
self ret , ret
T g -T —A /L , (23)
(13) then becomes
self mix , mix . mech
T µ + Tµ — A A = A µ . (24)
Similarly Mx / consists of two parts
ret mix
M7, A = 1,A A + M
1
selfin in in in /
Mlµ +g2 Sx ° Faµ — Sµ, ° Fax + V X S µ° V v^v — v SA ° F v v ( 31)
—I (vX S µ — vµ S^ ' ) = I SA µ .
'
self
Except for the term Mx,., which we have to investigate, (31) is just of the
ret ret
required form (21). We see now that A p, and BAµ have to be so chosen
self
that M aµ is of the form
self
(F pto- si =F PS SP°
we get
vµ T
self + 2 g2 (F sp a ) + g2 vµ S pa B°° v v = I (vfF s • (33)
(vv) = 0
(vv) + v 2 = 0
(vv)-} 4(vv)+3v==0.
Contract (31) with — (S'PS p µ — SµpS p ^`). Using the relations (22) and
remembering (1), (2) and (32) it follows quite easily that
1 2 922 (D ' , v\l 2 g2 (F inpv S pQ ) + -2 92 (vX F... v v — va F'avvv) AQ
Hence in order that (33) and (35) should be consistent, we must have
self
vµ T, = 4g 2 (DS). (36)
mix
We now see that the choice of A v, was unique. The first term in it
was chosen so as to change (28) into the form of the last two terms on the
mix
right-hand side of (21), while the second term in A. had then to be such
as to make (33) and (35) consistent for the term containing g 2 .
mech mech
The choice of A,,, and B aN, is not unique. There are other possi-
bilities which will be discussed in the paper of Bhabha and Corben. The
choice (28) is equivalent to treating the spin as a pure gyroscope with no
mechanical moment of inertia perpendicular to the axis of the spin. M plays
Classical Theory of Spinning Particles 255
the part of the mechanical mass. Both M and I may be given any values
we please, and there is no connection between them.
The equations (30) and (31) neglecting the damping terms have already
been derived by Frenkel (1926) by another method.
D pµ vµ = 0,
and
ret .ret 1
V ^T µ — A A ) 2- (Dx A J (36)
It will appear presently that these two conditions are so stringent as practi-
ret ret
cally to determine A. and B AA uniquely.
As is well known, the retarded field of a dipole has a singularity of the
ret ret
order e. Hence both T µ and MA, will have singular terms which tend to
infinity as e --)- 0. We therefore write
ret (e) (0)
Tµ, = g2 2 (Tµ + Tµ ) (37)
(E)
where T all the terms which are singular.
Similarly we write
ret (e) (0)
Mx µ = g 22 (MxA + M (38
38 ))
(E) ret
where M aµ contains all the singular terms in M.
In the paper of Bhabha and Corben* it is shown that
(E)
E2s v µ S2 — 1 5 Sµo s)
Tµ d'r rE s v µ S2) +
E ^v µ ^ 5 v2 S2 —' 3r S2
15 „/2 J 15 Sp? ^' aY v Y + 5 ^^p,0 ^'/^ 1 J (39)
(40)
*v ^µ^ S+ { 25 vA S ` s Y 5 Va S
J
The minus sign at the end of a bracket means that similar terms with A and
µ intercharged have to be subtracted.
ret (E)
Corresponding to (37) we split Aµ into two parts, of which A A contains
the singular terms
ret (E) (0)
A µ = g2 2 (A M, + Aµ ) • (41)
(e)
A µ is chosen so that
T(, _ A(E)= 0.
µ µ (42)
The fact that (39) is a perfect differential makes this possible. Thus by (23)
self (0) . (0)
T µ = g2 2 (T µ — A µ ), (43)
and is now entirely free from singularities. It remains finite when E —). 0.
ret
(40) shows that all the singular terms in M aµ are not perfect differ-
entials. However, the terms in square brackets in (40) which are not
perfect differentials are just sufficient to compensate the singular terms
(e) (E)
— ( V A AA — A1)
in (32 a). Thus we may write
Tet (e) (0)
B Aµ = g2 2 (B A ,., + BA,.) (44)
(0) (e)
where B kµ has no singular term and B Aµ is just equal to the term in curly
brackets in (40) which is a perfect differential. It then follows that
self (0) (0) (0) . (0)
+ 2 (d + 3) ( S vµ — S 22 (Sr" v, — Sµ" v)
5
vY ) (46)
5k1 vanishes, then S,j also vanishes. Thus in order that Skl should continue
to be zero in the absence of an external field, we must have initially
vµ = 0 , S, = 0, Ski = 0.
It is shown in the appendix, that if further v µ vanish, then the trans-
lational equation (30) requires that V. should vanish. vµ is still undetermined.
Thus, in order that in the absence of an external field the dipole should
continue in a state of uniform motion, it is necessary that initially vµ , -µ
and v should all vanish. And the same condition for the rotational equa-
tions requires that S, and S N ,, should vanish. Thus the solution of the
equations of motion (30) and (47) is completely determined in any given
external field if initially the velocity and direction of the spin are given,
and finally, when the external field was died down, z, v,,, v, S N„ and S µ „ are
made to vanish.
Lastly we give the equation (33) explicitly.
1 p2 (.r n C^ — g2 ( s' rl v) p22 { 3 ( c+) _ (s' s") + 3 S2 v2
+td— 2 —2 (d -^-- 3) (^ 7 s v
3 (S" S
! + ( d --1 s'2 v J
\ +d(S'Sv)}=0. (49)
The damping term in equation (48) can be derived directly from quite
elementary considerations. It must clearly be proportional to g 22 , and
Classical Theory of Spinning Particles 259
The average work done per unit time by this field on the dipole, which equals
the total energy radiated, is
f, gHo si ne B sin
g2 (F^) = ,g2 (HM) = g2w (M1H0) cos Wt =_ -- - - (55c)
Dividing this by H 02 18,r and using (55 b) we get the scattering cross-section
w2 Sin2 0
67r 9 I
2 4 (56)
4 V 24+
To find the intensity scattered in different directions, we must calculate
the retarded field due to the oscillating dipole. It can easily be shown by
(4) and (5) that at large distances the electric field E' and the magnetic field
H' at a point r are
r•M Lr
y2 ^ret H' — 92 L r M I ret . )
_ - ba [ ^y3 ] J ( 57 )
92
2 )I4 2 Sln2 O _ 2w4
a2
M1 2 sin g
(57 a)
4Tr 72 8a y
a = I , fl' H
ag2 2age
The work done on the average by the incident light on the dipole per
unit time according to (55 c) is
g 2 fi' H„ sin2 0 sin Ho w 2 sin 2 B
--- 2 \/ i ,I - - — 2a (a 2 ± W4 )
On the other hand the total energy radiated by the dipole is given by (57 a)
integrated over all directions of emission.
and thus decreases inversely as the square of the frequency just like the
scattering by a point charge. (56) thus shows that radiation damping will
be important when
A w A 3I(.
2(60)
^2g2
This formula may be applied to get a rough estimate of the point at
which the quantum theory of the interaction of neutral mesons with neu-
trons (or protons) may be expected to fail due to its neglect of the effects
of radiation damping. In applying (56) which has been calculated for light
to mesons, it should be remembered, that, as has been shown in an earlier
262 H. J. Bhabha
paper (Bhabha, 1939 a), the only effect of the finite mass of the meson is
to diminish the cross-section (56) for small It w — µ while for /t w > µ no
appreciable difference is introduced. Thus, putting the empirical values
I =h/2, g 2 =g2'Jtfµ, g2'2 //t— 1/15 (60) becomes
h u, µ 3µ. (61)
V 4 9 2 ' 1'
Thus the quantum theory does in fact fail where one expected it to fail, but
this limit has really nothing to do with the mass of the meson, and is only
determined by the empirical magnitude of the constant g 2 . The limit (61)
results in a very unusual state of affairs. We should not expect quantum
effects to come in till the momentum of the mesons becomes comparable
with the rest mass of the neutron, i.e., till It w ' M. The classical formula (56)
will therefore be valid upto this limit. Thus there is a region of frequencies
3 /t w M where the classical formula (56) is still right, but the quantum
formula (59) quite wrong.
It is well known that the electron has no explicit g 2 interaction with the
radiation field. If, however, the above theory be applied to its spin to find
the limitations of the quantum theory due to the neglect of radiation
damping, we find, putting I = Ii/2, g 2, = elt!2m
^w ^ ,^ / 6e µ ^ U6 x 137 µ•
— i5S 21)
1 2 — 3g
2 4 v B
S2
} VM( Y S S µv *^Pv p S ^^v 2s v
5µv SPv
P
2S µ + 2
T `8 pP SPX + 2 SX" S S' Sµv 71v — S V2],
The terms which are perfect differentials in (63) can be eliminated at once
(0)
by introducing corresponding terms in A. The only effect of this is
therefore to change the coefficients of the corresponding five terms in (64)
(0) (0)
due to the presence of — (v ? A v µ A x ). In the subsequent work we will
consider this to have been done.
(0)
Now A X must be proportional to g 22 , and hence it must be quadratic in
Sµ ,,. Further, from dimensional reasons, each term must have three dots over
(0)
it, since each term of T µ has four dots over it. Hence remembering (1) the
(0)
most general substitution for A A is
(0)
(S' S”) + C 3 9J µ (S ' S v) + C4 7l µ S 2 (v v)
A µ, C1 v µ (S .S) + C 2 v µ
C5 vµ S2 Cs vµ S 2 .' C ? v µ S2 v2 +c8 vµ S2 Cy ^ µv S v
(0) (0)
The expressions for T Mx p, were calculated independently by each of
us as a check.
261 H. J. Bhabha
The c's are arbitrary numerical coefficients. Any other terms of the required
type can at once be absorbed into the above expression by the use of the
relations (2) and (16 a). Similarly, it can be seen that the most general
(0)
expression for B x ^,, is
BAµ
()
_ [d1 vA S,,, + d2 vA SAP S „ v + d3 Z,,k SAv S , y
+ d 4 vA S
+ d 5 vA v A S 2 + ds Sap vP S Pµ _ (66)
Using (64), (65) and (66), we now form the expression
self (0) (0) (0) .(0)
The various terms which occur in this expression are given in the accom-
panying table with their respective coefficients. The coefficients are so
(0)
;,X ^ e S2 —d5 —6 =0
V\ S A5 S" — C1 2— d2 =— c12 +
vX S µ „ S'° — c 9d3 + r5 =0
v\ y 8 v P vP — C11 — d3 — 6t4 — 15 = 0
vX S vP — C14 — d4 + 35 -- C 14 — dl — 15
v x vµ S2 —05 -- g =_= O
V S'2 —C° +2 =0
v \ V N, S 2 - 2 — Cq +5 `_ O
— C15 + 3 = C1 5 + 3
V\ S,u , S IP v2
v\ vµ S2 —c8 —d5 =O
SxP SP —d6 + 2 =0
S ?, P SP A —d 6 =—2
S\ "S' u 3 =—d 2 — 15
S' S µ „ v" I _ — d1 + d 4 + 5
S^, P SP µ v2 —2 -- — 3
Classical Theory of Spinning Particles 265
spaced in the table as to show the respective contributions of the different
parts. The first column gives the contribution to the coefficients due to
(0) ro^ rol
— (v a A v1 AX ), the second column that due to B x ,, and the last column
(0)
that due to M x given by (64) with the changes mentioned above. Some of
self
these terms cannot possibly occur in M xµ , since it has to be of the form (32).
The coefficients of all these terms must therefore be equal to zero, and this
at once gives us a number of relations between the above c's and d's. For
example, the coefficient of S^, p SP, is — (d6 — -) and this must vanish, from
which it follows that ds = s . The coefficient of Sx p SPA is then uniquely
determined.
Further, the term — d lv x S,, V S'° cannot occur by itself, since it is not
of the correct form (32). It can only occur in the form
— dl S vA — SA ' v").
By (2), the last part of this becomes — d 1 S a ' (S p " + 2 Sµ ,, vv). Similarly
the term — (d l + d 2 + 36) v X Sµv S"' can only occur in the combination
— (d 1 + d 2 + r5) S (5 ,,, v a — S"A v").
The last part of this can be put in the form — (d 1 + d 2 + i; ) S X " S^'.
Thus the actual coefficient of Sr" S N,' must be equal to the sum of the
coefficients of this term as derived in the above way. In other words
3 =d 1 —(d 1 + d2 +) — —(d2 + Y5)•
This at once determines d 2 . Similarly it can be shown for the coefficient
of S x ' 5.. v" that
_ —d l -f-d a -I-S
which at once determines d 4 in terms of d 1 . Proceeding in this way, we can
self
reduce M a to the form
self /
Maµ = g2 2 (Sa p DP — S p DPA).
The expressions to which these coefficients must be equal are given in the
self
last column of the table. Contracting M X, with v and remembering (1)
we get,
self
MAp' V P =g22 S^v [(d1 — c15 + s) S' v2 — (C 12 -- 15) S"
e4 = 15 c° = d + 3 C14 = — d —1
C5 =— 8 X10= d + 6 X15 -= d +
- (d + ) S µ vs y — (d+ 2 ) S µ , SP v P —2 (d -}-1) S S” _ (d + 3) S
+ 2 (a! is)
—
Sµv S"P vP+ (d+is) SµvS°P v P — ( d+ ) S µ v S'v v2
—a! Sµ, S ,,,, v 2 — (d + 1) S AS, S"P vP v 2 — 2 (d+ ) SA , S'" (v v)] .
Classical Theory of Spinning Particles 267
There is no particular value of d which simplifies this expression consider-
self
ably. Perhaps d = effects the greatest simplification both here and in M aµ
self
When v w = v µ = 0 and S xµ = 0, T µ reduces to
2
g22 L — is vµ S2 Sµv S v
All the other terms in equation (30) also vanish in the absence of an external
field, so that this equation requires that the above expression shall be zero.
Using the last of equations (2) it becomes
I — i5 Vµ 8 2 + — 5 S µv SP v] = 0.
On contracting this with S^,w and using (22) we get
(_ is S Aµ vi') S 2 = 0
from which it follows that the second term in the previous equation
vanishes, and hence v µ = 0, which is the result quoted in the text.
Summary
The exact relativistic classical equations taking radiation reaction into
account for the rotation and translation of a point dipole are given for the
case where the dipole is always a pure magnetic dipole in the rest system.
These equations are entirely free from any singularities. It is shown that the
mass M, angular momentum of the spin I and magnetic moment g 2 are three
entirely independent constants with no connection between them. The cross-
section for the scattering of light by a dipole is given by formula (56).
This formula shows that due to radiation reaction the scattering actually
decreases as w 2 for very high frequencies w, instead of increasing as a
when radiation reaction is neglected. The quantum mechanical formula
for the scattering of neutral mesons by neutrons is shown to go wrong at
energies h w '^ 3 µ due to neglect of the effects of radiation damping. The
classical formula (56) can still be correctly applied in the range 3 µ < A w
< M, where the quantum formula is wrong, M being the neutron mass.
Finally reasons are given for thinking t ha t the quantum theory of the
electron fails at energies above about 1/3 x 137 m due to neglect of the
effect of radiation damping on the spin, and the quantum theory of the
meson and its inter-action with the electromagnetic field at -/6x17 µ.
REFERENCES
Bhabha .. Proc. Roy. Soc., (A), 1939a, 172, 384-409.
Proc. Ind. Acad. Sci., (A), 1939b, 10, 324-32.
Dirac . .. Proc. Roy. Soc., (A), 1938, 167, 148-69.
Frenkel .. Z. f. Phys., 1926, 37, 243-62.
Heisenberg .. Ibid., 1939, 113, 61-86.
Kramers .. Physica, 1934, 1, 825-28.
Pryce .. Proc. Roy. Soc., (A), 1938, 168, 389-401.
A2 . s