Eur. I. Phys. 2 (1981) 168-170.

Printed in Northern Ireland

168

The Feynman paradox revisited

J M Aguirregabiria and A Hernandez
Departamento de Fisica, Facultad de Ciencias, Universidad del Pais Vasco, Apdo 644, Bilbao, Spain

Received 23 July 1981

Abstract We propose a simplermodel in order to Laburpena Eremu elektromagnetiko estatikoen

facilitate calculations of the Feynman paradox concern- momentu angeluarrari buruzko Feynman-en paradoxa-
ing the angular momentum of a static electromagnetic ren kalkuluak errazteko, eredu sinplifikatu bat aurkez-
field. Whenan angular momentum is attached to the ten da. Horrelako eremuei momentu angeluarra egokit-
static electromagnetic field the paradox disappears. The zen bazaie, era kuantitatiboan ikusten da paradoxa de-
storage of the angular momentum in the field during sagertu egiten dela. Dispositiboaren eratze-prozesuan
the assembling process is also analysed, zehar, momentu angeluarra eremuan nola metatzen den
ere aztertzen da.

1. Introduction
It is well known that, for systems of particles satis- when the current falls off is easily checked. Finally,
fyingNewton’ssecondandthird laws, thetime $4 is devoted to a discussion emphasising that the
derivatives of the total linear and angular momenta electromagneticangular
momentum does
not
areequal,respectively,tothenetexternalforce satisfy the superposition principle.
and torque. For isolated systems we conclude that
thetotallinearandangularmomentaarecon-
served.However,theseconservation laws canbe 2. Feynman paradox: simplified version
consideredasindependentprinciplesforisolated Let us consider a thin circular plastic disc of radius
systems, owing to their relationship with the sym- a lying on the X Y plane of a reference frame, such
metries of the system. that its onlyallowedmotion is rotation about the
If theinternalforces of thesystemsare of an OZ symmetry axis. At the centre of thedisc is a
electromagnetic nature,
Newton’s third law no small hole in which a small circular ring is placed
longer holds, and, if the conservation principles of centrally so that it is free to movewith respect to the
linearandangularmomenta(andenergy)must disc.
remain valid, these magnitudes must be assigned to Let us assume that the ring is made of a super-
theinteraction field,giving risetothePoynting conducting material and that a constant current of
theorems in electrodynamics and to the formalism intensity Z is flowing in it, giving rise to a magnetic
of the energy-momentum tensor (Mbller 1972). moment m = mk, k being the unit vector along the
However, in thestaticelectromagneticcase, it 0 Z axis. The ring is to be assumed small enough to
seems ‘intuitively’ difficult to assign linear and an- be considered point-like.
gularmomentato fields thatdonotchange with A charge q is located at the point with coordi-
time. It will be helpful for pedagogical reasons to nates (a, 0,0) on the edge of the disc. This system is
analyse some examples in order to show the coher- initiallystatic:the fields donotdependontime,
ence of thegeneraltheory.Some of themhave and there are no forces on the charge or the ring.
been given by Romer (1966, 1967), Pugh and Pugh Let us assume that the magnetic moment starts
(1967) and Corinaldesi (1980). The recent experi- decreasing slowly atthe small rate m =dm/dt,
ments of Graham and Lahoz (1980) have revital- owing to, for instance, a small increaseof tempera-
ised the subject. ture; this gives rise to the appearance of a non-zero
In thisarticle we makeadetailedstudy of a resistivity in thering.Weshallacceptthatthis
simplified model of theFeynmanparadox(Feyn- process is slow enoughtoneglectradiationand
man1964). relativistic and retarded effects.
The paradox is presented in 32. In 33 the elec- The
electric field induced by the changing
tromagneticangularmomentum is computed;the magnetic field acts on the charge, causing the disc
fact that this momentum is transferred to the disc torotate.Thevectorpotential A created by the

0143-0807/81/030168+03$01.50 0 T h e Institute of Physics & the European Physical Society

The Feynman paradox revisited 169

ringat a point r is, accordingtoPanofskyand Poynting vector (and also the remaining self-terms
Phillips (1975), of the energy-momentum tensor); in order to take
theseintoaccount,usehastobemade of the
Lorentz-Dirac equation.
Because B = V X A, (5) can be rewritten as
When m changes,theinducedelectric field E =
-aA/at has a tangentialcomponent,atthepoint rx[Ex(VxA)]dV.
where the charge is, given by /R (6)
Using the vector identities
Cx(VxD)+Dx(VxC)=(V.C)D+(V.D)C
The torque exerted by the electric force along the + V * T (7)
0 2 axis is rxV*T=V.R (8)
where
T,, = ( C . D ) S , , - ( C , D , + C , D , )
such that the angular momentum of the disc when and
themagneticmomenthasfallentozero is Lzk, R , = EF‘x~T,,
where
Eikl being the Levi-Civita completely antisymmetric
tensor,andusingGauss’stheorem in differential
form, (6) can be expressed as
But, if ‘initially’ theangularmomentum of the
system is zero and must be conserved, the disc must
remainstatic.This is in brief theessence of the
L,,= jR3 ( r x p A ) d V + E O[ ( V * A ) r x E ] d V
/R
paradox proposed by Feynman.
V-QdV.
+cOJR3

3. Explanation of the paradox Here p is the charge density and

It can be argued that at the beginning the angular
momentumassociatedwiththechargecarriers of
thesuperconductingringmustbetakenintoac- Thevectorpotential A given by (1) is di-
count. Stedman (1981) has remarked that thisis the vergenceless, since itsatisfies the Coulomb gauge,
case in someactualexperiments,suchasthat of andthesecondterm in (9) vanishes.By’Gauss’s
Graham and Lahoz (1980). However, in our model theorem, the third term in (9) appears as
this is not relevant, since it has to be compensated
by an external torque on the ring in order to leave
V * Q d V = lim dS-Q (11)
the ring at rest. There is another basis on which to L 3 R-x [(R)
rejectthisinterpretation:wecan, in fact,change
the sign of the charge and reverse the velocity of where S(R)is the sphere of radius R centred at the
thechargecarrierswithoutmodifying (4), but origin. It is easy to see that this integral vanishes
changingtheangularmomentum of thecarriers. because Q falls off as R - 3 while d S increases as R‘,
Similarly, in more general terms, we can change the when R goes to infinity.
charge-mass ratio of the carriers. Thus, in princi- In our problem p = q8‘3’(r - a ) , 8(3)(r-a) being
ple, we neglect the angular momentum contribution Dirac’s delta function, where a is the charge posi-
of the charge carriers in order to give an explana- tion vector, and then
tion of the paradox.
The principle of angular momentum conservation
can still hold if the electromagnetic field in vacuum
has an associated angular momentum with respect
L,,= L3 ( r x p A ) d V = q 5R 3
= qa x A ( a )
(r~A)~?(~)(r-a)dV

(12)
to the origin given by
i.e.

where S = (l/po)Ex B is the Poynting vector and c

is the speed of light in vacuum. At the end of the which coincides with (4).
process, when m = 0, L,, is also null, since the only Inthequasi-staticlimit,thelinearmomentum
remaining fields are those created by the moving associatedwithanarbitraryelectromagnetic field
charge whose evolution is governed by the Lorentz can be written, according to Calkin (1966), in the
equation, which excludestheself-terms of the form
170 J M Aguirregabiria and A Hemcindez

.-=IR3 E O ( E X B ) d V = j R 3p A d V (14)
which is null on the OX axis, and in particular at
the point where the ring is, so that no torque is
acting on it. However, the charge at the point x will
which according to (12) could be interpreted as be affected by a magnetic force according to
being PA, the linear momentum density in quasi-
static situations. But this interpretation is wrong as
far as angular momentum is concerned, because the
surface integral (11) does not in general vanish. In It is interesting to observe that in the charge-ring
our particular case it has vanished because, when R interaction Newton’s third law does not hold.
goes to infinity, A decreases as R-’. In order to move the charge at a constant speed,
Furry (1969) has calculated the angular momen- some external force must be applied on it, F,,, =
tum of a system consisting of a small magnet of -Fm, and its torque with respect to the origin will
magnetic moment m, placed atthe origin, and a be
point-like charge q located at r = a. He makes use
of an expansion in terms of Legendre polynomials,
giving a result which reduces to (13) when m = mk
and a = ai. To arrive at (12 ) from (9) we have made since r = xi.
use of a way of reasoning like that of Panofsky When the charge has reached the point ai, ac-
(1975). cording to the principle that the external torque is
In the preceding calculus, the ring has been equal to the time derivative of thetotal angular
considered strictly point-like because of the factor r momentum, there is stored in the electromagnetic
in expression (5). Otherwise, in order to compute field an angular momentum
the linear momentum, the exact form of the fields
in the neighbourhood of the ring must be taken
into
account, as has been remarked by Furry
(1969). which coincides with (13).
It should be remarked that (17) is independent of
U , justifying the assumption that the charge stops
4. Comments and discussion when it arrives at the disc.
One could object that, if there were present only a Finally, the result (17) is independent of the way
charge or only a magnetic dipole, the angular of assembling the system, if radiation is neglected,
momentum of the correspondingelectromagnetic ascan be shown by means of the formalism of
field would be zero; and even if both parts were conservation laws expressed in terms of the energy-
present, it would also be null because of the ‘super- momentum tensor.
position principle’.
However, the superposition principle does not
References
apply here, since Poynting’s vector and the elec- Calkin M G 1966 Am. J. Phys. 34 921
tromagnetic energy-momentum tensor are not Corinaldesi E 1980 Am. J. Phys. 48 83
linear in thetotalelectromagnetic field, and thus Feynman R P 1964 The Feynman Lectures on Physics
this principle does not hold for the electromagnetic v01 I1 (Reading, Mass.: Addison-Wesley) 917.4
angular momentum. Another intuitive wayof con- Furry W H 1969 Am. J. Phys. 37 621
vincing oneself of this and of the result (4) is to Graham G M and Lahoz D G 1980 Nature 285 154
follow the assembling process of the system, in a Mdler C 1972 The Theory of Relatiuity 2nd edn (Ox-
similar way to that developed by Calkin (1966) for ford: Clarendon) ch 5
Panofsky W H and Phillips M 1975 Classical Electricity
the electromagnetic linear momentum. and Magnetism 2nd edn (Reading, Mass.: Addison-
In fact, let us assume we have initially the ring Wesley) pp 133, 97
and charge infinitely far apart, so that no angular Pugh E M and Pugh G E 1967 Am. J. Phys. 35 153
momentum exists, and we move the charge from Romer R H 1966 Am. J. Phys. 34 772
infinity to the disc edge at a constant speed along - 1967 Am. J. Phys. 35 445
the OX axis. The chargecreatesa magnetic field Stedman G E 1981 Phys. Lett. 81A 15