Feynman Cylinder Paradox

John Belcher
Division of Astrophysics, Massachussetts Institute of Technology, Cambridge, MA 02139
Kirk T. McDonald
Joseph Henry Laboratories, Princeton University, Princeton, New Jersey 08544
(1983; updated May 25, 2015)

1 Problem
An infinitely long wire with linear charge density −λ lies along the z axis. An insulating
cylindrical shell of radius a and moment of inertia I per unit length is concentric with the
wire, and can rotate freely about the z axis. The areal charge density on the cylinder is
σ = λ/2πa and is uniformly distributed.

The cylinder is immersed in an external magnetic field Bexẑ, and is initially at rest.
Starting at t = 0 the external magnetic field is slowly reduced to zero over a time T  a/c,
where c is the speed of light. What is the final angular velocity ω of the cylinder?

2 Solution
This problem is a version of the Feynman disk paradox [1]-[17] that is particularly easy to
analyze. However, it avoids a subtle point related to the return flux of the external magnetic
field, as discussed in sec. 2.4. This problem is based on earlier discussions by McKenna
[18, 19] and by Romer [3].

1
2.1 Solution Via Conservation of Angular Momentum
The initial angular momentum Li (per unit length) of the system is entirely due to the
electromagnetic field,
  ∞
E×B
Li,field = r × p dArea = 2π r× rdr, (1)
0 4πc

recalling that the field momentum density is the Poynting vector S = cE×B/4π (in Gaussian
units) divided by c2 , where c is the speed of light.
In the present problem, an electric field exists only for r < a, since the charge density σ
on the cylinder has been chosen to cancel the field from the charged wire for r > a. From
Gauss’ law we obtain
2λ
E = − r̂ (r < a), (2)
r
and hence the field momentum density in a cylindrical coordinate system (r, φ, z) is
λBex
p= φ̂ (r < a). (3)
2πcr
The initial angular momentum is therefore1
 a
λBex r × φ̂ λa2 Bex
Li,field = 2π rdr = ẑ. (5)
2πc 0 r 2c
The angular momentum when the external magnetic field is zero is due to the rotation of
the cylinder at angular velocity ω. There is now the mechanical angular momentum Iω as
well as the field angular momentum due to the solenoidal magnetic field inside the rotating,
charged cylinder. The azimuthal current (per unit length) is
Q ω λω
Jφ = = 2πaσ = . (6)
T 2π 2π
The resulting final magnetic field is along the z axis, with strength
4πJφ 2λω
Bf = ẑ = ẑ (r < a), (7)
c c
independent of radius for r < a according to Ampere’s law. Since this field is in the same
sense as the original field, we can immediately use eq. (5) to find the final field angular
momentum:
λa2Bf λ 2 a2 ω
Lf,field = ẑ = ẑ. (8)
2c c2
1
As discussed, for example, in [20], the field angular momentum in quasistatic examples can also be
computed via

ρA(C) aBex φ̂ λa2 Bex
Li,field = r× dVol = a r̂ × λ = ẑ, (4)
c 2c 2c

noting that the vector potential of the external magnetic field is A(C) = rBex φ̂/2.

2
The total angular momentum in the final state is therefore
 
λ2 a2
Lf = Lf,mechanical + Lf,field = I + 2 ωẑ. (9)
c

Since there is no frictional torque in this problem (and we ignore radiation), angular
momentum is conserved. Hence,
 
λa2Bex λa2 Bex λ 2 a2
ω= ≈ 1− 2 . (10)
2cI(1 + λ2 a2/c2 I) 2cI c I

The presence of c2 in the denominator of the last term of eq. (10) indicates the presence of
relativistic effects in this problem.

2.2 Solution Via Faraday’s Law

As the magnetic field drops, its time derivative Ḃ results in an induced electric field in the
azimuthal direction. According to Faraday’s law, we have

rḂz
Eφ (r) = − . (11)
2c
This field acts on the charged cylindrical shell to produce an azimuthal torque (per unit
length) of
λa2Ḃz dLmechanical dω
Nφ = aEφ(a)2πaσ = − = =I . (12)
2c dt dt
We integrate to find the final angular velocity:
 ∞ 
dω λa2 ∞ λa2(Bex − Bf )
ω= dt = − Ḃz dt = , (13)
0 dt 2cI 0 2cI
Again, we must note that the final magnetic field is not zero, but is given by eq. (7). With
this, eq. (13) becomes
λa2(Bex − 2λω/c)
ω= , (14)
2cI
which again leads to eq. (10).

2.3 Another Relativistic Correction

This section was written April, 2002.
In addition to the above accounting of angular momentum, there is a small amount
of initial angular momentum associated with the motion of the conduction current that
produces the field Bex. Furthermore, in the final state the cylinder of radius
 a is rotating
angular velocity ω, so its moment of inertia increases by the factor γ = 1/ 1 − a2ω 2/c2 due
to the relativistic increase of mass.2
2
The related issue of the relation between mechanical kinetic energy of electrical currents and magnetic
field energy is considered in [21].

3
 To characterize the initial mechanical angular momentum, we suppose the magnetic field
Bex is produced by a long cylinder of radius b > a, which must therefore carry azimuthal
current (per unit length along the z axis)
c
Iex = Bex. (15)
4π
This current is due to an areal number density ne of conduction electrons that we take to
have velocity ve . Then, the current Iex is also related by

Iex = −ene ve , (16)

writing e > 0 as the magnitude of the charge of the electron. Hence,

c Bex
ne ve = − . (17)
4π e
The initial mechanical angular momentum (per unit length) associated with conduction
electrons is
me c 2
Li,mech = 2πbne γ e me ve b ẑ = −γ e b Bex ẑ, (18)
2e
where the total number of conduction
 electrons per unit length is 2πbne , me is the rest mass
of the electron, and γ e = 1/ 1 − ve2/c2 ≈ 1. Combining this with eq. (5), the total initial
angular momentum is
   
λa2Bex m e c2 b 2 λa2Bex e b2
Li = 1 − γe ẑ = 1 − γe ẑ, (19)
2c λe a2 2c λre a2

where re = e2/me c2 is the classical electron radius. The last term in eq. (19) is not necessarily
small, since e/re corresponds to ≈ 1013 electrons/cm.
Reviewing the argument of sec. 2.2, we see that in eq. (12) the derivative dω/dt should
really be dγω/dt, with the moment of inertia I being calculated using the rest mass of the
cylinder. However, eq. (7) for the final magnetic field remains the same, so eq. (14) becomes

λa2(Bex − 2λω/c)
γω = , (20)
2cI
Expanding γ as approximately 1 + a2ω 2 /2c2 , we find
 
λa2Bex λ2 a2 λ2 a6Bex2
ω≈ 1− 2 − . (21)
2cI cI 2c4 I 2

2.4 A Subtle Point

This section was updated May, 2015.
This example, and near equivalents [3, 7, 18], are crafted so as to avoid a complication
associated with the return flux of the magnetic field.

4
 To see the difficulty, suppose instead that the linear charge density on the central wire
were λ0 , while that on the cylinder of radius a is still called λ. Then, according to eq. (1),
the initial field angular momentum would be
Bex Bex
Li,field = − [λ0 b2 + λ(b2 − a2)]ẑ = [λa2 − (λ + λ0)b2 ]ẑ. (22)
2c 2c
where b is the radius of the solenoid that provides the external field. Here, we make the usual
(but as we will see, unwarranted) assumption that the field of a long solenoid is essentially
zero outside the solenoid.
The final magnetic field is still given by eq. (7), so the final field angular momentum
would be
λω
Lf,field = 2 [λa2 − (λ + λ0 )b2]ẑ. (23)
c
The total final angular momentum would be
 
λ
Lf = I + 2 [λa − (λ + λ0 )b ] ωẑ.
2 2
(24)
c

Equating (22) and (24) the final angular velocity would be

Bex[λa2 − (λ + λ0 )b2]
ω= . (25)
2c{I + [λa2 − (λ + λ0 )b2 ]/c2}
However, the argument in sec. 2.2 based on Faraday’s law is exactly the same as before,
which again implies that the final angular velocity is given by eq. (10).
The argument based on Faraday’s law seems the more robust, so I conclude that eq. (10)
is correct for any value of the charge density λ0 on the central wire.
The field angular momentum calculations must be in error.

Real solenoids have only finite length, and the magnetic field is not quite zero outside the
solenoid since all of the magnetic flux inside the solenoid must be returned on paths outside
the solenoid. As discussed in [20], computations of field angular momentum associated with
long solenoids are more reliably made using eq. (4) than eq. (1). We see that the form (4)
predicts a field angular momentum that is independent of the charge density λ0 along the
axis, which restores agreement with the argument based on Faraday’s law.3

2.5 Do the Electric and Magnetic Field Lines Rotate When the
Cylinder Rotates?
This section is based on a query by Michael Romalis, May 20, 2015.
Suppose the charged cylinder were rotating with angular velocity ω in the absence of any
external magnetic field. The electric field is again given by eq. (2), and the magnetic field is
given by eq. (7). Do the lines of these electric and magnetic fields also rotate with angular
velocity ω?
3
In examples where the source of the magnetic field has only a finite extent, an analysis in spherical
coordinates is possible [1, 17] using both eqs. (1) and (4), given the same results.

5
 An appealing view of electric field lines is that they begin/end on electric charges, such
that if charges are in motion so are the electric fields lines associated with them. Hence, when
the charged cylinder rotates with final angular velocity ω we interpret the radial electric field
lines of eq. (2) as rotating with this angular velocity.
In contrast, magnetic field lines always form close loop, as magnetic charges do not exist
(as far as we know). Hence, it is less clear that the magnetic field lines rotate along with the
charged cylinder. Indeed, Faraday’s view (secs. 218 and 220 of [22], and sec. 3090 of [23])
was that the magnetic field lines do not rotate in this case.4
If we follow Einstein [25] in supposing that the density u = (E 2 + B 2)/8π of energy in the
electric and magnetic fields corresponds to density u/c2 of effective mass, and also suppose
that this energy density rotates along with the charged cylinder, then there are densities of
momentum and angular momentum associated with the fields. In particular, the angular
momentum per unit length associated with the rotating electric field lines is
  2   a 2  2
E r ω 2λ λ2 a2ω
LE = r × r × ω dArea = 2πr dr ẑ = ẑ, (26)
8πc 0 8πc r 2c2
and that associated with the magnetic field (if it rotates) is
  2   a 2  2
B r ω 2λω λ 4 a2 ω 3
LB = r × r × ω dArea = 2πr dr ẑ = ẑ, . (27)
8πc 0 8πc c 2c4
In contrast, the field angular momentum per unit length computed according to eq. (1)
is
  ∞  a
E×B r 2λ 2λω λ2 a2ω
Lfield = r × p dArea = r× 2πrdr = 2πr dr ẑ = ẑ. (28)
0 4πc 0 4πc r c c2
The supposed contribution (27) to the field angular momentum due to the possibly
rotating magnetic field lines does not have the same functional form as the “standard” result
(28), which reinforces Faraday’s view that the magnetic fields lines are not actually rotating
in this case.
On the other hand, the result (26) obtained by assuming that the electric field lines do
rotate is 1/2 of the “standard” result (28). This suggests that there is some validity to
regarding the rotating electric field as carrying momentum and angular momentum with it.

We noted above that the most reliable computation of the field angular momentum
associated with a long/infinite solenoid is via the vector potential,

ρA(C) aB(r < a)φ̂ λ2 a2ω
Lfield = r × dVol = a r̂ × λ = ẑ, (29)
c 2c c2
independent of the value of the charge density −λ0 on the wire.
In particular, if λ0 = 0, then the electric field is zero for r < a, and Er = 2λ/r for r > a,
and the field angular momentum associated with the rotating electric field lines is infinite,
  2   ∞ 2  2
E r ω 2λ λ2 (∞2 − a2)ω
LE = r × r × ω dArea = 2πr dr ẑ = ẑ. (30)
8πc a 8πc r 2c2
4
For a review of this issue, see sec. 2 of [24].

6
 Note also that the velocity of rotation of the electric field lines is v = ωr at radius r,
which exceeds the speed of light for r > c/ω. Hence, the interpretation of the rotating
field energy density u = E 2 /8π as being associated with an effective, rotating mass density
E 2 /8πc2 is doubtful for r > c/ω.

We conclude that the form (1), or better (4), should be used for computation of the
field angular momentum, rather than supposing that the rotating electric field lines can be
associated with a rotating, effective mass density E 2/8πc2.

References
[1] R.P. Feynman, R.B. Leighton and M. Sands, The Feynman Lectures on Physics, Vol. II,
Secs. 17-4, 27-5,6 (Addison Wesley, 1964),
http://www.feynmanlectures.caltech.edu/II_17.html#Ch17-S4
http://www.feynmanlectures.caltech.edu/II_27.html#Ch27-S5
http://www.feynmanlectures.caltech.edu/II_27.html#Ch27-S6

[2] E.M. Pugh and G.E. Pugh, Physical Significance of the Poynting Vector in Static Fields,
Am. J. Phys. 35, 153 (1967),
http://physics.princeton.edu/~mcdonald/examples/EM/pugh_ajp_35_153_67.pdf

[3] R.H. Romer, Angular Momentum of Static Electromagnetic Fields, Am. J. Phys. 34, 772
(1966), http://physics.princeton.edu/~mcdonald/examples/EM/romer_ajp_34_772_66.pdf
Electromagnetic Angular Momentum, Am. J. Phys. 35, 445 (1967),
http://physics.princeton.edu/~mcdonald/examples/EM/romer_ajp_35_445_67.pdf
Electromagnetic Angular Momentum, Am. J. Phys. 53, 15 (1985),
http://physics.princeton.edu/~mcdonald/examples/EM/romer_ajp_53_15_85.pdf

[4] E. Corinaldesi, Angular momentum of a static electromagnetic field, Am. J. Phys. 48,
83 (1980), http://physics.princeton.edu/~mcdonald/examples/EM/corinaldesi_ajp_48_83_80.pdf
[5] J.M. Aguirregabiria and A. Hernández, The Feynman paradox revisited, Eur. J. Phys.
2, 168 (1981), http://physics.princeton.edu/~mcdonald/examples/EM/aguirregabiria_ejp_2_168_81.pdf
[6] G.G. Lombardi, Feynman’s disk paradox, Am. J. Phys. 51, 213 (1983),
http://physics.princeton.edu/~mcdonald/examples/EM/lombardi_ajp_51_213_83.pdf

[7] F.L. Boos, Jr, More on the Feynman’s Disk Paradox, Am. J. Phys. 52, 756 (1984),
http://physics.princeton.edu/~mcdonald/examples/EM/boos_ajp_52_756_84.pdf

[8] T. Bahder and J. Sak, Elementary solution to Feynman’s disk paradox, Am. J. Phys.
53, 495 (1985), http://physics.princeton.edu/~mcdonald/examples/EM/bahder_ajp_53_495_85.pdf
[9] O.B. Keyes, Comments on “Feynman’s Disk Paradox”, Am. J. Phys. 52, 680 (1984),
http://physics.princeton.edu/~mcdonald/examples/EM/keyes_ajp_52_680_84.pdf

[10] T.-C.E. Ma, Field angular momentum in Feynman’s disk paradox, Am. J. Phys. 54,
949 (1986), http://physics.princeton.edu/~mcdonald/examples/EM/ma_ajp_54_949_86.pdf

7
[11] H.S.T. Driver, Angular momentum in static electric and magnetic fields: A simple case,
Am. J. Phys. 55, 755 (1987),
http://physics.princeton.edu/~mcdonald/examples/EM/driver_ajp_55_755_87.pdf

[12] N.L. Sharma, Field versus action at a distance in a static situation, Am. J. Phys. 56,
420 (1988), http://physics.princeton.edu/~mcdonald/examples/EM/sharma_ajp_56_420_88.pdf

[13] D.J. Griffiths, Note on “Field versus action-at-a-distance in a static situation” by

N.L. Sharma, Am. J. Phys. 57, 558 (1989),
http://physics.princeton.edu/~mcdonald/examples/EM/griffiths_ajp_57_558_89.pdf

[14] A.S. de Castro, Electromagnetic angular momentum for a rotating charged shell, Am. J.
Phys. 59, 180 (1991), http://physics.princeton.edu/~mcdonald/examples/EM/decastro_ajp_59_180_91.pdf

[15] F.S. Johnson, B.J. Cragin and R.R. Hodges, Electromagnetic momentum density and
the Poynting vector in static fields, Am. J. Phys. 62, 33 (1994),
http://physics.princeton.edu/~mcdonald/examples/EM/johnson_ajp_62_33_94.pdf

[16] O. Rojo, On the Energy Flow Vector of the E-M Field, Rev. Bras. Ens. Fis. 19, 396
(1997), http://physics.princeton.edu/~mcdonald/examples/EM/rojo_rbef_19_396_97.pdf

[17] See also Prob. 11 of Ph501 Set 5, which is based on [10],

http://physics.princeton.edu/~mcdonald/examples/ph501set5.pdf

[18] R.P. McKenna, Analog 75, No. 1, 14 (Mar. 1965),

http://physics.princeton.edu/~mcdonald/examples/EM/mckenna_analog_75_1_14_65.pdf

[19] K.T. McDonald, McKenna’s Paradox: Charged Particle Exiting the Side of A Solenoid
Magnet (Apr. 12, 2006), http://physics.princeton.edu/~mcdonald/examples/mckenna.pdf

[20] K.T. McDonald, Electromagnetic Field Angular Momentum of a Charge at Rest in a

Uniform Magnetic Field (Dec. 21, 2014),
http://physics.princeton.edu/~mcdonald/examples/lfield.pdf

[21] K.T. McDonald, How Much of Magnetic Energy is Kinetic Energy? (Sept. 12, 2009),
http://physics.princeton.edu/~mcdonald/examples/kinetic.pdf

[22] M. Faraday, Experimental Researches in Electricity.–Second Series, Phil. Trans. Roy.

Soc. London 122, 163 (1832),
http://physics.princeton.edu/~mcdonald/examples/EM/faraday_ptrsl_122_163_32.pdf

[23] M. Faraday, Experimental Researches in Electricity.–Twenty-eighth Series, Phil. Trans.

Roy. Soc. London 142, 25 (1852),
http://physics.princeton.edu/~mcdonald/examples/EM/faraday_ptrsl_142_25_52.pdf

[24] A.I. Miller, Unipolar Induction: a Case Study of the Interaction between Science and
Technology, Ann. Sci. 38, 155 (1981),
http://physics.princeton.edu/~mcdonald/examples/EM/miller_as_38_155_81.pdf

8
[25] A. Einstein, Ist die Trägheit eines Körpers von seinem Energieinhalt abhängig?, Ann.
Phys. 18, 639 (1905), http://physics.princeton.edu/~mcdonald/examples/EM/einstein_ap_18_639_05.pdf
Translation: Does the Inertia of a Body Depend upon its Energy-Content?,
http://physics.princeton.edu/~mcdonald/examples/EM/einstein_ap_18_639_05_english.pdf