Hidden momentum, field momentum, and electromagnetic impulse

David Babson and Stephen P. Reynolds

Department of Physics, North Carolina State University, Raleigh, North Carolina 27695
Robin Bjorkquist and David J. Griffiths
Department of Physics, Reed College, Portland, Oregon 97202
共Received 1 February 2009; accepted 20 May 2009兲
Electromagnetic fields carry energy, momentum, and angular momentum. The momentum density,
⑀0共E ⫻ B兲, accounts 共among other things兲 for the pressure of light. But even static fields can carry
momentum, and this would appear to contradict a general theorem that the total momentum of a
closed system is zero if its center of energy is at rest. In such cases, there must be some other
共nonelectromagnetic兲 momenta that cancel the field momentum. What is the nature of this “hidden
momentum” and what happens to it when the electromagnetic fields are turned off? © 2009 American
Association of Physics Teachers.
关DOI: 10.1119/1.3152712兴

I. INTRODUCTION would appear to be the simplest context: the parallel-plate

The linear momentum density carried by electromagnetic
fields is related to the Poynting vector1
1
℘em = S = ⑀0共E ⫻ B兲. 共1兲
c2
The classic example is an electromagnetic wave 共see Fig. 1兲.
When the wave strikes an absorber, its momentum is passed
along in the form of the pressure of light. But there are other
examples in which the fields are perfectly static, and yet the
electromagnetic momentum is not zero. Consider, for in-
stance, the following configurations.
Capacitor in a magnetic field. A charged parallel-plate ca-
pacitor 共with uniform electric field E = −Eŷ兲 is placed in a
uniform magnetic field B = Bẑ, as shown in Fig. 2.2,3 Naively,
the electromagnetic momentum is4
pem = − ⑀0EBAdx̂ = − BQdx̂, 共2兲
where A is the area of the plates, d is their separation, and Q
is the charge on the upper plate.
Magnetic dipole and electric charge. A magnetic dipole
m = mŷ is situated a distance a from a point charge q, as
shown in Fig. 3.5 The electromagnetic momentum is
␮0 qm
1
pem = x̂ = 2 共E ⫻ m兲, 共3兲
4␲ a2 c I=
where E is the electric field at the location of the dipole.
Polarized magnetized sphere. A sphere of radius R carries
a uniform polarization P and a uniform magnetization M
共see Fig. 4兲.6 The momentum carried by the fields is
pem = 94 ␲␮0R3共M ⫻ P兲. 共4兲
Coaxial cable. A long coaxial cable 共length l兲 is connected
capacitor in a uniform magnetic field. We are led to a sur-
prising paradox. In Sec. III we return to the first question 共“Is
this the whole story?”兲, to which the answer is no. Here we
develop the theory of “hidden momentum.” In Sec. IV we
work out the details for an electric dipole at the center of a
spinning, uniformly charged spherical shell, and resolve the
apparent paradox from Sec. II. In Sec. V we do the same for
an electric dipole inside a long solenoid. In Sec. VI we dem-
onstrate that hidden momentum always cancels electromag-
netic momentum, in the static case,7 and draw some general
conclusions about the nature of the hidden momentum.
II. CAPACITOR IN A UNIFORM MAGNETIC FIELD
If the electric or magnetic field is turned off, the momen-
tum originally stored in the fields must 共one would think兲 be
converted into ordinary mechanical momentum. For ex-
ample, in the case of the capacitor in a magnetic field, we
might connect a wire between the plates, allowing the ca-
pacitor to discharge slowly8 共see Fig. 6兲. According to the
Lorentz law, this wire will experience a force F = IBd to the
left, where I is the current in the wire. The net impulse de-
livered to the capacitor—which is to say, the mechanical
momentum it acquires—is
冕 冕冉 冊 冕
dq 0
Fdt = − Bd − dtx̂ = Bd dqx̂ = − BQdx̂,
dt Q
共6兲
which is precisely the momentum originally stored in the
fields 关see Eq. 共2兲兴.
Alternatively, we might turn off the magnetic field. Ac-
cording to Faraday’s law, the changing magnetic field will
induce an electric field,

冖
to a battery of voltage V at one end and a resistor R at the
other 共see Fig. 5兲. The momentum carried by the fields is d⌽
E · d艎 = − 共7兲
2
dt
lV
pem = x̂. 共5兲 共which, by Lenz’s law, runs counterclockwise in Fig. 7兲. The
c 2R
x component of the force on the strip shown is dFx
It seems strange 共to say the least!兲 for purely static fields = Ex␴wdx, where ␴ is the surface charge density, and the net
to carry momentum. Can this possibly be the whole story? force on the capacitor is Fx = ␴w养E · d艎, where the integral is
And what happens to the momentum when we turn off the taken clockwise around the dotted loop, and we have ignored
fields? In Sec. II we explore the latter question in what the two short ends. The magnetic flux through this loop

826 Am. J. Phys. 77 共9兲, September 2009 http://aapt.org/ajp © 2009 American Association of Physics Teachers 826
 E
pem
Fx = ␴w 冉冕 top
E · d艎 − 冕
bottom
E · d艎 冊 共11a兲

B = ␴w 冋1 dB d 1 dB
l−
2 dt 2 2 dt
d 1
冉 冊册
dB
− l = ␴wld ,
2 2 dt
共11b兲

Fig. 1. An electromagnetic wave carries momentum in the direction of
propagation; when it hits an absorber, this momentum is transferred in the
form of the pressure of light.
共counting inward as positive, for consistency兲 is ⌽ = −Bld, so
Fx = ␴wld共dB / dt兲, and the impulse delivered to the capacitor
is
which is half of what we obtained before. The two ends are
shorter 共兰dy = d兲, but they are farther out 共x = l / 2兲, and their
contribution to 养E · d艎 is the same as the top and bottom; but
they do not contribute to the force because there is no charge
there. Apparently, the impulse delivered to the capacitor
when B is turned off is not the same as the momentum origi-
nally stored in the fields.
But that is not all. Our naive expression for the momentum

冕冉 冊
in the fields in Eq. 共2兲 ignored the fringing field of the ca-
I= 冕 Fdt = ␴wld
dB
dt
dtx̂ pacitor, and when this field is correctly included,3 the answer
is half as great 共see Table II兲. Now lines 1 and 3 are consis-

冕 0 tent, but 2 is off! There is evidently a problem here, but it

= ␴wld dBx̂ = − BQdx̂, 共8兲 runs much deeper than that factor of 1/2, as we shall see in
B Sec. III.
the same as Eq. 共6兲. Everything seems to be in order: when
either the electric field is turned off 共by discharging the ca-
III. HIDDEN MOMENTUM
pacitor兲 or the magnetic field is removed, the momentum
originally stored in the fields is converted into ordinary me- There is a very general principle in special relativity,10
chanical momentum, and the capacitor moves off to the left which we shall call the center of energy theorem: if the cen-
共see Table I兲. Sounds good, but it is almost entirely wrong.9 ter of energy12 of a closed13 system is at rest, then its total
In the first place, 养E · d艎 includes the two vertical seg- momentum is zero. The center of energy of a capacitor in a
ments, which do not contribute the force. Of course, we as- static magnetic field is certainly at rest, so if there is momen-
sume that the plates are very close together; doesn’t that tum in the fields, there must be some compensating nonelec-
mean the “extra” piece is negligible? Unfortunately, it does tromagnetic momentum elsewhere. Where is this “hidden
not. Suppose we use a solenoid to establish the magnetic momentum” located, and what is its nature? When we turn
field 共see Fig. 8兲. Because of the azimuthal symmetry, we off E or B, and the capacitor gets a kick, something else 共in
can calculate the induced electric field explicitly, E共2␲r兲 = this case the solenoid兲 must get an equal and opposite kick so
−␲r2dB / dt, which implies that that the total momentum remains zero. But there is no a
r dB ˆ priori reason that the impulse to the capacitor should equal
E=− ␾ 共9兲 the momentum originally stored in the fields, or that it should
2 dt be the same when we turn off the electric field as when we
and turn off the magnetic field.
The cleanest example of hidden momentum is the follow-
r dB ing: Imagine a rectangular loop of wire carrying a steady
E · d艎 = − 共− sin ␾x̂ + cos ␾ŷ兲 · 共dxx̂ + dyŷ + dzẑ兲
2 dt current.14 Picture the current as a stream of noninteracting
共10a兲 positive charges that move freely within the wire. When a
uniform electric field E is applied 共see Fig. 9兲, the charges
1 dB will accelerate up the left segment and decelerate down the
= 共r sin ␾dx − r cos ␾dy兲 right one. Notice that there are fewer charges in the upper
2 dt segment but they are moving faster. Question: What is the
1 dB total momentum of all the charges in the loop?
= 共ydx − xdy兲. 共10b兲 The momenta of the left and right segments cancel, so we
2 dt
need only consider the top and bottom segments. Say there
So the force on the capacitor is are Nt charges in the top segment, going to the right at speed
vt, and Nb charges in the lower segment, going at speed vb to
the left. The current 共I = ␭v兲 is the same in all four segments
y 共otherwise charge would be piling up somewhere, and it
would not be a steady current兲. Thus
+Q pem
E _Q Table I. Capacitor in a magnetic field—naive solution.
B x

z Momentum initially stored in fields −BQdx̂

Momentum delivered to capacitor as it discharges −BQdx̂
Fig. 2. A charged capacitor in the presence of an external magnetic field Momentum delivered to capacitor as B decreases −BQdx̂
carries electromagnetic momentum even though nothing is moving.

827 Am. J. Phys., Vol. 77, No. 9, September 2009 Babson et al. 827
 y Table II. Capacitor in a magnetic field—corrected.
1
Momentum initially stored in fields − 2 BQdx̂
pem Momentum delivered to capacitor as it discharges −BQdx̂
m 1
Momentum delivered to capacitor as B decreases − 2 BQdx̂
a x
q
z
1
Fig. 3. There is electromagnetic momentum in the fields of a point charge phid = 共m ⫻ E兲, 共16兲
near a magnetic dipole even though both are stationary. c2
where m is the magnetic dipole moment of the loop 共兩m兩
= Ilw兲. If we integrate ℘em = ⑀0共E ⫻ B兲 for a magnetic dipole
qNt qNb in an electric field,17 we obtain18
I= vt = vb , 共12a兲
l l 1
pem = − 共m ⫻ E兲. 共17兲
c2
so
The hidden momentum exactly cancels the field momentum,
Il as the center of energy theorem requires.
N tv t = N bv b = , 共12b兲
q It is easy to generalize this result. In terms of the electric
potential V, 共␥t − ␥b兲mc2 = −q共Vt − Vb兲, and hence19
where q is the charge of each particle and l is the length of
the rectangle. Nonrelativistically, the momentum of a single mIl q
phid = − 共Vt − Vb兲x̂
particle is p = mv, where m is its mass, so the total momen- q mc2
tum 共to the right兲 is
Il Il
pclass = mNtvt − mNbvb = m − m = 0, 共13兲
=−
c
I
c
I
2 共Vt − Vb兲lx̂ = − 2 冖 Vd艎, 共18兲
q q
where the integral is around the loop in the direction of the
as we would expect 共after all, the loop as a whole does not current 共the ends cancel兲. For surface or volume currents, we
move兲. But relativistically the momentum of a particle is p have
= ␥mv, and we get

prel = ␥tmNtvt − ␥bmNbvb =

mIl
共␥t − ␥b兲, 共14兲
phid = −
1
c2
冕 VKda 共19a兲

q and
which is not zero because the particles in the upper segment
move faster.
As a particle goes up the left side, it gains energy equal to
phid = −
1
c2
冕 VJd␶ . 共19b兲

the work done by the electric force, This model of electric current is artificial, and one might
prefer to treat it as an incompressible fluid.20 In that case,
␥tmc − ␥bmc = qEw,
2 2
共15a兲
assuming that the wire has a constant cross section, the speed
so and spacing of the charges are the same all the way around
the loop, but those in the top segment are under higher pres-
IlEw sure. Now, a moving fluid under high pressure carries greater
prel = , 共15b兲 momentum than the same fluid under low pressure. The
c2
quickest way to see this is by examining the stress-energy
and hence tensor for a simple fluid,21

T ␮␯ = ␳ 0v ␮v ␯ + P 冉 v ␮v ␯
c2
− g ␮␯ , 冊 共20兲

where ␳0 is the mass density in the rest frame of the fluid, P

M is the 共scalar兲 pressure, and v␮ = ␥共c , v兲 is the local
P
B pem

I
E + + I
_
+ +
x

V _ _ _ _ _ R

Fig. 5. A coaxial cable, with current flowing from a battery at one end
Fig. 4. A sphere with uniform polarization P and uniform magnetization M through a resistor at the other and returning, carries electromagnetic momen-
carries electromagnetic momentum proportional to M ⫻ P. tum in the direction of the high-voltage current.

828 Am. J. Phys., Vol. 77, No. 9, September 2009 Babson et al. 828
 y

+
I F y
d φ
B _ x
x
E r
z B
z
Fig. 6. A fine wire between the plates allows current I to flow, discharging
the capacitor. The magnetic force on the current delivers an impulse to the
capacitor.
Fig. 8. In this model the magnetic field is produced by a long solenoid, and
the induced electric field can be calculated explicitly 关Eq. 共9兲兴.

4-velocity. The 0i components of T␮␯ give the momentum

density, vl IlEw

冉
T0i = ␥2 ␳0c +
P i
c
冊
v = 㜷ic. 共21兲
phid =
c 2 ␳EAw =
c2
,

the same as before 关Eq. 共15b兲兴.

共27兲

Thus, the momentum density of the fluid is

℘fluid = ␥2 ␳0 +冉 P
c2
冊
v. 共22兲 IV. ELECTRIC DIPOLE AT THE CENTER
OF A SPINNING, UNIFORMLY CHARGED
The first term represents the ordinary flow of mass; the sec- SPHERICAL SHELL
ond is the “extra” momentum associated with pressure.22 The What about a capacitor in the magnetic field of a solenoid?
latter accounts for the hidden momentum, In this case the hidden momentum is located in the solenoid
␥2 共that is where the moving charges are兲, and the electric field
phid = 共㜷t − 㜷b兲lA = 共Pt − Pb兲vlA, 共23兲 responsible for the variation in ␥ must be the exterior 共fring-
c2
ing兲 field of the capacitor.24 The fringing field is notoriously
where A is the cross-sectional area of the wire. difficult to calculate, and the results are independent of the
The difference in pressure between the top and bottom geometry, so we begin with a simpler model, replacing the
segments is due to the 共electric兲 force on the charges in 共say兲 capacitor by an electric dipole with dipole moment p = pŷ,25
the left segment of the loop. In relativity, the force per unit and the solenoid by a spherical shell of radius R which car-
volume acting on a fluid is given by23 ries a uniform surface charge ␴ and spins at a constant an-
⳵ ␮␯ gular velocity ␻ = ␻ẑ 共see Fig. 10兲.26 This configuration pro-
f␮ = T . 共24兲 duces a uniform magnetic field,
⳵ x␯
B = 32 ␮0␴R␻ 共28兲
If x is the vertical direction, then from Eq. 共20兲
1

冉 冊
for points inside the sphere, and a dipole field
dP ␥2v2
1
f = 1 +1 共25a兲 ␮0 1 4␲ 4
dx c2 B= 关3共m · r̂兲r̂ − m兴, where m = ␴R ␻ 共29兲
4␲ r3 3
or
for points outside. The field of the electric dipole is
f = ␥2 ⵜ P. 共25b兲
1 1 1
Equation 共25b兲 is the 共relativistic兲 relation between the force E= 3 关3共p · r̂兲r̂ − p兴 − p␦3共r兲. 共30兲
density and the pressure gradient. In our case the force per 4␲⑀0 r 3⑀0
unit volume is ␳E, so ␥2 ⵜ P = ␳E, or integrating over the
volume of the left segment,
vt
␥2共Pt − Pb兲A = ␳EAw. 共26兲
Putting Eq. 共26兲 into Eq. 共23兲 and using I = ␳Av, we get I

w I E I
y
l I
w
dx +σ
vb
d _σ
B x l
z Fig. 9. The simplest model for hidden momentum: a steady current I con-
sisting of noninteracting charges constrained to move around a rectangular
Fig. 7. If the magnetic field is gradually reduced, an electric field is induced, tube. In the presence of an external electric field E, the charges accelerate up
in the counterclockwise direction, imparting an impulse to the capacitor. the left segment and decelerate down the right segment.

829 Am. J. Phys., Vol. 77, No. 9, September 2009 Babson et al. 829
 y
phid = −
1
c2
冕 VKda

σ =−
1
c2
冕冉 p · r̂
冊
4 ␲ ⑀ 0r 2
关␴共␻ ⫻ r兲兴da 共35a兲
p
B
x
=− ⑀0␮0
p
4␲⑀0
␴␻R 冕 sin ␪ sin ␾关sin ␪ cos ␾ŷ
z − sin ␪ sin ␾x̂兴sin ␪d␪d␾ 共35b兲

冉 冊
Fig. 10. A uniformly charged spherical shell spins at constant angular ve-
locity, producing a uniform magnetic field at interior points; an electric ␮0 4 1
dipole is located at the center. = p␴␻R ␲x̂ = 共p ⫻ B兲. 共35c兲
4␲ 3 2
So far, so good: the momentum, pem + phid, is zero.
To calculate the electromagnetic momentum, we integrate Now let’s connect a wire between the ends of the electric
⑀0共E ⫻ B兲 over the interior and exterior of the sphere,27

再冕 冋
dipole, allowing it to discharge. The impulse delivered to the
1 1 dipole is
pin = ⑀0 关3共p · r̂兲r̂ − p兴
4␲⑀0 r3
冕 冕
册冎
Idip = Fdt = I共d ⫻ B兲dt
1

冉冕 冊
− p␦3共r兲 d␶ ⫻ B 共31a兲
3⑀0 0
= dq d ⫻ B = − 共p ⫻ B兲, 共36兲
1 q
=− 共p ⫻ B兲. 共31b兲
3 where d is the separation of the charges, and p = qd. At the
This part comes exclusively from the delta function because same time, the changing electric field of the dipole induces a
the angular integral of the first term 共using spherical coordi- magnetic field, which exerts a force on the spinning sphere.
To calculate the induced magnetic field, we orient the polar
nates and p = pŷ兲,
axis along p 共see Fig. 11兲 and apply the Ampere–Maxwell
冕 关3共p sin ␪ sin ␾兲共sin ␪ cos ␾x̂ + sin ␪ sin ␾ŷ
law 共养B · d艎 = ␮0Ienc + ␮0⑀0d⌽E / dt兲 to the dotted “Amperian
loop” using the spherical cap to determine ⌽E,
+ cos ␪ẑ兲 − pŷ兴sin ␪d␪d␾ ,
is zero. The contribution from outside the sphere is
共32兲
⌽E = 冕 E · da

pout = ⑀0
1 ␮0
4␲⑀0 4␲
冕 1
r6
关3共p · r̂兲r̂ − p兴 =
1
4␲⑀0
冕 1
r3
关3共p · r̂兲r̂ − p兴 · r2 sin ␪d␪d␾r̂ 共37a兲

⫻ 关3共m · r̂兲r̂ − m兴r2 sin ␪drd␪d␾

冕
共33a兲
1 p sin2 ␪
= 共2p cos ␪兲sin ␪d␪d␾ = . 共37b兲
␮0mp 1 4 ␲ ⑀ 0r 2 ⑀ 0r
=− 3 x̂ = − 共p ⫻ B兲, 共33b兲
12␲R 6
B points in the ␾
ˆ direction,

冉 冊
where B is the magnetic field inside the sphere. So
d p sin2 ␪
pem = pin + pout = − 21 共p ⫻ B兲. 共34兲 B共2␲r sin ␪兲 = ␮0⑀0 , 共38兲
dt 2⑀0r
The hidden momentum in the spinning sphere is 关see Eq.
共19a兲兴 so

␮0 ṗ sin ␪ ␮0 兩ṗ ⫻ r̂兩

B= = , 共39兲
z 4␲ r2 4␲ r2

where ṗ = dp / dt. The force on the spinning sphere 共reverting

p = pŷ and ␻ = ␻ẑ兲 is
r sin θ
θ
r
F= 冕 共K ⫻ B兲da =
␮ 0␴
4␲R2
冕 关共␻ ⫻ r兲 ⫻ 共ṗ ⫻ r̂兲兴da

p =−
␮0␴R␻ ṗ
x̂, 共40兲
3
Fig. 11. Geometry for calculating the magnetic field of a discharging electric
dipole. and the impulse delivered to the shell is

830 Am. J. Phys., Vol. 77, No. 9, September 2009 Babson et al. 830
Table III. Electric dipole inside spinning sphere—naive solution. y

1 1
Initial momentum pem = − 2 共p ⫻ B兲 phid = 2 共p ⫻ B兲 ptot = 0
1 1 rmax
Dipole discharges Idip = −共p ⫻ B兲 Isphere = 2 共p ⫻ B兲 ptot = − 2 共p ⫻ B兲 R
1 1 θ p
Sphere slows Idip = − 2 共p ⫻ B兲 Isphere = 0 ptot = − 2 共p ⫻ B兲
z

Isphere = 冕 1
Fdt = 共p ⫻ B兲.
2
共41兲
Fig. 12. A long solenoid produces a uniform magnetic field at interior
points; an electric dipole is located on the axis.

Curiously, this impulse is only half as great as the impulse to

the dipole 关see Eq. 共36兲兴.
Alternatively, we could turn off the magnetic field by prevails outside this sphere. It is safest to do the calculation
bringing the spinning sphere to rest. During this process the in spherical coordinates even though this makes the upper
changing magnetic field induces an electric field, limit on the r integral awkward: rmax = R / sin ␪,

E = − 21 Ḃr sin ␪␾
ˆ, 共42兲
pem = ⑀0 再冕 冋 1 1
4␲⑀0 r3
关3共p · r̂兲r̂ − p兴

册冎
which exerts a force on the two ends of the dipole,

F = 2q 冉 冊Ḃ d
22
1
x̂ = pḂx̂.
2
共43兲
−
1
3⑀0
p␦3共r兲 d␶ ⫻ B 共46a兲

The net impulse is

Idip = − 21 共p ⫻ B兲 共44兲
=
B
4␲
冕 1
r
1
3 关3共p · r̂兲共r̂ ⫻ ẑ兲 − 共p ⫻ ẑ兲兴d␶ − 共p ⫻ B兲.
3
共which is the same as the momentum originally stored in the 共46b兲
fields兲. But this time there is no electromagnetic impulse to
the sphere 共see Table III兲. It is still inconsistent! The im- Now,
pulses to the dipole and the spinning sphere do not balance,
and it seems that the total momentum of the system, after p · r̂ = p sin ␪ sin ␾ ,
turning off the fields, is not zero. 共47兲
But wait: What happened to the hidden momentum when r̂ ⫻ ẑ = sin ␪ sin ␾x̂ − sin ␪ cos ␾ŷ, p ⫻ ẑ = px̂,
we turned off the fields? This was purely mechanical mo-
mentum associated with the motion of the charges that con- so the integral is

冕
stitute the current; as the fields are reduced, the excess mo-
1
mentum of the charges is delivered to the structure that keeps p 关3 sin ␪ sin ␾共sin ␪ sin ␾x̂ − sin ␪ cos ␾ŷ兲
them on track 共in this case the spherical shell兲.28 Thus the r3
hidden momentum should be added to the impulse delivered
− x̂兴r2 sin ␪drd␪d␾

冕 再冕 冎
to the sphere as the fields are turned off, and finally every-
␲ R/sin ␪
thing works out 共see Table IV兲. 1
= ␲ px̂ dr 共3 sin2 ␪ − 2兲sin ␪d␪
0 ⑀ r
V. ELECTRIC DIPOLE IN THE FIELD
共48a兲
OF A LONG SOLENOID
The calculations are no more difficult for the solenoid
model 共see Fig. 12兲.29 In this case, the magnetic field is = ␲ px̂ 冕 ␲
关ln R − ln共sin ␪兲 − ln ⑀兴

B共s兲 = 再 Bẑ 共s ⬍ R兲
0 共s ⬎ R兲,
冎 共45兲
0

⫻共1 − 3 cos2 ␪兲sin ␪d␪ 共48b兲

where R is the radius of the solenoid, s is the distance from

the axis, and B = ␮0K = ␮0nI 共K is the surface current density,
n is the number of turns per unit length, and I is the current兲.
=− ␲ px̂ 冕 ␲

0
ln共sin ␪兲共1 − 3 cos2 ␪兲sin ␪d␪

The delta function term in the dipole field in Eq. 共30兲 must 2
be handled with care: this represents the field inside a sphere = − ␲ px̂. 共48c兲
of radius ⑀ 共in the limit ⑀ → 0兲; the “ordinary” dipole field 3

Thus
Table IV. Electric dipole inside spinning sphere—corrected.

Initial momentum
1
pem = − 2 共p ⫻ B兲
1
phid = 2 共p ⫻ B兲 ptot = 0
pem =
B
4␲
− 冉
2␲
3
1
3
1
2
冊
px̂ − 共p ⫻ B兲 = − 共p ⫻ B兲, 共49兲
Dipole discharges Idip = −共p ⫻ B兲 Isphere + phid = 共p ⫻ B兲 ptot = 0
Sphere slows
1
Idip = − 2 共p ⫻ B兲
1
phid = 2 共p ⫻ B兲 ptot = 0 which is the same as for the spinning sphere 关see Eq. 共34兲兴.
The hidden momentum in the solenoid is

831 Am. J. Phys., Vol. 77, No. 9, September 2009 Babson et al. 831
Table V. Electric dipole in the field of a solenoid. VI. CONCLUSION
1 1
Initial momentum pem = − 2 共p ⫻ B兲 phid = 2 共p ⫻ B兲 ptot = 0 We considered a rectangular current loop in a uniform
Dipole discharges Idip = −共p ⫻ B兲 Isol + phid = 共p ⫻ B兲 ptot = 0 electric field, an electric dipole at the center of a spinning
1 1
Current decreases Idip = − 2 共p ⫻ B兲 phid = 2 共p ⫻ B兲 ptot = 0 charged spherical shell, and an electric dipole in the field of
a long solenoid; in each case the hidden momentum exactly
balances the electromagnetic momentum, so the total is zero,

冕冉 冊
consistent with the center of energy theorem. Can we prove
phid = −
1
c2
冕 VKda = − ␮0⑀0
p · r̂
4 ␲ ⑀ 0r 2
共K␾
ˆ 兲Rd␾dz.
that this cancellation always works when the fields are static?
Yes, for if we use Ampere’s law 共⵱ ⫻ B = ␮0J兲 to replace the
current density in the general expression for hidden momen-
共50兲 tum 关see Eq. 共19b兲兴 and integrate by parts, we obtain

冕
In this case, it is simplest to use cylindrical coordinates
共s , ␾ , z兲, phid = −
1
VJd␶
c2
r = 冑R2 + z2,
pR
p · r̂ =
r
sin ␾, ␾ˆ = − sin ␾x̂ + cos ␾ŷ,

共51兲
=−
⑀ 0␮ 0
␮0
冕 V共ⵜ ⫻ B兲d␶ 共57a兲

冕
so

phid = −
␮0
4␲
pKR2 冕 sin ␾共− sin ␾x̂ + cos ␾ŷ兲
共R2 + z2兲3/2
d␾dz
=− ⑀0 关ⵜ ⫻ 共VB兲 + 共B ⫻ 共ⵜV兲兲兴d␶ 共57b兲

共52a兲
= ⑀0 冖 VB ⫻ da + ⑀0 冕 共B ⫻ E兲d␶
1
= ␮0 pKR2x̂
2
冕 0
⬁
dz
共R + z2兲3/2
2
= − ⑀0 冕 共E ⫻ B兲d␶ = − pem . 共57c兲
1 1
= pBx̂ = 共p ⫻ B兲, 共52b兲
2 2 共For a localized distribution the boundary term is zero.兲
When the electric or magnetic field is turned off, the field
as in the spherical model 关Eq. 共35兲兴.
momentum disappears, the hidden momentum is absorbed,
Now we discharge the dipole. The impulse to the dipole
and some element共s兲 in the system may receive an impulse.
itself is the same as in Eq. 共36兲, and the induced magnetic
But there is no obvious reason why this impulse should equal
field is again given by Eq. 共39兲. The force on the solenoid is
the momentum originally stored in the fields—all we can say
F= 冕 共K ⫻ B兲da =
␮0
4␲
冕 1
r2
关K ⫻ 共ṗ ⫻ r̂兲兴Rd␾dz. 共53兲
in general is that the total momentum afterward, like the total
momentum before, is zero.30
The hero 共or is it the villain?兲 of this story is hidden mo-
But mentum. What can be said about the nature of hidden mo-
mentum in general? It seems to share three general features:
ṗK
K ⫻ 共ṗ ⫻ r̂兲 = 共− R cos2 ␾x̂ − R sin ␾ cos ␾ŷ • It is purely mechanical.31 Although it arises most often in
r
electromagnetic contexts, it has nothing to do with electro-
− z cos ␾ẑ兲, 共54兲 dynamics. The force involved in Fig. 9 could just as well
so be gravity, or little rockets attached to the particles, and
Eq. 共19b兲 could be written more generally as
1
F = − ␮0 ṗKR2x̂
2
冕 0
⬁
1
共R + z 兲
2
1
2 3/2 dz = − Bṗx̂.
2
共55兲
phid = −
1
冕 uvd␶ , 共58兲
c2
The impulse delivered to the solenoid is

Isol = 冕 1
Fdt = − Bx̂
2
冕 dp
dt
dt
where u is the potential energy density 共of whatever form兲
and v is the local velocity.
• It occurs in systems with internally moving parts 共such as
1
= − Bx̂
2
冕 p
0
1
2
1
dp = Bpx̂ = 共p ⫻ B兲,
2
共56兲
current loops兲.32
• It is intrinsically relativistic.

again reproducing the result for the spinning sphere in Eq.
共41兲.
If we turn off B 共by reducing the current in the solenoid兲,
the induced electric field is the same as in Eq. 共42兲, and the
impulse delivered to the dipole is the same as in Eq. 共44兲;
again, there is no impulse to the solenoid. The entire table is
essentially unchanged 共compare Tables IV and V兲.
A definitive characterization of the phenomenon remains
elusive, and some have suggested that the term should be
expanded to include all strictly relativistic contributions to
momentum 关including electromagnetic momentum, the 共␥
− 1兲mv piece of particle momentum, and the ␥2 Pv / c2 portion
of the momentum density of a fluid under pressure兴; others
urged that the term be expunged altogether.33

832 Am. J. Phys., Vol. 77, No. 9, September 2009 Babson et al. 832
 1
The relation between momentum density and energy flux is not peculiar
to electrodynamics. See R. P. Feynman, R. B. Leighton, and M. Sands,
The Feynman Lectures 共Addison-Wesley, Reading, MA, 1964兲, Vol. 2,
Eq. 共27.21兲.
2
F. S. Johnson, B. L. Cragin, and R. R. Hodges, “Electromagnetic momen-
tum density and the Poynting vector in static fields,” Am. J. Phys. 62,
33–41 共1994兲.
3
K. T. McDonald, “Electromagnetic momentum of a capacitor in a uni-
form magnetic field” ⬍www.hep.princeton.edu/~mcdonald/examples/
cap_momentum.pdf⬎.
4
As we shall see, this turns out to be incorrect.
5
W. H. Furry, “Examples of momentum distributions in the electromag-
netic field and in matter,” Am. J. Phys. 37, 621–636 共1969兲.
6 error 共Ref. 3兲.
R. H. Romer, “Question #26. Electromagnetic field momentum,” Am. J.
Phys. 63, 777–779 共1995兲. Romer did not use a polarized/magnetized
sphere but a sphere carrying the same surface charge and current distri-
butions as they would produce. This avoids the awkward question of
whether we should use D ⫻ B in place of Eq. 共1兲 and preempts conceptual
problems about hidden momentum in bound currents. If such issues arise,
this example is to be interpreted in Romer’s way, as a configuration of
free charges and free currents.
7
Hidden momentum occurs in moving systems as well. See, for instance,
E. Comay, “Exposing ‘hidden momentum’,” Am. J. Phys. 64, 1028–
1034 共1996兲, but it is most striking in static configurations, and we shall
concentrate on such cases.
8 27
In this paper all such processes will be carried out quasistatically to avoid
electromagnetic radiation 共which would remove momentum from the sys-
tem兲.
9 28
This, incidentally, is the answer provided in the Solution Manual to D. J.
Griffiths, Introduction to Electrodynamics, 3rd ed. 共Prentice-Hall, Upper
Saddle River, NJ, 1999兲, Problem 8.6. The error was caught by David
Babson.
10
See S. Coleman and J. H. Van Vleck, “Origin of ‘hidden momentum
forces’ on magnets,” Phys. Rev. 171, 1370–1375 共1968兲 and Refs. 5 and
11.
11
M. G. Calkin, “Linear momentum of the source of a static electromag-
netic field,” Am. J. Phys. 39, 513–516 共1971兲.
12
Center-of-energy is the natural relativistic generalization of center-of-
mass: 兰rud␶ / 兰ud␶, where u is the energy density.
13
It is notoriously dangerous to speak of the momentum 共or energy兲 of a
configuration that is not localized in space. In this context a uniform field
should always be interpreted to mean locally uniform, but going to zero at
infinity.
14
See Refs. 11, 15, and 16.
15
L. Vaidman, “Torque and force on a magnetic dipole,” Am. J. Phys. 58,
978–983 共1990兲.
16
V. Hnizdo, “Hidden momentum of a relativistic fluid carrying current in
an external electric field,” Am. J. Phys. 65, 92–94 共1997兲.
17
D. J. Griffiths, “Dipoles at rest,” Am. J. Phys. 60, 979–987 共1992兲.
18
Because electromagnetic momentum 关see Eq. 共1兲兴 is linear in B, this
result—expressed in terms of the total magnetic dipole moment—holds
for any collection of dipoles, and hence in particular for the rectangular
loop 共which could be built up as a tessellation of tiny squares兲.
19
See Refs. 11, 15, 16, and 3.
20
See Ref. 15 for the nonrelativistic argument and Ref. 16 for the relativ-
istic version.
21
See, for example, R. J. Adler, M. J. Bazin, and M. Schiffer, Introduction
to General Relativity, 2nd ed. 共McGraw-Hill, New York, 1975兲, Sec. 9.2.
22
Why should a moving fluid under pressure carry extra momentum? This is
a surprisingly subtle relativistic effect. For a lovely explanation, see K.
Jagannathan, “Momentum due to pressure: A simple model,” Am. J.
833 Am. J. Phys., Vol. 77, No. 9, September 2009
Phys. 77, 432–433 共2009兲. The essence of the argument appears also in
Ref. 7, Sec. IIB. A variation is suggested by K. Szymanski, “On the
momentum of mechanical plane waves,” Physica B 403, 2996–3001
共2008兲, Sec. V; see also R. Medina, “The inertia of stress,” Am. J. Phys.
74, 1031–1034 共2006兲.
23
A more general version of the following argument is found in Ref. 16 and
in a slightly different form in V. Hnizdo, “Hidden mechanical momentum
and the field momentum in stationary electromagnetic and gravitational
systems,” Am. J. Phys. 65, 515–518 共1997兲.
24
This realization casts doubt on our original naive expression for the field
momentum in Eq. 共2兲, which took the electric field to be uniform inside
the capacitor and zero outside, and as we have seen, that equation is in
25
Because electromagnetic momentum 关see Eq. 共1兲兴 is linear in E, our
result—expressed in terms of the total electric dipole moment—will hold
for any collection of dipoles, and hence in particular for the original
capacitor model. This is how McDonald 共Ref. 3兲 fixed the error in Eq.
共2兲.
26
More precisely, we want an electrically neutral spherical shell that carries
a surface current density K = ␴共␻ ⫻ r兲. In the presence of an electric field,
the charges constituting this current will speed up and slow down 共in the
first model of Sec. III兲, spoiling the simple picture of a rigid spinning
sphere. The incompressible fluid model is, in this sense, closer in spirit to
the spinning sphere.
More simply, we can use the general expression for the momentum of an
electric dipole in a magnetic field 共Ref. 17兲: pem = 共p · ⵜ兲A, where in this
1
case A = − 2 共r ⫻ B兲.
This is the “hidden momentum force” that began the whole story. W.
Shockley and R. P. James, “ ‘Try simplest cases’ discovery of ‘hidden
momentum’ forces on ‘magnetic currents’,” Phys. Rev. Lett. 18, 876–
879 共1967兲.
29
One might worry, of course, about assuming that the solenoid is infinitely
long, or—if finite—the approximations involved in treating its field as
uniform inside and zero outside.
30
The rotational analog is strikingly different: Angular momentum in static
fields is not 共typically兲 compensated by hidden angular momentum 共there
is no rotational analog to the center-of-energy theorem兲, and when the
fields are removed, the system starts to rotate, with angular momentum
equal to that originally stored in the fields. The classic case is the “Feyn-
man disk paradox” 共Ref. 1, Sec. 17-4兲. It is easy to construct configura-
tions with hidden angular momentum, but because there is no analog to
the center of energy theorem, hidden angular momentum is not forced on
us as dramatically as hidden linear momentum.
31
A possible counterexample is given in Ref. 7, Sec. III, where the role of
hidden momentum is played by standing electromagnetic waves.
32
For example, if the magnetic field is produced by stationary magnetic
monopoles, instead of electric currents, then there is no hidden momen-
tum, but in that case the electromagnetic momentum also vanishes 共Ref.
17兲.
33
Although hidden momentum was first discovered 40 years ago 共Ref. 28兲,
it continues to carry a mysterious aura and has been misunderstood and
misused by a number of authors 共including one of us, D.J.G.兲. Just last
year Jon Thaler 共personal communication, August 26, 2007兲 and Timothy
Boyer, “Concerning ‘hidden momentum’,” Am. J. Phys. 76, 190–191
共2008兲 pointed out that the coaxial cable is not an example of hidden
momentum—there is momentum in the fields, but the center of energy is
not at rest. The battery at one end is losing energy, and the resistor at the
other end is gaining energy, and the momentum in the fields is precisely
the momentum associated with the motion of the center of energy.
Babson et al. 833