This is the accepted manuscript made available via CHORUS.

The article has been

published as:

Electric Power Generation from Earth’s Rotation through its

Own Magnetic Field
Christopher F. Chyba and Kevin P. Hand
Phys. Rev. Applied 6, 014017 — Published 29 July 2016
DOI: 10.1103/PhysRevApplied.6.014017
 Electric Power Generation from Earth’s Rotation Through its
Own Magnetic Field

Christopher F. Chyba∗
Department of Astrophysical Sciences and Woodrow
Wilson School of Public and International Affairs,
Princeton University, Princeton, NJ 08544 USA

Kevin P. Hand†
Jet Propulsion Laboratory, California Institute
of Technology, Pasadena, California 91109 USA
(Dated: June 27, 2016)

Abstract
We examine electric power generation from Earth’s rotation through its own non-rotating mag-
netic field (that component of the field symmetric about Earth’s rotation axis). There is a simple
general proof that this is impossible. However, we identify a loophole in that proof and show
that voltage could be continuously generated in a low-magnetic-Reynolds number conductor rotat-
ing with Earth, provided magnetically permeable material were used to ensure curl(v × B0 ) 6= 0
within the conductor, where B0 derives from the axially symmetric component of Earth’s magnetic
flux density and v is Earth’s rotation velocity at the conductor’s location. We solve the relevant
equations for one laboratory realization, and from this solution predict voltage magnitude and sign
dependence on system dimensions and orientation relative to Earth’s rotation. The effect, which
would be available nearly globally with no intermittency, requires testing and further examination
to see if it could be scaled to practical emission–free power generation.

1
I. INTRODUCTION

Barnett showed in 1912 that when an axially symmetric electromagnet rotates about
its north-south axis, its magnetic field does not rotate with the magnet [1], resolving a
controversy [2] that had its origins in Faraday’s interpretation of his rotating disk experiment
[3, 4]. Here, in Section II we first review Faraday’s results and carefully address the definition
of electromotive force (emf) and the historical meaning of saying a magnetic field “rotates
with the magnet.” We then describe the compelling experimental evidence for the non-
rotation of axially symmetric magnetic fields. In Section III, we consider the particular case
of the non-rotation of the axisymmetric component of Earth’s magnetic field. The rotation
of Earth’s surface through that non-rotating component yields a steady v × B force that one
might hope to use to generate electric power. However, in Section IV we present a simple,
and seemingly general, proof that power generation in this way is impossible.
Nonetheless, in Section V we show that this proof has a loophole, suggesting that con-
tinuous power generation is possible if two unusual conditions are both met. These two
conditions will not simultaneously hold in any typical natural or laboratory circuit, but it is
possible to create them together. The first condition is that the current path must lie within
a magnetically permeable conductor the topology of which is such that ∇ × (v × B) 6= 0 in
its interior, where B is the magnetic flux density and v is the velocity of the conductor. The
second is that this conductor must have a magnetic Reynolds number Rm  1, which on a
laboratory scale excludes all common metal and mu-metal conductors.
In Sections VI to IX, we fully calculate one realization of such a system: a low-Rm mag-
netically permeable cylindrical shell. Section VI first considers the case when the shell is
stationary (v = 0) with respect to a constant background magnetic field (with zero back-
ground electric field). Then the current density J = 0 and it is straightforward to derive
the corresponding magnetic flux density B0 within the shell. We prove that in general,
∇ × (v × B0 ) 6= 0, so that if the shell’s composition, dimensions, and velocity can be cho-
sen to yield Rm  1, the system would fulfill the necessary conditions for emf generation.
In Sections VII through IX, we demonstrate that for this system these conditions are
also sufficient. In Section VII, we show that if the shell is put into motion transverse to
its long axis, B0 can no longer be a solution. We calculate the B that does satisfy the
induction equation within the moving shell when v 6= 0 and Rm  1, and find that the

2
time-dependent part of B goes to zero extremely rapidly. In Section VIII, we find that there
remains a time-independent solution, given by B0 plus a series of perturbation terms scaled
by successive powers of Rm . We use these results in Section IX to derive an expression for
the emf generated in the shell. Section X provides an intuitive discussion of the results of
the previous calculations.
In Section XI, we present a parallel analysis in a frame co-moving with the translating
shell, and demonstrate that there is a net non-zero Poynting vector flux delivering power
into the shell. We show in Section XII that in the frame in which the shell is translating at
v, a magnetic braking term arises in Poynting’s theorem, and the generated electrical power
equals the braking loss from Earth’s rotational kinetic energy.
We make quantitative predictions for this system in Section XIII, including the striking
prediction that the voltage generated should change sign when the cylindrical shell (together
with its attached leads and voltmeter) is rotated by 180◦ . Section XIV begins the discussion
of whether such systems might be scaled up to generate useful amounts of electric power.

II. HISTORICAL BACKGROUND AND DEFINITIONS

In December 1831, Faraday experimented with a conducting disk rotating near a magnet
[3, 4]. The disk connected via brushes to a simple galvanometer, with leads running to the
disk’s axle and edge. The galvanometer circuit was stationary in the laboratory. Current
flowed when the magnet was stationary and the disk rotated, or when the disk and magnet
rotated coaxially along the magnet’s north-south axis of symmetry; but not when the magnet
rotated about this axis and the disk was stationary [3, 5].
Faraday subsequently experimented with a rotating conducting magnet connected to a
galvanometer via brushes on the magnet’s axle and rim [5, 6]. Current flowed when the mag-
net rotated around its north-south axis but the galvanometer circuit remained stationary,
or when the magnet was stationary but the circuit rotated. Subsequent researchers have
explored additional permutations in the configurations of Faraday’s experiments [7].
In modern terms, the conducting magnet rotates at velocity v = ω × ρ (for angular
velocity ω and cylindrical radius ρ) through its own magnetic field H (or equivalently,
through its own magnetic flux density B = µH, where µ is the magnetic permeability),
generating a v × B Lorentz force that drives the current.

3
 The electromotive force (emf) around a path C with line element dl is given by [8–10]:
I Z
emf = (E + v × B) · dl = [−∂B/∂t + ∇ × (v × B)] · da, (1)
C S

where E is the electric field, and the area element da is right-hand normal to the surface S
bounded by C. The second equality in Eq. (1) holds via Stokes’ theorem and the Faraday’s
law Maxwell equation provided there is no jump discontinuity on S [11]. This condition will
be met in our work below, and we will calculate the emf using Eq. (1), which we take as
the definition of the term.
For the Faraday disk, for which B is spatially constant and ∂B/∂t = 0, only the v × B
term contributes to the integral. Were the entire circuit rotating at constant ω, the curl of
v × B would be zero and we would have emf = 0. But because the galvanometer circuit is
stationary while the disk rotates in the laboratory frame, the line integral of v × B around
C is nonzero. In any frame at least part of C is in motion. A Poynting theorem analysis
of the Faraday disk shows that the energy for the electric current flowing between axle and
rim in the disk comes from the disk’s kinetic energy of rotation [12]. Taking into account
the small magnetic perturbations to the applied B due to the current that flows in C does
not change these conclusions [12].
The emf in Eq. (1) is often identical to an electromotive force defined by the “flux rule”:

emf Φ = −dΦ/dt, (2)

B · da. Inequality between emf and emf Φ in Eqs. (1) and (2)
R
where magnetic flux Φ = S

in certain circumstances give rise to so-called Faraday paradoxes. Auchmann et al. [11],
consistent with some earlier discussions [13], show that equality requires the path velocity
of the moving surface S (and its boundary C) to be equal to the material velocity of the
conducting medium in which S is embedded. Our applications below meet this requirement.
Faraday concluded from his experiments that magnetic field lines do not rotate with a
magnet when the magnet rotates around its axis of symmetry [3, 4, 6]. But Preston [2]
showed in 1885 that Faraday’s results are equally explained if the magnetic field does rotate
with the magnet, producing a v × B force on the stationary part of C, giving an emf identical
to that of a non-rotating field with the rotating disk. The idea of the field “rotating with
the magnet” was understood [2, 14, 15] to mean that a force qv × B would be experienced
by an electric charge q if q had a velocity v relative to axes fixed in (so co-rotating with) the

4
rotating magnet. This differs from the current understanding of the qv × B force, in which
v is the velocity of q in the frame in which the magnetic flux density is B [13, 16].
Poincaré [17] and others [18] asserted that since both the rotating and non-rotating pic-
tures appeared to give identical results, the distinction between them was meaningless. But
Barnett’s experiments in 1912 [1], reproduced by Kennard [15] and improved upon by Pe-
gram in 1917 [19], demonstrated a difference and resolved the question for the magnetic
fields of electromagnets by using an open circuit. Barnett placed a cylindrical capacitor
axially in the field of a solenoid (or, in analogous experiments, between two large iron flat-
pole electromagnets); a thin wire connected the two concentric cylinders of the capacitor.
Co-rotation of the cylinders and their connecting wire while holding the solenoid stationary
charged the capacitor (due to the v × B force on the wire). After charging, the connecting
wire was disconnected, the system despun, and opposite charges on the cylinders were mea-
sured by electrometer. But rotating the solenoid (or flat-pole electromagnets) while holding
the cylindrical capacitor and connecting wire stationary generated no charge. Co-rotation
of the capacitor and connecting wire together with the solenoid charged the capacitor [19].
Barnett and his contemporaries thereby proved that the field of a rotating axially symmetric
electromagnet does not itself rotate [1, 5, 15, 18, 19].

III. NON-ROTATION OF EARTH’S AXISYMMETRIC FIELD

The Barnett [1], Kennard [15], and Pegram [19] experiments with electromagnets sug-
gest that those components of Earth’s magnetic field that are axisymmetric about Earth’s
rotation axis will be stationary with respect to (do not rotate with) the rotating Earth,
understood in the sense described for rotating electromagnets in the previous section [20–
22]. These would be, for example, the axially symmetric dipole, quadrupole, and octopole
components, with coefficients in the usual Schmidt-normalized Legendre-function expansion
of g10 , g20 and g30 , respectively [23]. Components with coefficients gnm or hm
n where m 6= 0

depend on azimuthal angle ϕ like cos(mϕ) and sin(mϕ), respectively, and therefore rotate
with Earth.
To our knowledge, no experiment has been performed that demonstrates that Earth’s
axisymmetric field does not rotate with Earth. Non-rotation of Earth’s axisymmetric field is
the conservative expectation given the experimental results for electromagnets [1, 5, 15, 19],

5
and this has been taken to be the case by many authors [20–22, 24]. Were the effect we
predict in this paper to be demonstrated, an ancillary consequence would be an experimental
demonstration of the non-rotation of Earth’s axisymmetric field.
Earth’s rotation carries its surface through the non-rotating component of Earth’s mag-
netic flux density B with an azimuthal speed v = 465 sin θ m s−1 at colatitude θ. The
resulting v × B force generates position-dependent volume charge densities (of order 1 e−
m−3 [22]) whose electric field perfectly cancels v × B [20, 24, 25]. A resulting latitude-
dependent surface charge density maintains overall charge balance, with a corresponding
electric potential at Earth’s surface. Any additional motion of individual conductors or
conducting fluids leads to continuous extremely rapid charge redistribution with resulting
perfect cancellation of fields.
However, Earth is surrounded by a conducting ionosphere co-rotating with Earth. Does
this external conducting spherical shell mean that Earth’s axisymmetric magnetic field is
somehow “dragged” into co-rotation with Earth? One might imagine that this is an implica-
tion of Alfvén’s “frozen-flux” theorem [26], which considers Ohm’s law (for current density
J) for a moving conductor
E + v × B = J/σ (3)

in the limit σ → ∞ (a so-called perfect conductor), so that E = −v × B. Then Eqs. (1)

and (2) imply that the magnetic flux Φ cannot change through the surface S as C moves
along — in the usual picturesque language, the flux is “frozen in.”
But consider an axially symmetric conductor rotating with angular speed ω about the
axis of Earth’s axisymmetric magnetic field. Clearly ∂B/∂t = 0 in such a case. This is the
first term in the integrand of the surface integral in Eq. (1). In spherical polar coordinates
(r, θ, ϕ) we have v = ωr sin θϕ̂ and find

∇ × (v × B) = −ω∂B/∂ϕ, (4)

using ∇ × (v × B) = (B · ∇)v − (v · ∇)B + v(∇ · B) − B(∇ · v) and ∇ · B = 0. Therefore

the second term in the integrand in Eq. (1) is also zero due to axisymmetry, and by Eq. (2)
this means dΦ/dt = 0. But this has nothing to do with σ → ∞; it would be just as true for
a very poor conductor as for a perfect conductor. It is therefore not a consequence of the
“frozen-flux” theorem. It is simply a consequence of the symmetry involved, and there is no
reason to view the field as somehow being dragged around with the ionosphere.

6
 In fact, there are well-known examples where translating or rotating conductors do not
“drag” magnetic fields at all, and cause no distortions in the magnetic fields through which
they are moving [22, 27, 28]. This is because the background field is only modified if a
current density J is induced in the conductor; J then induces a magnetic field of its own
via Ampère’s law and it is this induced field that distorts the shape of the overall B field
away from the background field. This distortion often, though not always [27], leads to field
lines that have the appearance of being dragged by the moving conductor. But Van Bladel
has proven that it is impossible to induce a non-zero J for any axially symmetric conductor
rotating in an axially symmetric field [29]. By this theorem a conducting ionosphere rotating
about Earth’s axially symmetric field components cannot induce a J, and therefore cannot
distort (“drag” along with it) Earth’s axially symmetric field. To the contrary, Appendix A
describes how it is Earth’s non-rotating axially symmetric field that brings charged particles
in a conducting plasma around Earth into co-rotation [24].

IV. A PROOF THAT ELECTRIC POWER GENERATION IS IMPOSSIBLE

Could we construct a circuit C in the lab whose rotation along with Earth’s surface
through Earth’s axially symmetric field would generate a continuous electric current via the
v × B force? The emf around any path C is given by Eq. (1). The v × B force experienced
as the conductor C rotates through Earth’s magnetic field drives electron redistribution until
the resulting electrostatic field E perfectly cancels the v × B field: E = −v × B everywhere
within C [5, 22, 30]. Redistribution of charge occurs extremely rapidly, on a classical charge
relaxation timescale τe ∼ 0 /σ ≈ 10−11 (1 S m−1 /σ) s [31]. For very good conductors such as
typical metals for which σ ∼ 107 S m−1 , the relaxation time is given by the electron collision
timescale τc ∼ 107 τe , or ∼ 10−11 s [32]. Since charge redistributes rapidly and continuously
to maintain E = −v × B, emf = 0 by Eq. (1) always. Electric power generation therefore
appears impossible for uniform rotation about an axially symmetric field.
However, this argument contains hidden assumptions. The electric field of a static charge
distribution may always be written as a potential of a scalar field: E = −∇V . But since
∇ × ∇V = 0 always, the equation E = −∇V = −v × B can hold only if ∇ × (v × B) =
0. We will use magnetically permeable materials to violate this requirement, providing a
necessary, but not sufficient, condition for generating a non-zero emf.

7
 Of course, one can always choose to transform to a gauge in which the transformed
scalar potential Ṽ = 0. But then the vector potential transforms to à = (∇V )t, and
E = −∂ Ã/∂t − ∇Ṽ = −∇V , as before [33]. So once again E = −v × B can only hold if
∇ × (v × B) = 0.

V. THE LOOPHOLE IN THE PROOF

Magnetically permeable materials channel magnetic flux, and can be used to alter B
to give ∇ × (v × B) 6= 0. This guarantees that the electrons in such a conductor cannot
rearrange themselves to generate an electrostatic field E = −∇V that satisfies E = −v × B
in Eq. (1). This is the first of our two necessary conditions for electric power generation.
But one could still have E = −v × B if E were no longer purely electrostatic, i.e. if one had
E = −∂A/∂t − ∇V = −v × B, where A is the magnetic vector potential. If this equality
were always to hold for our system, power generation would still be impossible. Are there
circumstances where this equality can be circumvented? This may be answered using the
advection-diffusion equation for A, to which we now turn.
Consider two inertial frames. In frame K at infinity there is a constant background
magnetic flux density (B∞ ), and no electric field (E∞ = 0). A conductor is moving at
constant velocity v = vŷ in K. Frame K 0 is the frame co-moving with the conductor.
Frame K 0 approximates our frame on Earth’s surface (the laboratory frame), translating
through the non-rotating component of Earth’s field. Frame K approximates a non-rotating
frame fixed at Earth’s center and moving with Earth in its orbit.
Frames K and K 0 are not exactly related by a Lorentz boost because of Earth’s rotation.
In K 0 , Maxwell’s equations incorporate rotation via the metric tensor gµν , introducing factors
√
g00 ≈ 1 − 21 (v/c)2 when (v/c)  1 [29]. For v = 465 m s−1 , (v/c)2 ≈ 10−12 . We show
below that these corrections are negligible compared to the effects of interest. We assume
(v/c)2  1 throughout. We may therefore approximate K and K 0 as two inertial frames in
relative linear motion.
Coordinates in the two frames are then related by t0 = t, x0 = x, y 0 = y − vt, and z 0 = z.
We have ∂xµ /∂xν = δνµ and ∂x0µ /∂x0ν = δνµ , ∂t/∂t0 = 1, v = ∂y/∂t0 , ∂/∂t0 = ∂/∂t + v∂/∂y,
and ∂/∂x0 = ∂/∂x, ∂/∂y 0 = ∂/∂y, ∂/∂z 0 = ∂/∂z, so ∇02 = ∇2 . The fields are related by

E0 = E + v × B, (5)

8
B0 = B, and A0 = A. For our system, the curl of Eq. (5) yields ∂B/∂t0 = ∂B/∂t + v∂B/∂y,
i.e. the curl of the field transformation for E0 is just the advective derivative for B. While
B = B0 , Eq. (5) means that E = 0 in K implies E0 = v × B in K 0 .
We begin with an analysis in frame K, and examine results in K 0 in Section XI. Ohm’s
law in K 0 is E0 = J0 /σ, so by Eqs. (3) and (5), J0 = J. We have B = µH and B0 = µH0 [29].
Using E = −∇V − ∂A/∂t and J = ∇ × H, Eq. (3) yields the advection-diffusion equation
for A in K:
−∇V − ∂A/∂t + v × (∇ × A) = η∇ × ∇ × A, (6)

where η = (σµ)−1 is the magnetic diffusivity, here assumed constant. The displacement
current does not appear in Ampère’s law because | 0 ∂E/∂t |  | J | (0 is the vacuum
permittivity) for timescales t  τe [10, 34]. Ohm’s law in K 0 also yields Eq. (6) because of
the field transformation Eq. (5).
The curl of Eq. (6) yields the advection-diffusion equation for B, or “induction equation”:

−∂B/∂t + ∇ × (v × B) = −η∇2 B. (7)

Integrated over S, Eq. (7) is identical to Eq. (1). Therefore

Z I
emf = −η ∇2 B · da = η (∇ × B) · dl. (8)
S C

Whether η∇2 B is negligible in Eq. (7) depends on the magnetic Reynolds number Rm =
τD /τv = σµvξ, where τD = ξ 2 /η is the magnetic diffusion time, and τv = ξ/v the transport
time, for a system that varies over a characteristic length scale ξ. Then | η∇2 B |∼ ηB/ξ 2
and | ∇ × (v × B) |∼ vB/ξ so Rm = | ∇ × (v × B) | / | η∇2 B | [35, 36]. If Rm  1, η∇2 B
is negligible in Eq. (7), so emf = 0. If Rm  1 however, we may have emf 6= 0. Rm  1 is
the second of our two necessary conditions for electric power generation.
In Eq. (1) consider a path C lying within a conducting slab, made say of aluminum for
which σ = 4 × 107 S m−1 and relative permeability µr = 1 (µr = µ/µ0 where µ0 = 4π × 10−7
H m−1 ) [37]. Then for v = 465 m s−1 , Rm  1 if ξ > 1 mm so emf = 0. We instead explore
a system satisfying Ohm’s law with Rm  1 and ∇ × (v × B) 6= 0. As we show below,
one realization is a path C lying within a long cylindrical shell made of an appropriate
magnetically permeable MnZn ferrite [38]. We first consider this system at rest in a frame
K in which B∞ is constant and there is no background electric field (E∞ = 0), in which
case emf = 0. We then give this system a velocity v and show that a nonzero emf will be

9
generated. Such a system at rest in a laboratory on Earth’s surface would therefore generate
electrical power as Earth rotates.

VI. MAGNETICALLY PERMEABLE CYLINDRICAL SHELL

Consider an infinitely long magnetically permeable conducting cylindrical shell with axis
of symmetry along the z-axis, and inner and outer radii a and b, respectively. The back-
ground fields at infinity are B∞ = B∞ x̂ and E∞ = 0 in a frame in which the shell has
v = 0. Of course v × B = 0 and with E∞ = 0 we must have by Eq. (3) J = 0. Therefore
∇ × H = J = 0, so H = −∇W , where W is a magnetic potential. We designate H when
v = 0 as H0 (and define B0 = µH0 ) so ∇ · H0 = ∇2 W = 0, whose solution for a magneti-
cally permeable cylindrical shell for all space is well known in cylindrical (ρ, φ, z) coordinates
[39, 40]. In Cartesian coordinates, the resulting magnetic flux densities exterior to the shell,
within its conducting body, and within its hollow interior are:

B0x (ρ > b) = B∞ + β3 (b/ρ)2 cos 2φ, (9a)

B0y (ρ > b) = β3 (b/ρ)2 sin 2φ; (9b)

B0x (a ≤ ρ ≤ b) = β1 − β2 (a/ρ)2 cos 2φ, (10a)

B0y (a ≤ ρ ≤ b) = −β2 (a/ρ)2 sin 2φ; (10b)

and
B0x (ρ < a) = 2β1 (µr + 1)−1 , (11a)

B0y (ρ < a) = 0. (11b)

Here
β1 = 2B∞ µr (µr + 1)ζ, (12)

β2 = 2B∞ µr (µr − 1)ζ, (13)

β3 = B∞ [1 − (a/b)2 ](µ2r − 1)ζ, (14)

and
ζ = [(µr + 1)2 − (a/b)2 (µr − 1)2 ]−1 . (15)

If a = 0, Eq. (10) collapses to that for a solid magnetically permeable cylinder:

B(ρ ≤ b) = β1 (a = 0)x̂ = 2µr (µr + 1)−1 B∞ x̂, (16)

10
for which the magnetic field is constant in the interior, although of course the exterior (ρ > b)
field is distorted according to Eq. (9) with a = 0.
Because Bz = 0, only the z-component of A is non-zero [9, 27], so

Bx = ∂Az /∂y, (17a)

and
By = −∂Az /∂x. (17b)

Eqs. (9) to (11) then correspond to the vector potential A0 = A0 ẑ, with

A0 (ρ > b) = B∞ y + β3 (b2 /ρ) sin φ, (18)

A0 (a ≤ ρ ≤ b) = β1 y − β2 (a2 /ρ) sin φ, (19)

and
A0 (ρ < a) = 2β1 (µr + 1)−1 ρ sin φ, (20)

with the usual gauge ambiguity allowing the addition of a gradient of a single-valued function.
Moreover, because of Eq. (17), any function of z alone may be added to A0 without affecting
B0 (or E). A0 must be continuous across the boundaries at ρ = a and ρ = b; this is easy
to verify for Eqs. (18) to (20). This requirement means that a choice of gauge on one side
of a boundary restricts the choice of gauge on the other [41]: one cannot arbitrarily assign
different gradient terms (or functions of z) to A0 in each of Eqs. (18) to (20), and this will
prove important below.
From Eq. (19) we see that a solid permeable cylinder has A0 (ρ ≤ b) = β1 y in its interior.
That is, the first term on the right in Eq. (19) is that for a solid cylinder; when a 6= 0 a
second term enters as a modification of this first a = 0 term.
For the region within the body of the cylindrical shell (a ≤ ρ ≤ b), we find by Eq. (10):

∇ × (v × B0 ) = 2vβ2 a2 ρ−3 [(3 sin φ − 4 sin3 φ)x̂ + (3 cos φ − 4 cos3 φ)ŷ)] 6= 0, (21)

using ∂ρ/∂x = cos φ, ∂φ/∂x = −ρ−1 sin φ, ∂ρ/∂y = sin φ, and ∂φ/∂y = ρ−1 cos φ. Such
a translating shell, if made out of conducting material satisfying Rm  1, would therefore
satisfy our two necessary criteria for electric power generation. We show below that in this
case these conditions are also sufficient.

11
 Instead of an infinite shell, consider a finite shell lying along the z axis from −L/2 to L/2,
with L  2b. The magnetic field in the interior (ρ < a) of a finite permeable cylindrical
shell may be written as the sum of two contributions: the field corresponding to the shielded
interior of an infinitely long shell, plus a contribution from the field penetrating in from the
openings [42]. For ρ < a near z = ±L/2, B deviates from B0 , but moving inward this
deviation falls off rapidly like exp(−3.83z/a) [42], so in the interior the result for an infinite
<
shell should hold for a finite shell provided |z| ∼ L/2 − a. For this result to hold for ρ < a,
the field in the region a ≤ ρ ≤ b must be similarly undisturbed, so we take the result for a
<
finite shell in this region to correspond to those for an infinite shell provided |z| ∼ L/2 − a.

VII. TIME-DEPENDENT SOLUTION FOR v 6= 0 AND Rm  1

It is clear that B0 in Eq. (10) can no longer be a solution for our system once v 6= 0,
since B0 could only solve Eq. (7) were ∇ × (v × B0 ) 6= 0, in contradiction to Eq. (21).
We show in Appendix B that B0 (x, y 0 ) = B0 (x, y − vt), i.e. the advecting version of Eq.
(10), also cannot be a general solution when v 6= 0. Any traveling wave solution of the form
B(x, y − vt) solves the transport equation, so will solve Eq. (7) in the limit Rm  1. We
wish to solve Eq. (7) for smaller Rm , when the diffusion term is not negligible.
It is easiest first to solve for A. We therefore solve Eq. (6) explicitly to find A (and so
B) for the magnetically permeable cylindrical shell in the case v 6= 0 (Fig. 1) with Rm  1.
We impose the requirement that in the limit v → 0 we must have A → A0 and B → B0 .
We first work in K, and examine the picture in K 0 in Sec. XII.
Our calculations can be facilitated by a choice of gauge to simplify Eq. (6). We choose
a gauge sometimes used in eddy current [43] or magnetohydrodynamic (MHD) [44] applica-
tions that relates the potentials by the gauge condition:

∇ · A = −V /η. (22)

Because Eq. (22) is less familiar than the more commonly used Lorenz (∇ · A = −µ0 0 ∂V /∂t)
or Coulomb (∇ · A = 0) gauges [45], we discuss it further in Appendix C. In the gauge of
Eq. (22), Eq. (6) simplifies to

−∂A/∂t + v × (∇ × A) = −η∇2 A (23)

12
using the identity

∇ × ∇ × A = ∇(∇ · A) − ∇2 A. (24)

A complete solution to the system is given by Eqs. (22) and (23) together. We have
A = Az ẑ and ∇ · A = ∂Az /∂z. Because v × (∇ × A) = −v∂Az /∂y ẑ, Eq. (23) reduces to
a single non-trivial equation:

∂Az /∂t + v∂Az /∂y = η∇2 Az . (25)

By Eq. (17), a function f (z) may be added to Az without altering B, so f (z) may be chosen
to yield in Eq. (22) the appropriate V expected by physical arguments. However, by Eq.
(1), the emf around C is independent of V , so for the emf it is enough to solve Eq. (23).
If Rm  1, | η∇2 Az || v∂Az /∂y| and Eq. (25) collapses to a transport equation whose
solution is a function of the form Az (x, y − vt). We are interested in the case Rm  1, for
which we expect diffusion to be important. The advection term v∂Az /∂t in Eq. (25) cannot
be neglected even with Rm  1 because η∇2 A0 = 0 and by analogy to MHD [10, 34, 41], we
expect (or at least must not exclude ab initio) | v∂A0 /∂t |∼| η∇2 A1 |, where A1 is a small
perturbation term satisfying | A1 |∼ Rm | A0 |. We seek a solution Az to Eq. (25) that holds
when Rm  1 for any v = vŷ with the requirement Az → A0 as v → 0. Henceforth we set
ξ = b as the relevant diffusion length scale, so put Rm = µσvb, with advection timescale
τv = b/v, and diffusion timescale τD = b2 /η = Rm τv .
We solve Eq. (25) exactly using cylindrical coordinates in K with the origin centered
in the shell at some particular instant; such a solution will hold only for a time short
compared with τv , after which the shell will have moved sufficiently far from the origin that
the cylindrical symmetry assumed in Eqs. (9) to (11) is broken. However, we will see that
the system reaches steady-state extremely rapidly with τD  τv , meaning that B extremely
rapidly adapts itself via diffusion to the shell’s motion [34, 41]. For any location of the
translating shell, we may choose the origin in K to coincide with the center of the shell at
that instant. Since there was nothing special about the instant chosen, this should represent
the steady-state for the system.
The solution to Eq. (25) may in general be written

Az = As (ρ, φ) + At (ρ, φ, t), (26)

13
where As (ρ, φ) solves the steady-state equation

v∂As /∂y = η∇2 As , (27)

and At (ρ, φ, t) solves the time-dependent equation

∂At /∂t = −v∂At /∂y + η∇2 At . (28)

When Rm  1, naive inspection of Eq. (28) suggests At will exponentially decay away
on a timescale ∼ τD [9, 44]. We explicitly solve Eq. (28) by separation of variables using
At = G(ρ, φ)W (t). With separation constant −α2 , this gives

η −1 ∂W (t)/∂t = −α2 W (t), (29)

and
∇2 G − (v/η)∂G/∂y + α2 G = 0. (30)

By Eq. (29),
2
W (t) = C0 e−ηα t ; (31)

all Ci are constants. The alternative choice of separation constant +α2 yields an At (hence
B) exponentially growing with time, so we exclude this solution on physical grounds. Were
Rm  1, Eq. (28) would become the transport equation for which separation of variables
yields a traveling wave solution.
Putting G = g(ρ, φ)eky (a standard technique from MHD [10, 27]) in Eq. (30) yields

∇2 g + λ2 g = 0, (32)

with
k = v/2η (33)

and λ2 = α2 − k 2 . Therefore Eq. (31) becomes

2 +λ2 )t
W (t) = C0 e−η(k . (34)

Solving Eq. (32) by putting g = m(ρ)n(φ) with separation constant ν 2 yields

m(ρ) = C1 Jν (λρ) + C2 Yν (λρ) (35)

and
n(φ) = C3 cos(νφ) + C4 sin(νφ), (36)

14
where the Jν and Yν are Bessel functions of the first and second kinds of order ν. Therefore
2 +λ2 )t
At = C0 m(ρ)n(φ)ekρ sin φ e−η(k . (37)

Since η(k 2 + λ2 ) > 0 always, in Eq. (37) At decays exponentially and the system over
time goes to the steady-state solution As (ρ, φ) in Eq. (26). We therefore cannot choose the
trivial solution As = 0 in Eqs. (26) and (27) since for v = 0 we must have Az = A0 with
A0 given by Eq. (19). Therefore As (v = 0) = A0 . The condition Rm  1 requires v 6= 0 so
solutions for Rm  1 need not satisfy this constraint.
We use boundary conditions and Eqs. (35) to (37) to solve for λ. This allows us to show
explicitly that the exponential in Eq. (37) does indeed decay on a timescale (even faster
than) ∼ τD , consistent with more general arguments [9, 44]. In Eq. (35), we set C2 = 0
so that our solutions remain bounded in the case a → 0. By Gauss’ law, we know that
Bρ must be continuous across the boundary of the cylindrical shell at ρ = a. In the case
of a static external transverse magnetic field, Bρ (ρ < a) ∼ 10−3 B∞ for µr ∼ 5 × 103 , a
value typical of the magnetically permeable materials we will discuss here. That is, the shell
acts as a magnetic shield for its hollow interior [39, 40, 42, 46]. For time-varying fields, the
shielding is as good or better than it is for the static field case [46–48]. We may then take
as a boundary condition for the time-dependent part Bt of B that Btρ (ρ = a) → 0 for all φ
as one approaches the boundary from ρ > a within the shell. Since Btρ = ρ−1 ∂At /∂φ and
At evolves independently of As , this boundary condition then implies that, for all φ,
1 ∂At
= 0. (38)
ρ ∂φ ρ=a

One could try to satisfy Eq. (38) for all φ by setting the product of constants in Eq.
(37), either C0 C1 C3 or C0 C1 C4 , to be proportional to some negative power of µr that goes
to zero for large µr . But this is an unphysical choice, since its effect is to force At to 0
for all ρ within a ≤ ρ ≤ b, meaning that the cylindrical shell magnetically shields itself
throughout its entire volume, as well as its hollow interior. If this unphysical choice were
made nonetheless, it would render At negligible so that only the steady-state solution As
would remain. This conclusion would be the same as that we obtain below, but below it
will be for the reason that At decays away extremely quickly.
Using Eqs. (35) with C2 = 0 and Eq. (37), requiring that Eq. (38) be true for all φ in
turn requires
Jν (λa) = 0 (39)

15
for all ν. Choosing ν = 0, the first zero of the Bessel function gives λ = 2.40/a, and Eq.
(34) becomes
2 t/τ
W (t) = C0 e−(Rm /4)t/τv e−(2.4b/a) D
, (40)

using the definitions of Rm , τv , and τD . The first exponential in Eq. (40) is a decay that
is slow with respect to the translation timescale τv . The second exponential is a decay that
is much faster than the diffusion timescale τD . Since τD = Rm τv , this decay for Rm  1
is extremely fast with respect to τv . Choosing any higher value of ν (or values, were one
to make the solution a series of terms in ν) would yield larger values for λ, leading to even
faster exponential decays in Eq. (40). In the special case a = 0 in Eq. (40), W (t) = 0 so
At = 0 and the full solution is is just the steady-state solution As (a = 0) from Eq. (26).

VIII. TIME-INDEPENDENT SOLUTION FOR v 6= 0, Rm  1

Clearly W (t) → 0 as t → ∞ in Eq. (40), with As in Eq. (26) the steady-state solution
that remains. When Rm  1, W (t) → 0 on a timescale < τD  τv , so At in Eq. (37)
decays rapidly away in the time that it takes the shell to move a distance < vτD , where
vτD  vτv = b. That is, we are — as expected — in a quasi-stationary situation where at
any point in the shell’s translation, Az = As (ρ, φ) to a good approximation, with As given
by Eq. (27). We now solve Eq. (27).
First consider the special case where our cylinder is solid (a = 0) and translating at
v = vŷ through the background field B∞ = B∞ x̂. Then A0 (a = 0) must be given by Eq.
(19) with a = 0, i.e A0 = β1 y + h(z), where h(z) is an arbitrary function of z. This satisfies
Eq. (27) provided h(z) = kβ1 z 2 with k given by Eq. (33), so that

A0 (a = 0) = kβ1 z 2 + β1 y. (41)

By the gauge condition Eq. (22), we then have

V (a = 0) = −vβ1 z. (42)

For a finite solid cylinder, charges in the cylinder experience a v × B force and flow in
response, redistributing on an extremely short timescale τe until an electric field E = −∇V
is established that perfectly cancels v × B. In particular, we see that Eq. (42) gives the
physically correct answer for the special case of a translating finite solid cylinder.

16
 Now what happens when our cylinder becomes a cylindrical shell with a 6= 0? We
anticipate from Eq. (19) that A0 (a 6= 0) will be given by A0 (a = 0) plus additional terms.
We solve Eq. (27) for the general (a 6= 0, v 6= 0) case, with the requirements that we recover
Eq. (41) when a = 0 and Eq. (19) when v = 0.
Eq. (27) is solved by f (ρ, φ)eky , where the function f satisfies

∇2 f − k 2 f = 0 (43)

with k given by Eq. (33), so

f (ρ, φ) = [C5 cos(νφ) + C6 sin(νφ)][C7 Iν (kρ) + C8 Kν (kρ)], (44)

where the separation constant is ν 2 and Iν and Kν are modified Bessel functions of order ν
of the first and second kind. We therefore write the general solution as

As = kβ1 z 2 + β1 y + f (ρ, φ)eky , (45)

where the first two terms provide the solution to Eq. (27) for the case a = 0 and the final
term modifies that solution, analogously to Eq. (19), for the case a 6= 0. Eq. (45) must go
to Eq. (19) in the v = 0 limit. Noting that as kρ → 0 [49]:

K1 (kρ) = (kρ)−1 + kρ(2γ − 1) + (kρ/2) ln(kρ/2) + O(kρ)2 , (46)

where γ = 0.5772... is the Euler constant;

Iν (kρ) = (kρ)ν /(2ν ν!) + O(kρ)ν+2 ; (47)

and
eky = 1 + kρ sin φ + (1/2)(kρ)2 sin2 φ + O(kρ)3 , (48)

requiring Eq. (45) to equal Eq. (19) for v = 0 fixes in Eq. (44) C5 = 0 = C7 and ν = 1,
with C6 = 1 and C8 = −β2 ka2 . Then the solution to Eq. (27) is

Az (a ≤ ρ ≤ b) = As = kβ1 z 2 + β1 y − β2 ka2 K1 (kρ)eky sin φ. (49)

For kρ → 0, Eq. (49) becomes

Az (a ≤ ρ ≤ b) = As = A0 + A1 + O(Rm )2 (50)

17
where A0 is given by Eq. (19),

A1 = −(Rm /2)b−1 β2 a2 sin2 φ, (51)

and Rm = 2kb = µσvb. That is, when Rm  1, Az for v 6= 0 is perturbed away from the
v = 0 solution (A0 ) by a series whose terms are scaled by powers of Rm .
Finally, applying the gauge condition Eq. (22) to A = Az ẑ with Eq. (49), we find

V (a ≤ ρ ≤ b) = −vβ1 z, (52)

so that even when a 6= 0,

∇V = −vβ1 ẑ. (53)

IX. GENERATION OF AN EMF

Eqs. (17a) and (49) yield

Bx (a ≤ ρ ≤ b) = β1 −β2 (a/ρ)2 eky {[kρ cos 2φ+(kρ)2 sin φ]K1 (kρ)−(kρ)2 sin2 φK0 (kρ)}, (54)

using the identities [50]

∂K1 (kρ)/∂(kρ) = −[K0 (kρ) + K2 (kρ)]/2 (55)

and
K0 (kρ) − K2 (kρ) = −(2/kρ)K1 (kρ), (56)

so that
∂K1 (kρ)/∂(kρ) = −K0 (kρ) − (kρ)−1 K1 (kρ). (57)

Noting that [49]

K0 (kρ) = −γ − ln(kρ/2) + O(kρ), (58)

as v → 0 we have by Eq. (54) for Rm  1:

Bx (a ≤ ρ ≤ b) = B0x + B1x + O(Rm )2 , (59)

with
B1x = −Rm b−1 β2 a2 ρ−1 sin φ cos2 φ. (60)

Fig. 2 shows for a particular case how Bx differs from B0x to O(Rm ).

18
 Similarly, Eqs. (17b) and (54) yield

By (a ≤ ρ ≤ b) = B0y + B1y + O(Rm )2 , (61)

with
B1y = −Rm b−1 β2 a2 ρ−1 sin2 φ cos φ. (62)

Fig. 3 shows for a particular case how By differs from B0y to O(Rm ).
We see that B1x = ∂A1 /∂y and B1y = −∂A1 /∂x. Eqs. (60) and (62) are readily checked
to verify that they do solve Eq. (7); e.g. v∂B0x /∂y = η∇2 B1x as required. Eqs. (59)
and (61) show that the effect of v 6= 0 for Rm  1 is to perturb B away from B0 by a
series scaled by successive powers of Rm . The asymmetry of Bx about y = 0 leads to the
continuous generation of an emf within the cylindrical shell. Consider the current path C
in Fig. 1 for x0 = b cos φ0 , and y0 = b sin φ0 . Then Eq. (8) for this path gives:
I
emf(x0 , y0 ) = −η ∇2 Az ẑ · dl = −2Rm vβ2 l(a/b)2 sin φ0 cos2 φ0 + O(Rm )2 , (63)
C

using
η∇ × B = −∇V − η∇2 A (64)

from Eqs. (22) and (24), Az = As , and Eqs. (27), (17), and (59). Eq. (63) is only valid for
Rm  1; if Rm  1, emf = 0 by Eq. (7). Even for Rm  1, emf = 0 in Eq. (63) if v = 0,
or a = 0, or µr = 1. The emf in K 0 is the same as that in K provided (v/c)2  1 [13].
The result in Eq. (63) is for one designated current path C. For an arbitrary C
with segments parallel to the z axis, the integration underlying Eq. (63) leads to emf ∝
[Bx (x1 , y1 ) − Bx (x2 , y2 )], where the coordinates designate the (x, y) coordinates of the two
segments of C parallel to the z axis. (Arbitrary current paths can then be built up by
fusing such rectangular sub-paths.) Because of the symmetry in φ of B0x , Eq. (10a), it
is clear that for every such circuit with 0 < φ < π there is a corresponding circuit with
π < φ < 2π that yields an emf of opposite sign with respect to B0x . This is because the
scalar product (v × B0 ) · dl = vB0x ẑ · dl has the opposite sign in the two cases, since the
circuits are mirror reflections across the y = 0 plane and in each case C is traversed using a
right-hand rule. Over the entire shell, these therefore average to zero. It is the component
of Bx of O(Rm ) that makes a non-zero contribution, because of the asymmetry in φ of B1x ,
with B1x switching sign at the y = 0 plane (Fig. 2). With respect to B1x , for every current

19
path C with 0 < φ < π, there is a corresponding path with π < φ < 2π that yields an emf
of identical sign, so the two do not cancel. To O(Rm ) therefore, the average emf around the
shell cannot be zero. We will see in Sec. XII that, consistent with a non-zero emf, there is
a net absorption of power by the shell from Poynting vector inflow.
An infinite solid conducting bar moving through a background magnetic field will (in
principle) generate a current due to v × B since its infinite extent prevents the accumulation
of the charges at its ends that for a finite bar generates the electric field E = −∇V =
−v × B. However, even for the infinite solid bar, ∇ × (v × B) = 0 so there will be emf = 0
about any closed path C lying within the bar. The nonzero emf in Eq. (63) is not, therefore,
attributable to the fact that our formalism began with an infinite cylindrical shell.

X. INTUITIVE PHYSICAL PICTURE

A simple physical picture offers insight into why a magnetically permeable cylindrical
shell moving at velocity v and satisfying Rm  1 would be expected to generate an emf
according to Eq. (8). In frame K, picture a finite cylindrical shell moving transversely to
its long axis (Fig. 1). Assume a 6= 0. By Eq. (21), we know that it is impossible for the
shell’s electrons to establish a configuration such that −∇V = −v × B.
Imagine beginning with the cylindrical shell at rest and then placing it into motion at
velocity v. The magnetic flux density within the shell itself is initially B0 , given by Eq.
(10). B0 results from Maxwell’s equations requiring the continuity of the normal component
of B and tangential component of H at the surfaces ρ = a and ρ = b. As the shell moves
it attempts, so to speak, to enforce B = B0 throughout a ≤ ρ ≤ b. If τD /τv = Rm  1,
the diffusion timescale τD for the magnetic flux density is much longer than the advection
timescale τv . That is, diffusion is negligible compared to advection and the field B(a ≤ ρ ≤ b)
advects along with the shell, so that B = B0 to very high precision. (Alfvén’s frozen-flux
theorem [26] holds.) Since ∇ × B0 = 0, by Eq. (8) we must have emf = 0 when Rm  1.
Contrast this with the case τD /τv = Rm  1. Now the timescale τD for diffusion is
much shorter than τv . That is, as the shell moves, the field’s adjustment is dominated not
by advection but by diffusion, toward a field configuration at which diffusion would stop,
i.e. toward B0 (a ≤ ρ ≤ b), where the “destination” B0 is the value that would apply for a
stationary shell at the location to which the shell has just moved. The field can never reach

20
this end point since the shell keeps moving even as the field diffuses, so a steady-state is
reached in which the diffusing field differs slightly from B0 . The field within the shell does
not adjust instantaneously to the shell’s motion, so never (unless the shell is brought to rest
in K) fully “catches up” to that motion. A closed path C moving with the shell constantly
experiences a field that is diffusing across its boundaries, and Eq. (8) in general is nonzero.

XI. ANALYSIS IN THE LABORATORY FRAME

We now consider our system in the laboratory frame K 0 , where v = 0 so there is no

magnetic v × B force, but there is instead an electric field given by the field transformation
Eq. (5): E0 = E + v × B. Ohm’s law in K 0 is simply E0 = J0 /σ, which leads to the induction
equation in K 0 :
∂B/∂t0 = η∇2 B. (65)

Since J0 = J and B0 = B when (v/c)2  1, the emf is given by

I I I
0 0 0 −1
emf = E · dl = σ J · dl = η (∇ × B) · dl. (66)
C C C

We have used the fact that E0 and therefore J must be parallel to ẑ, so the relevant part
of dl is perpendicular to y and therefore we can put dl0 = dl. Eq. (66) is identical to Eq.
(8) and therefore to Eq. (63), so emf 0 = emf, as expected [13]. Eq. (66) is nonzero in K 0
provided the same conditions hold as those needed for Eq. (63) to give emf 6= 0.
It might nevertheless seem puzzling that an emf could be generated in K 0 . We intuitively
expect ∂B/∂t0 = 0 in steady state, so that by Eqs. (8) and (65), emf 0 = 0. But care must
be taken with B in Eq. (65): because B is not rotating with Earth, it cannot be treated
implicitly as B(x, y 0 ), where y 0 = y − vt relates the coordinates in K 0 and K. If B were
rotating with Earth, then in K 0 we would simply have B = B(x, y 0 ) and ∂B(x, y 0 )/∂t0 =
(∂B/∂y 0 )(∂y 0 /∂t0 ) = 0, using the chain rule and ∂y 0 /∂t0 = 0.
But treating B as rotating with Earth is inconsistent with the clear expectation from
the results of the Barnett [1], Kennard [15], and Pegram [19] experiments. Rather, in K 0
we must treat B as advecting through the cylindrical shell at velocity v = −vŷ, which
we capture by writing B = B(x, y) with y = y 0 + vt. The time-dependence of B(x, y) is
driven by the advection of B through the shell; this dependence is included implicitly by

21
y = y(y 0 , t) = y 0 + vt. Then by the chain rule and ∂y/∂t0 = v,

∂B/∂t0 = ∂B(x, y(y 0 , t))/∂t0 = (∂B/∂y)(∂y/∂t0 ) = v∂B/∂y. (67)

That is, we do have ∂B(x, y 0 )/∂t0 = 0, but we also have ∂B(x, y)/∂t0 = v∂B/∂y 6= 0, and we
must distinguish between the two. Only the second representation for B in K 0 is consistent
with experiment.
Substituting Eq. (67) into (65) gives a time-independent equation for B:

v∂B/∂y 0 = η∇2 B. (68)

Recalling ∂/∂y 0 = ∂/∂y, Eq. (68) yields Eq. (27) for A given the gauge choice Eq. (22).
Physically, the induction (advection-diffusion) equation concerns the steady-state that is
reached in B as it advects through the Rm  1 cylindrical shell and undergoes concomitant
diffusion; as a result B (as we know from Eqs. (59) to (62)) is slightly perturbed away from
B0 . Were B instead advecting along with the shell, there would be no emf.
A Poynting vector and flux transport analysis [41, 51, 52] in K 0 make it clear that energy is
flowing into our Rm  1 cylindrical shell, providing the power required to sustain emf 0 6= 0.
The Poynting vector in K 0 is

S0 = µ−1 (E0 × B) = µ−1 η(∇ × B) × B, (69)

where we have used E0 = J/σ and Ampère’s law. Were it the case that B = B0 , we would
have S0 = 0 by ∇ × B0 = 0 in Eq. (69) and there would be no energy input to the cylindrical
shell. However, ∇ × B1 6= 0 and using Eqs. (24), (23), (53), (27), and (17), we find

η∇ × B = v(β1 − Bx )ẑ, (70)

giving
S0 = vµ−1 (β1 − Bx )(Bx ŷ − By x̂). (71)

We perform our calculations at the instant at which the origins of the K 0 and K frames
coincide. The net energy flux PS0 into the shell’s surface within l/2 ≤ z ≤ l/2 (where l/2 is
chosen to be sufficiently far in from the shell’s edge at L/2) is given by:
Z 2π Z l/2
PS0 = S0 ·ρ̂ ρdφdz, (72)
0 −l/2

22
where ρ̂ = cos φ x̂ + sin φ ŷ and the boundaries at both ρ = b and ρ = a must be taken into
account by summing the contributions from evaluating Eq. (72) at ρ = b and at ρ = a. The
boundary at ρ = a enters with a negative sign, opposite from that at ρ = b. The calculation
is simplified by noting

(Bx ŷ − By x̂) · ρ̂ = −By cos φ + Bx sin φ = [β1 + β2 (a/ρ)2 ] sin φ + O(Rm )2 , (73)

i.e. the O(Rm )1 terms cancel in Eq. (73). In Eq. (72), nearly all terms integrate to zero,
and
PS0 = (π/4)σv 2 β22 a2 [1 − (a/b)2 ]l + O(Rm )2 . (74)

PS0 = 0 if v = 0, or µr = 1 (because then β2 = 0), or a = 0. Otherwise, the Poynting vector

S0 in K 0 gives a net energy flow into the cylindrical shell that sustains the emf 0 .
It is interesting to ask which terms within S0 provide this energy. Nearly every term in
Eq. (71) makes zero contribution to Eq. (72), either because it cancels an identical term of
opposite sign, or because it integrates to zero over φ: the energy from most terms simply
flows through the shell, with as much energy leaving as entering. The only term in S0 that
makes a nonzero contribution is

S0x = 2σv 2 β22 a4 ρ−3 sin2 φ cos φ(4 cos2 φ − 1) x̂. (75)

Eq. (65) can be written [41, 52]:

∂B/∂t0 = ∇ × (w × B) (76)

for an appropriate velocity w. By Ampère’s and Ohm’s law in K 0 , we have

E0 = η∇ × B, (77)

so E0 · B = 0 since B = (Bx , By , 0). When E0 · B = 0, w in Eq. (76) is [41, 52]:

w = (E0 × B)/B 2 , (78)

which is called the “flux transporting velocity” [52], meaning that w for the case σ 6= ∞
preserves flux, because it satisfies Eq. (76) in the same way that v satisfies Eq. (7) for the
case σ = ∞. A contour C in the cylindrical shell will therefore have dΦ/dt = 0 through its
corresponding surface S if C is moving at w, where w may vary point-to-point along C.

23
 By Eqs. (68) and (78),
w = µS0 /B 2 . (79)

Direct calculation using Eqs. (79) and (71) reveals w to be algebraically complicated with
w 6= v, even while satisfying ∇ × (w × B) = ∇ × (v × B), as it must. When w 6= v, then
if C is simply being transported with the conductor at v 6= 0, there must be an emf around
C. This is the case, for example, for the contour C in Fig. 1. By Eqs. (78), the zero-velocity
solution B0 in K 0 is not transported through the shell — because in this case E0 = 0 by
Eq. (77), so w = 0 in Eq. (78). Of course, B0 does not diffuse: ∇2 B0 = 0. Only the
perturbations B1 and higher orders will have w 6= 0.
Eq. (79) means that magnetic flux is transported proportionally to the transport of
energy defined by the Poynting vector. Since S0 integrated over the cylindrical shell is non-
zero, there is a corresponding net flow of magnetic flux into the shell. We have a picture
in K 0 in which by Eq. (75), near the x = 0 plane for π/3 < φ < 2π/3 and again for
4π/3 < φ < 5π/3, magnetic field lines are diffusing (transported at velocity w) in the x
direction vertically toward the x = 0 plane from above and below. These lines annihilate
[9, 52, 53] in the x = 0 plane, providing energy that drives the current flow in C. The
cancellation (annihilation) of the magnetic filed lines in the x = 0 plane preserves the
gradient, which in turn maintains the continuing inward diffusion of the field.
We note an analogy to the homopolar generator. By Eqs. (67) and (76),
I
0
emf = (w × B) · dl, (80)
C

where w is given by Eq. (78). In the homopolar generator, the analog to Eq. (80) gives
emf 0 6= 0 because only part of C is rotating, so v varies (stepwise) around C and v × B does
not integrate to 0 around C. In the Rm  1 cylindrical shell of Fig. 1, emf 0 6= 0 because w
varies around C according to Eq. (79), so w × B does not integrate to 0 around C.

XII. POYNTING’S THEOREM AND MAGNETIC BRAKING

When Rm  1, in the steady-state (e.g. Eq. (45)) we have ∂A/∂t = 0, so E = −∇V =

vβ1 ẑ by Eq. (53). Together with Ampère’s law and Eq. (69), we have

E · J = σv 2 β1 (β1 − Bx ). (81)

24
By Eq. (3) we also have
E · J = σ −1 J 2 + (J × B) · v. (82)

Integrating Eq. (81) over the volume dV = ρdφdρdz gives zero, so Eq. (82) implies
Z Z
−1 2
σ J dV = − (J × B) · v dV, (83)
V V

where the integral on the right-hand side is the familiar expression for magnetic braking.
In K, Joule heating therefore derives from the energy made available by magnetic braking
of the cylindrical shell. Since the shell is being carried by Earth, it is clear that electrical
power in our system derives ultimately from the kinetic energy of Earth’s rotation. This is
analogous to the Poynting theorem analysis of the homopolar generator [12].
Explicitly integrating (J × B) · v in Eq. (83) over the volume V of the shell with J =
µ−1 ∇ × B shows the power removed from Earth’s rotational kinetic energy to be:
Z b Z 2π
2
Pk = −σv l Bx (β1 − Bx )ρdρdφ = (π/2)σv 2 β22 la2 [1 − (a/b)2 ] + O(Rm )2 . (84)
a 0

Were B = B0 , we would have Pk = 0 since ∇ × B0 = 0. If v = 0 or µr = 1 or a = 0, then

Pk = 0. The power in K must equal that in the laboratory frame K 0 to O(v/c)2 [13].
The manner in which this power arises in K 0 is of interest. Poynting’s theorem [54] states
that the rate at which work is done on the electrical charges within a volume V of surface
area Σ is equal to the decrease in energy stored in the electric and magnetic fields, minus
the energy that flowed out through the surface bounding the volume. In K 0 , Poynting’s
theorem is:
Z Z Z
0 −1 0
E · J dV = −µ B · ∂B/∂t dV − S0 · dΣ, (85)
V V Σ

where as before the displacement current is negligible. By Eq. (65), E0 · J = σ −1 J 2 . The

second term on the right of Eq. (85) is just Eq. (72). The first term on the right may
be evaluated using Eq. (76), a calculation most easily performed with B in cylindrical
coordinates (Appendix D). The result is identical to Eq. (74).
Therefore in K 0 , Eq. (85) gives the power PP0 provided to the shell, to be
Z
PP0 =σ −1
J 2 dV = (π/2)σv 2 β22 a2 [1 − (a/b)2 ]l + O(Rm )2 , (86)
V

with the energy for electrical power being provided in Poynting’s theorem coming equally
from Poynting vector inflow and the ∂B/∂t0 term. This is indeed equal to the expression

25
found in Eq. (84) by calculating in the K frame: Pk = PP0 . For a given choice of b, Eqs.
√
(84) and (86) reach their maximum values for a = b/ 2.
An important question is whether the slight de-spinning of Earth caused by the mag-
netic braking found here is consistent with angular momentum conservation. Mechanical
systems can come in or out of rotation solely via the transfer of angular momentum between
mechanical rotation and the electromagnetic field. (For the fully calculated example of a
charged magnetized sphere exhibiting this behavior, see [55, 56].) The angular momentum
of the electromagnetic field is proportional to r × S (with r the usual radial component in a
spherical coordinate system). For the system in our thought experiment, the analogous issue
is linear momentum conservation. In K, the mechanical momentum of the cylindrical shell
lies in the ŷ direction, and the braking force per unit volume, given by J × B, acts in the
−ŷ direction. The momentum (per unit volume) of the electromagnetic field associated with
this system is p = ε0 µ0 S, with S = µ−1 E × B = (µσ)−1 J × B, i.e. p is proportional to, and
lies in the direction of, the magnetic braking vector. Therefore, as the system is braked,
positive mechanical linear momentum is lost from the cylindrical shell while negative linear
momentum is lost from the electromagnetic field, and momentum conservation is possible.

XIII. EXPERIMENTAL PREDICTIONS

Evaluating the emf expected to be measured in a laboratory test of these claims requires
that part B∞ of Earth’s total field that is axially symmetric about the planet’s rotation axis.
This is well approximated by summing the axisymmetric dipole, quadrupole, and octupole
components of the total field to yield the northward (X) and downward (Z) components of
B∞ at a point on the surface of the Earth. These are [23], at colatitude θ,

X = −g10 sin θ − 3g20 sin θ cos θ − (3/2)g30 sin θ(5 cos2 θ − 1) (87a)

and
Z = −2g10 cos θ − (3/2)g20 (3 cos2 θ − 1) − 2g30 cos θ(5 cos2 θ − 3), (87b)

giving
B∞ = (X 2 + Z 2 )1/2 , (87c)

where the Gauss coefficients g10 = -29496.5 nT, g20 = -2396.6 nT, and g30 = 1339.7 nT [57].
Then for, say, Princeton’s colatitude θ = 49◦ 390 (for which v = 354 m s−1 ) B∞ = 45 µT ,

26
pointed downward into Earth’s surface at an angle (from the horizontal when facing the
north geographic pole) tan−1 (Z/X) = 57.5◦ .
√
Suppose that our cylindrical shell had dimensions l = 20 cm, b = 1 cm and a = b/ 2,
and was made of MN60 MnZn ferrite, with data sheet values given to be µr = 6, 500 ± 3, 000
√
and σ ≈ 0.5 S m−1 [58]. Then Rm = 1.4 × 10−2  1, while Rm  (v/c)2 ensures that g00
effects are small compared to first-order perturbations scaled by Rm . For φ0 = 45◦ , Eq. (63)
gives emf = 65 µV .
By inspection of the integral in Eq. (63), the emf should reverse sign when the shell
(together with the attached measuring apparatus, a digital voltmeter; see Fig. 1) is ro-
tated by 180◦ . This is a striking prediction that should separate an emf generated by the
effect predicted here from other types of emf generation. Our derivation is only valid for
v transverse to the shell, but the emf must pass through zero between the two transverse
orientations that are separated by 180◦ . A voltmeter across d and f in Fig. 1 would measure
half the emf around C in Eq. (63). We caution that C may “choose itself” under rotation,
and experiment will show whether a voltage measurement actually somehow averages over
many possible current paths. If so, we may approximate the expected emf by averaging over
ρ and φ in the calculation leading to Eq. (63):
Z bZ π
1
< emf >= − vBx l dρdφ = −(4/3π)Rm vβ2 l(a/b)2 (1 − a/b)−1 ln(b/a), (88)
π(b − a) a 0

which for the identical parameter values as above gives < emf >= 46 µV. Once again, emf
measured as in Fig. 1 yields half this value, and the sign reverses under 180◦ rotation.

XIV. SCALING AND CONCLUSIONS

The cylindrical shell was chosen as an especially simple realization of a conductor with
∇ × (v × B) 6= 0, and MN60 material was chosen to provide Rm  1 on a laboratory scale.
For the MN60 device considered above, Pk ≈ 16 nW by Eq. (84). By the maximum power
transfer theorem at most half of this power can be transferred to the load [59]. To be useful,
the effect must be scaled up greatly in voltage and power. One way might be to maintain
Rm  1 while increasing σ, by decreasing µr , b, and therefore a. Carbon nanotubes can be
coated with materials such as iron [60, 61], so very small low-Rm magnetically permeable
tubes seem plausible. One must also consider resistance and ohmic loss.

27
 It should be possible to separate the magnetic shield producing ∇×(v × B) 6= 0 from the
conductor providing Rm  1. For example, note that the functional form of Eq. (9), for the
magnetic flux density outside a magnetically permeable shell, is identical to that of Eq. (10)
for the flux density in the shell’s interior. This guarantees that Eq. (21) holds outside the
shell with the substitutions β2 → −β3 and a2 → b2 . Therefore we should be able to realize
the effect using a magnetically permeable cylinder surrounded by an insulated concentric
cylindrical shell of a non-permeable low-Rm material, and find results analogous to those
found above. Graphite has σ = 7.3 × 104 S m−1 [37], giving Rm ≈ 2 × 10−2 for b = 1 mm,
so we can hope to realize the effect for a mu-metal or ferrous cylindrical core surrounded
by a thin insulator with an overlying shell of graphite. Decreasing b to 5 µm would allow
copper (σ = 6.0 × 107 S m−1 [37]) or other common metals to be used for the outer layer,
with obvious advantages. Altogether different topologies and materials are possible.
The effect predicted here would be available nearly globally and with no intermittency,
but requires testing then further examination to see if it or some other configuration based
on broadly similar principles could be scaled to practical emission–free power generation.
Devices could have important practical implications even if only voltages of ∼ 1 volt could
be achieved. Such a device would represent a small-application power supply whose lifetime
would be limited only by material degradation. At the other, extreme end of speculations
regarding generated power, we note that global installed power generation capacity is pro-
jected to grow to 10,700 GW by 2040 [62]. Imagine as an upper limit that human civilization
generated this power entirely from Earth’s rotation through its magnetic field. Over a cen-
tury, the resulting kinetic energy loss would increase Earth’s rotation period by 7 ms. This
may be compared to fluctuations in the length of Earth’s day of 10 ms over time intervals of
several decades [63], and an observed long-term increase (dominated by lunar tidal recession)
of 2.5 ms per century [64].

ACKNOWLEDGMENTS

We thank three anonymous referees for their reviews, and are grateful for helpful discus-
sions with T. H. Chyba, R. L. Garwin, M. J. Rees, P. J. Thomas, E. L. Turner, and three
anonymous colleagues. We thank G. Z. McDermott and G. Cooper for administrative sup-
port, B. A. Lin for reference assistance, and M. Northrup of National Magnetics Group Inc.

28
for help with materials. CFC acknowledges research funds from the Woodrow Wilson School
and the Department of Astrophysical Sciences at Princeton University. KPH acknowledges
support through the Jet Propulsion Laboratory, California Institute of Technology, under a
contract with the National Aeronautics and Space Administration, and through the NASA
Exobiology Program (NNH09ZDA001N).

APPENDIX A: CO-ROTATION OF THE IONOSPHERE

What happens to charged particles in a conducting plasma around Earth in the presence
of Earth’s non-rotating axially symmetric field? Hones and Bergeson [24], building on Davis
[20, 25] and Backus [30] examined this question for the complicated general case of mag-
netic fields with both axisymmetric and non-axisymmetric components, treating the purely
axisymmetric case as a special case of their more general result. Here we follow their overall
logic but present the simpler calculation for the axisymmetric component only: the special
case of a magnetic dipole aligned antiparallel to Earth’s rotation axis.
An observer in a non-rotating frame sees a (non-rotating) dipole field anti-aligned with
Earth’s axis, given by Br = −(2M/r3 ) cos θ, Bθ = −(M/r3 ) sin θ, and Bϕ = 0, where M
is a constant proportional to the magnetic dipole moment. In general the electric field
seen in this frame is given by E = −∂A/∂t − ∇V . However, for our non-rotating dipole
we can put ∂A/∂t = 0 so E = −∇V . Earth’s rotation through its own dipole field
leads to an electrostatic field within the Earth that balances the resulting v × B force:
E = −∇V = −v × B, which gives V = −(M ω/r) sin2 θ, and a surface potential for Earth
(of radius R⊕ ) of V (r = R⊕ ) = −(M ω/R⊕ ) sin2 θ. Fields E and B in the plasma must sat-
isfy E · B = 0, due to the plasma’s near-infinite conductivity parallel to the magnetic field
lines. This condition gives 2 cos θ ∂V /∂r = −(1/r) sin θ ∂V /∂θ. The solution consistent with
the boundary condition at r = R⊕ is V = −(M ω/r) sin2 θ, so that Er = (M ω/r2 ) sin2 θ,
Eθ = −(2M ω/r2 ) sin θ cos θ, and Eϕ = 0. (Note that this satisfies E∞ = 0.) Charged
particles in this plasma drift azimuthally at a velocity v = (E × B)/B 2 ; direct calculation
gives v = −ωr sin θ φ̂ for particles of any charge or mass. That is, the ionosphere comes into
co-rotation with Earth because the charged particles composing it acquire exactly the neces-
sary co-rotation velocity from their interactions with Earth’s non-rotating axially symmetric
field together with the electric field induced in the ionosphere. Earth’s rotation through the

29
non-rotating axisymmetric component of its magnetic field drives ionospheric co-rotation.
The non-axisymmetric components — those components that give Earth’s magnetic field
its tilt away from Earth’s rotation axis — of course do rotate with Earth. Since magnetic
field lines are defined as lines everywhere tangent to the magnetic field, an observer well
away from Earth who could somehow see field lines would see Earth’s tilted dipole lines
rotating with Earth — the rotating lines being the vector sum of a non-rotating azimuthally
constant component plus a rotating azimuthally varying component.
There is a standard result that magnetic lines of force in a perfectly conducting fluid move
with the fluid — the fluid is “line-preserving.” [9, 51, 65]. (However, magnetic field lines are
not relativistically covariant [51], and their reality must be treated with care [21, 66, 67].)
When we calculate the equations for the magnetic field lines of a tilted dipole, we find that
these lines are described by axisymmetric time-independent terms (from the non-rotating
axisymmetric dipole) plus terms sinusoidal in ωt, i.e. terms that rotate with Earth. The field
lines do indeed vary sinusoidally with ωt, due to the superposition of a rotating component
on top of an underlying axially symmetric component.
Magnetic field lines must satisfy dl × B = 0, where dl is the arc length. This leads to
the usual condition
dr/Br = rdθ/Bθ . (A1)

Earth’s magnetic potential U , taking into account only the lowest-order terms for the ax-
isymmetric dipole (g10 ) and inclined dipole (g11 and h11 ) terms, is [23]:

U = g10 (a3 /r2 ) cos θ + (a3 /r2 )(g11 cos ϕ + h11 sin ϕ) sin θ, (A2)

where g10 = −29496.5 nT, g11 = −1585.9 nT, and h11 = 4945.1 nT [57]. Because of Earth’s
rotation, a non-rotating observer co-orbiting with Earth would see ϕ = ωt where ω is Earth’s
angular speed. Using Eq. (A1) with Br = −∂U/∂r and Bθ = −r−1 ∂U/∂θ, we find

Br = 2g10 (a/r)3 cos θ + 2(a/r)3 (g11 cos ϕ + h11 sin ϕ) sin θ, (A3)

Bθ = g10 (a/r)3 sin θ − (a/r)3 (g11 cos ϕ + h11 sin ϕ) cos θ, (A4)

and
dr g 0 cos θ + (g11 cos ϕ + h11 sin ϕ) sin θ
= 2dθ 10 . (A5)
r g1 sin θ − (g11 cos ϕ + h11 sin ϕ) cos θ

30
Now since g11 /g10 ≈ 0.05 and h11 /g10 ≈ 0.17, we may roughly approximate this as
dr
≈ 2dθ{cot θ + [(g11 /g10 ) cos φ + (h11 /g10 ) sin φ] csc2 θ}. (A6)
r
Integrating, then exponentiating both sides and using a Taylor expansion yields

r ≈ r0 sin2 θ − r0 [(g11 /g10 ) cos ϕ + (h11 /g10 ) sin ϕ] sin 2θ, (A7)

where r0 is a constant of integration and ϕ = ωt. The first term in Eq. (A7) is identical
to the usual equation for the field lines of an axisymmetric dipole field [9]. The next term
gives the inclined dipole and its rotation with Earth. An observer rotating with Earth at
a particular ϕ could interpret what he or she sees as co-rotating field lines with a shape
specific to that value of ϕ. An observer looking back at Earth who could see field lines
would see an inclined dipole rotating with Earth.

APPENDIX B: FAILURE OF THE v = 0 SOLUTION

We demonstrate that the v = 0 solution B0 (Eq. (10)) is no longer a solution for the
magnetically permeable cylindrical shell once the shell is moving with v = vŷ in K (Fig.
1). We assume B0 (or equivalently, A0 with allowance for gauge ambiguity) remains the
solution even though v 6= 0, and show that this leads to a contradiction.
When v = 0, we have B∞ (ρ  b) = B∞ x̂ and E∞ (ρ  b) = 0 in K. These must continue
to hold once v 6= 0, since the shell’s distortion of the fields must go to zero at infinity.
First assume B0 (x, y, z, t) to be a solution for the a 6= 0 cylindrical shell for v 6= 0.
Inserting Eq. (10) into Eq. (7) requires ∇ × (v × B0 ) = 0. But we know by Eq. (21) that
this is false in general. Therefore B0 (x, y, z, t) cannot be a solution when v 6= 0.
Rather than assuming a solution B0 (x, y, z, t), we instead treat the disturbance in the
background field B∞ as moving together with the cylindrical shell at v. We implement
this in Eqs. (9) to (11) by referring the coordinates of B0 to the K 0 system (x0 , y 0 , z 0 , t0 ) =
(x, y − vt, z, t). For example when v 6= 0, Eq. (9a) would be

0
B0x (ρ0 > b) = B∞ + β3 (b/ρ0 )2 cos 2φ0 , (B1)

where ρ0 = (x2 + y 02 )1/2 , φ0 = tan−1 (y 0 /x), and of course y 0 = y − vt. Correspondingly, Eq.
(18) becomes
A00 (ρ0 > b) = B∞ y 0 + β3 (b2 /ρ0 ) sin φ0 . (B2)

31
Henceforth in this discussion primed field quantities are understood to be written in terms
of the coordinate y 0 . In the limit v → 0, Eqs. (B1), (B2), and their analogs go to Eqs. (9)
to (11) and (18) to (20), as required.
In this appendix only we make the following simplifying choice of gauge [33, 41]:
Z
0 0 0 0
A → Ã = A + ∇ V 0 dt0 , (B3a)

so that
0 0 ∂ Z 0 00
V → Ṽ = V − 0 V dt = 0. (B3b)
∂t
I.e. the corresponding gauge condition is V 0 = 0. Because V 0 = V − vAy = V since only Az
is nonzero, we have V 0 = V = 0. Henceforth dropping the tilde on A0 , we have

E0 = −∂A0 /∂t0 . (B4)

Eqs. (B2) and (B4) give E0 (ρ0 > b) = 0, so by Eq. (5), E(ρ > b) = vB0x ẑ, where B0x is
given by Eq. (9a). But by taking ρ → ∞ in Eq. (9a), this means E∞ = vB∞ ẑ, which
contradicts our premise that E∞ = 0 in K. Therefore Eq. (B2) cannot be a solution for the
magnetically permeable cylindrical shell once v 6= 0.
But perhaps we can add a piece to A00 that preserves B00 while giving E∞ = 0? (This
would not be a gauge transformation, as we would be explicitly attempting to alter the field
quantity E while preserving B00 .) We now show that this is impossible, so that there is no
modification of Eqs. (18) to (20) that both maintains B0 = B00 and is consistent with the
requirement that B∞ (ρ  b) = B∞ x̂ and E∞ (ρ  b) = 0. Whatever term is added to Eq.
(B2) cannot vary with x or y 0 , or else B00 will be changed, in contradiction to our premise. If
we tried instead to add a spatially constant term vB∞ t0 to Eq. (B2) to alter E0 and thereby
0
E∞ , by Eq. (17) we would still change B0x by vB∞ (∂t0 /∂y) = B∞ , again contradicting
a premise. We have therefore shown that B00 (x0 , y 0 , z 0 , t0 ) = B00 (x, y − vt, z, t) cannot be a
solution when v 6= 0. In effect, the “solution” B00 (x, y − vt, z, t) is incompatible with the
premise that B∞ does not rotate together with the frame K 0 .

APPENDIX C: CHOICE OF GAUGE

While the gauge condition Eq. (22) is cited in the literature [43, 44], it is not included
in lists of standard electrodynamics gauges [68]. We therefore discuss it further here, and

32
show that it satisfies the requirements of gauge invariance. A gauge transformation leaves
B and E unchanged provided the transformed vector and scalar potentials satisfy

à = A + ∇χ (C1)

and
Ṽ = V − ∂χ/∂t. (C2)

Now take the divergence of Eq. (C1), multiply it by η, and add to this Eq. (C2), to obtain:

Ṽ + η∇ · Ã = V + η∇ · A + (−∂χ/∂t + η∇2 χ). (C3)

The gauge condition Eq. (22) therefore holds both before and after the gauge transformation
Eqs. (C1) and (C2) provided χ satisfies the diffusion equation

∂χ/∂t = η∇2 χ. (C4)

That is, gauge invariance with χ satisfying Eq. (C4) leads to the gauge condition Eq. (22).

APPENDIX D: CYLINDRICAL COORDINATE REPRESENTATION

Some calculations are most easily performed with B0 and B1 in cylindrical coordinates.
For convenience, we give this representation here. We have

Bρ = Bx cos φ + By sin φ, (D1a)

Bφ = −Bx sin φ + By cos φ, (D1b)

so that, from Eq. (10):

B0ρ (a ≤ ρ ≤ b) = [β1 − β2 (a/ρ)2 ] cos φ; (D2a)

and
B0φ (a ≤ ρ ≤ b) = [−β1 − β2 (a/ρ)2 ] sin φ. (D2b)

Using Eq. (D2) and v = vŷ = v sin φ ρ̂ + v cos φ φ̂ yields a simpler expression for Eq. (21):

∇ × (v × B0 ) = 2vβ2 a2 ρ−3 [sin 2φ ρ̂ + cos 2φ φ̂]. (D3)

We also have, from Eqs. (60) and (62):

B1ρ (a ≤ ρ ≤ b) = −(1/2)Rm b−1 β2 a2 ρ−1 sin 2φ; (D4a)

33
and
B1φ (a ≤ ρ ≤ b) = 0. (D4b)

The first term on the right hand side of Eq. (85) may then be evaluated via Eq. (65) and the
vector Laplacian in cylindrical coordinates. The calculation is tedious but straightforward.

∗ [email protected]; http://wws.princeton.edu/faculty-research/faculty/cchyba
† [email protected]; https://science.jpl.nasa.gov/people/Hand/
[1] S. J. Barnett, “On electromagnetic induction and relative motion,” Phys. Rev 35, 323 (1912).
[2] S. T. Preston, “On some electromagnetic experiments of Faraday and Plücker,” Phil. Mag.
19, 131–140 (1885).
[3] M. Farady, Faraday’s Diary of Experimental Investigations 1820 – 1862, edited by T. Martin,
Vol. I (HR Direct, 2008).
[4] M. Faraday, “Experimental researches in electricity,” Phil. Trans. R. Soc. Lond. 122, 125
(1832).
[5] A. I. Miller, “Unipolar induction: a case study of the interaction between science and tech-
nology,” Ann. Sci. 38, 155–189 (1981).
[6] M. Faraday, “Experimental researches in electricity: twenty-eighth series,” Phil. Trans. R.
Soc. Lond. 142, 25 (1852).
[7] A. G. Kelly, “Unipolar experiments,” Ann. Fond. Louis Broglie 29, 119–148 (2004).
[8] L. D. Landau, E. M. Lifshitz, and L. P. Pitaevskii, Electrodynamics of Continuous Media,
2nd ed. (Elsevier, 2004).
[9] G. K. Parks, Physics of Space Plasmas (Perseus, 1991).
[10] P. A. Davidson, An Introduction to Magnetohydrodynamics (Cambridge Univ. Press, 2001).
[11] B. Auchmann, S. Kurz, and S. Russenschuck, “A note on Faraday paradoxes,” IEEE Trans.
Magnetics 50, doi: 10.1109/TMAG.2013.2285402 (2014).
[12] C. F. Chyba, K. P. Hand, and P. J. Thomas, “Energy conservation and Poynting’s theorem
in the homopolar generator,” Am. J. Phys. 83, 72–75 (2015).
[13] P. J. Scanlon, R. N. Henriksen, and J. R. Allen, “Approaches to electromagnetic induction,”
Am. J. Phys. 37, 698–708 (1969).
[14] E. H. Kennard, “Unipolar induction,” Phil. Mag. 23, 937–941 (1912).

34
[15] E. H. Kennard, “On unipolar induction: another experiment and its significance for the exis-
tence of the aether,” Phil. Mag. 33, 179 (1917).
[16] I. Galili and D. Kaplan, “Changing approach to teaching electromagnetism in a conceptually
oriented introductory physics course,” Am. J. Phys. 65, 657–667 (1997).
[17] H. Poincaré, “Sur l’induction unipolaire,” L’Eclairage Electrique 23, 42–53 (1900).
[18] A. I. Miller, Albert Einstein’s Special Theory of Relativity: Emergence (1905) and Early In-
terpretation (1905-1911) (Springer, 1998).
[19] G. B. Pegram, “Unipolar induction and electron theory,” Phys. Rev. 10, 591 (1917).
[20] L. Davis, “Stellar electromagnetic fields,” Phys. Rev. 72, 632–633 (1947).
[21] H. Alfvén, “On frozen-in field lines and field-line reconnection,” J. Geophys. Res. 81, 4019–
4021 (1976).
[22] P. Lorrain, F. Lorrain, and S. Houle, Magneto-Fluid Dynamics: Fundamentals and Case
Studies of Natural Phenomena (Springer, 2006).
[23] W. D. Parkinson, Introduction to Geomagnetism (Scottish Academic Press, 1983).
[24] E. W. Hones and J. E. Bergeson, “Electric field generated by a rotating magnetized sphere,”
J. Geophys. Res. 70, 4951–4958 (1965).
[25] L. Davis, “D4. Stellar electromagnetic fields,” Phys. Rev. 73, 536 (1948).
[26] H. Alfvén, “On the existence of electromagnetic-hydrodynamic waves,” Arkiv för Matematik,
Astronomi och Fysik 29B, 1–7 (1942).
[27] J. A. Shercliff, A Textbook of Magnetohydrodynamics (Pergamon, 1965).
[28] P. Lorrain, J. McTavish, and F. Lorrain, “Magnetic fields in moving conductors: four simple
examples,” Eur. J. Phys. 19, 451–457 (1998).
[29] J. Van Bladel, Relativity and Engineering (Springer-Verlag, 1984).
[30] G. Backus, “The external electric field of a rotating magnet,” Astrophys. J. 123, 508–512
(1956).
[31] D. V. Redžić, “Conductors moving in magnetic fields: approach to equilibrium,” Eur. J. Phys.
25, 623 (2004).
[32] R. J. Gutmann and J. M. Borrego, “Charge density in a conducting medium revisited,” IEEE
Trans. Ant. Prop. 22, 635 (1974).
[33] D. H. Kobe, “Can electrostatics be described solely by a vector potential?” Am J. Phys. 49,
1075–1076 (1981).

35
[34] M. Kinet, MHD Turbulence at Low Magnetic Reynolds Number: Spectral Properties and Tran-
sition Mechanism in a Square Duct, Ph.D. thesis, Université Libre de Bruxelles (2009).
[35] H. K. Moffatt, Magnetic Field Generation in Electrically Conducting Fluids (Cambridge Univ.
Press, 1978).
[36] J. E. Allen, “A note on the magnetic Reynolds number,” J. Phys. D: Appl. Phys. 19, L133–
L135 (1986).
[37] J. Emsley, The Elements, 3rd ed. (Clarendon, 1998).
[38] E. C. Snelling, Soft Ferrites: Properties and Applications, 2nd ed. (Butterworth, 1988).
[39] T. Rikitake, Magnetic and Electromagnetic Shielding (Kluwer, 1987).
[40] J. Prat-Camps, C. Navau, D.-X. Chen, and A. Sanchez, “Exact analytical demagne-
tizing factors for long hollow cylinders in transverse field,” IEEE Magn. Lett. 3, doi:
10.1109/LMAG.2012.2198617 (2012).
[41] P. H. Roberts, An Introduction to Magnetohydrodynamics (Longmans, 1967).
[42] A. J. Mager, “Das Eindringen von Magnetfeldern in offene Abschirmzylinder,” Angew. Phys.
23, 381–386 (1967).
[43] T. Morisue, “A comparison of the Coulomb gauge and Lorentz gauge magnetic vector potential
formulations for 3D eddy current calculations,” IEEE Trans. Magnetics 29, 1372–1375 (1993).
[44] R. Moreau, Magnetohydrodynamics (Kluwer, 1990).
[45] J. D. Jackson, “From Lorenz to Coulomb and other explicit gauge transformations,” Am. J.
Phys. 70, 917–928 (2002).
[46] A. J. Mager, “Magnetic shields,” IEEE Trans. Magnetics 6, 67–75 (1970).
[47] J.F. Hoburg, “Principles of quasistatic magnetic shielding with cylindrical and spherical
shields,” IEEE Trans. Electromagn. Compat. 37, 574–579 (1995).
[48] S. Celozzi, R. Araneo, and G. Lovat, Electromagnetic Shielding (Wiley, 2008).
[49] F. W. J. Olver, D. W. Lozier, R. F. Boisvert, and C. W. Clark, NIST Handbook of Mathe-
matical Functions (Cambridge Univ. Press, 2010).
[50] D. H. Menzel, Fundamental Formulas of Physics, Vol. 1 (Dover, 1960).
[51] W. A. Newcomb, “Motion of magnetic lines of force,” Ann. Phys. 3, 347–385 (1958).
[52] A. L. Wilmot-Smith, E. R. Priest, and G. Hornig, “Magnetic diffusion and the motion of field
lines,” Geophys. Astrophys. Fluid. Dynam. 99, 177–197 (2005).

36
[53] E. Priest and T. Forbes, Magnetic Reconnection: MHD Theory and Applications (Cambridge
Univ. Press, 2000).
[54] J. H. Poynting, “On the transfer of energy in the electromagnetic field,” Phil. Trans. R. Soc.
Lond. 175, 343–361 (1884).
[55] N. L. Sharma, “Field versus action-at-a-distance in a static situation,” Am. J. Phys. 56,
420–423 (1987).
[56] D. J. Griffiths, “Note on ‘field versus action-at-a-distance in a static situation’,” Am J. Phys.
57, 558 (1989).
[57] C. C. Finlay, S. Maus, C. D. Beggan, T. N. Bondar, and A. Chambodut et al., “International
geomagnetic reference field: the eleventh generation,” Geophys. J. Int. 183, 1216–1230 (2010).
[58] National Magnetics Group, MN60 Material (www.magneticsgroup.com/pdf/MN60%20.pdf
www.magneticsgroup.com/pdf/MN60% 20.pdf).
[59] McLaughlin J. C. and K. L. Kaiser, “‘Deglorifying’ the maximum power transfer theorem and
factors in impedance selection,” IEEE Trans. Edu. 50, 251–255 (2007).
[60] Y. Zhang, N. W. Franklin, R. J. Chen, and J. Dai, “Metal coating on suspended carbon
nanotubes and its implication to metal-tube interaction,” Chem. Phys. Lett. 331, 35 (2000).
[61] S. M. Tripathi, T. S. Bholanath, and S Shantkriti, “Synthesis and study of applications of
metal coated carbon nanotubes,” Int. J. Control Automation 3, 53 (2010).
[62] International Energy Agency, World Energy Outlook 2014 (2014).
[63] K. Lambeck, “Changes in length-of-day and atmospheric circulation,” Nature 286, 104–105
(1980).
[64] K. Lambeck, “Effects of tidal dissipation in the oceans on the moon’s orbit and the earth’s
rotation,” J. Geophys. Res. 80, 2917–2925 (1975).
[65] T. J. Birmingham and F. C. Jones, “Identification of moving magnetic field lines,” J. Geophys.
Res. 73, 5505–5510 (1968).
[66] J. Slepian, “Lines of force in electric and magnetic fields,” Am. J. Phys. 19, 87–90 (1951).
[67] C. Fälthammar, “On the concept of moving magnetic field lines,” Eos Trans. AGU 88, 169
(2007).
[68] J. D. Jackson and L. B. Okun, “Historical roots of gauge invariance,” Rev. Mod. Phys. 73,
663–680 (2001).

37
FIG. 1. A magnetically permeable, low-Rm cylindrical shell with inner and outer radii a and b
and length L is moving at velocity v = vŷ through background fields B∞ = B∞ x̂ and E∞ = 0. A
rectangular current path C with vertices d, e, f, g is embedded in, and translating with, the shell.
An emf is generated around C according to Eq. (63). A digital voltmeter (DVM) measures half
this emf between d and f . C lies in the plane x = b cos φ0 ≡ x0 , with |x0 | ≥ a. It has right-angle
vertices at d = (x0 , y0 , −l/2), e = (x0 , y0 , l/2), f = (x0 , −y0 , l/2) and g = (x0 , −y0 , −l/2), where
y0 = b sin φ0 and l/2 < L/2 − a.
38
 0.02

0.01
∞
(B -B )/B
0x

0
x

-0.01

-0.02
-1.0 -0.5 0.0 0.5 1.0
y (cm)

FIG. 2. Deviations in the x component Bx of the magnetic flux density (Eq. 59) for the moving
cylindrical shell from that for the stationary shell (Eq. 10), relative to the flux density at infinity,
√
for a shell made of MN60 material (see text) with b = 1 cm. These results are for the x0 = b/ 2
√
plane, with (a/b) = 1/ 2.

39
 1

0.5
x (cm)

-0.5

-1
-0.02 -0.01 0.00 0.01 0.02

(B -B )/B
y 0y ∞

FIG. 3. Deviations in the y component By of the magnetic flux density (Eq. 61) for the moving
cylindrical shell from that for the stationary shell (Eq. 10), relative to the flux density at infinity,
√
for a shell made of MN60 material (see text) with b = 1 cm. These results are for the y0 = b/ 2
√
plane, with (a/b) = 1/ 2.

40