CHAPTER

6 Magnetic Fields in Matter

MAGNETIZATION

Diamagnets, Paramagnets, Ferromagnets

If you ask the average person what ―magnetism‖ is, you will probably be told
about refrigerator decorations, compass needles, and the North Pole—none of
which has any obvious connection with moving charges or current-carrying wires.
Yet all magnetic phenomena are due to electric charges in motion, and in fact, if
you could examine a piece of magnetic material on an atomic scale you would
find tiny currents: electrons orbiting around nuclei and spinning about their axes.
For macroscopic purposes, these current loops are so small that we may treat them
as magnetic dipoles. Ordinarily, they cancel each other out because of the random
orientation of the atoms. But when a magnetic field is applied, a net alignment of
these magnetic dipoles occurs, and the medium becomes magnetically polarized,
or magnetized.
Unlike electric polarization, which is almost always in the same direction as E,
some materials acquire a magnetization parallel to B (paramagnets) and some
opposite to B (diamagnets). A few substances (called ferromagnets, in defer-
ence to the most common example, iron) retain their magnetization even after the
external field has been removed—for these, the magnetization is not determined
by the present field but by the whole magnetic ―history‖ of the object. Permanent
magnets made of iron are the most familiar examples of magnetism, but from a
theoretical point of view they are the most complicated; I‘ll save ferromagnetism
for the end of the chapter, and begin with qualitative models of paramagnetism
and diamagnetism.

Torques and Forces on Magnetic Dipoles

A magnetic dipole experiences a torque in a magnetic field, just as an electric
dipole does in an electric field. Let‘s calculate the torque on a rectangular current
loop in a uniform field B. (Since any current loop could be built up from infinites-
imal rectangles, with all the ―internal‖ sides canceling, as indicated in Fig. 6.1,
there is no real loss of generality here; but if you prefer to start from scratch with
an arbitrary shape, see Prob. 6.2.) Center the loop at the origin, and tilt it an angle
θ from the z axis towards the y axis (Fig. 6.2). Let B point in the z direction. The
forces on the two sloping sides cancel (they tend to stretch the loop, but they don‘t

266
6.1 Magnetization 267

FIGURE 6.1

rotate it). The forces on the ―horizontal‖ sides are likewise equal and opposite (so
the net force on the loop is zero), but they do generate a torque:

N = a F sin θ x̂.

The magnitude of the force on each of these segments is

F = Ib B,

and therefore

N = I ab B sin θ x̂ = m B sin θ x̂,

or

N = m × B, (6.1)

where m =Iab is the magnetic dipole moment of the loop. Equation 6.1 gives
the torque on any localized current distribution, in the presence of a uniform field;
in a nonuniform field it is the exact torque (about the center) for a perfect dipole
of infinitesimal size.

z z
m m
B  B

F 

I
 y  y
a F
a b
x

(a) (b)

FIGURE 6.2
6.1 Magnetization 268

Notice that Eq. 6.1 is identical in form to the electrical analog, Eq. 4.4:
N = p E.× In particular, the torque is again in such a direction as to line the
dipole up parallel to the field. It is this torque that accounts for paramagnetism.
Since every electron constitutes a magnetic dipole (picture it, if you wish, as a
tiny spinning sphere of charge), you might expect paramagnetism to be a univer-
sal phenomenon. Actually, quantum mechanics (specifically, the Pauli exclusion
principle) tends to lock the electrons within a given atom together in pairs with
opposing spins,1 and this effectively neutralizes the torque on the combination.
As a result, paramagnetism most often occurs in atoms or molecules with an
odd number of electrons, where the ―extra‖ unpaired member is subject to the
magnetic torque. Even here, the alignment is far from complete, since random
thermal collisions tend to destroy the order.
In a uniform field, the net force on a current loop is zero:
. Σ
F = I (dl × B) = I dl × B = 0;

the constant B comes outside the integral, and the net displacement dl around a
closed loop vanishes. In a nonuniform field this is no longer the case. For example,
suppose a circular wire ring of radius R, carrying a current I , is suspended above
a short solenoid in the ―fringing‖ region (Fig. 6.3). Here B has a radial component,
and there is a net downward force on the loop (Fig. 6.4):

F = 2π I R B cos θ. (6.2)

For an infinitesimal loop, with dipole moment m, in a field B, the force is

F = ∇(m · B) (6.3)

B
I I

B I B

R
F F

FIGURE 6.3 FIGURE 6.4

1This is not always true for the outermost electrons in unfilled shells.
6.1 Magnetization 269

(see Prob. 6.4). Once again the magnetic formula is identical to its electrical
―twin,‖ if we write the latter in the form F= ∇ (p· E). (See footnote to Eq. 4.5.)
If you‘re starting to get a sense of déjà vu, perhaps you will have more respect
for those early physicists who thought magnetic dipoles consisted of positive and
negative magnetic ―charges‖ (north and south ―poles,‖ they called them), sepa-
rated by a small distance, just like electric dipoles (Fig. 6.5(a)). They wrote down
a ―Coulomb‘s law‖ for the attraction and repulsion of these poles, and devel-
oped the whole of magnetostatics in exact analogy to electrostatics. It‘s not a bad
model, for many purposes—it gives the correct field of a dipole (at least, away
from the origin), the right torque on a dipole (at least, on a stationary dipole),
and the proper force on a dipole (at least, in the absence of external currents). But
it‘s bad physics, because there’s no such thing as a single magnetic north pole or
south pole. If you break a bar magnet in half, you don‘t get a north pole in one
hand and a south pole in the other; you get two complete magnets. Magnetism
is not due to magnetic monopoles, but rather to moving electric charges; mag-
netic dipoles are tiny current loops (Fig. 6.5(c)), and it‘s an extraordinary thing,
really, that the formulas involving m bear any resemblance to the corresponding
formulas for p. Sometimes it is easier to think in terms of the ―Gilbert‖ model of
a magnetic dipole (separated monopoles), instead of the physically correct ―Am-
père‖ model (current loop). Indeed, this picture occasionally offers a quick and
clever solution to an otherwise cumbersome problem (you just copy the corre-
sponding result from electrostatics, changing p to m, 1/‹0 to μ0, and E to B). But
whenever the close-up features of the dipole come into play, the two models can
yield strikingly different answers. My advice is to use the Gilbert model, if you
like, to get an intuitive ―feel‖ for a problem, but never rely on it for quantitative
results.

N +
m

m p I

S 
(a) Magnetic dipole (b) Electric dipole (c) Magnetic dipole
(Gilbert model) (Ampère model)

FIGURE 6.5

Problem 6.1 Calculate the torque exerted on the square loop shown in Fig. 6.6, due
to the circular loop (assume r is much larger than a or b). If the square loop is free
to rotate, what will its equilibrium orientation be?
270 Chapter 6 Magnetic Fields in Matter

I I b

a b

FIGURE 6.6

Problem 6.2 Starting from the Lorentz force law, in the form of Eq. 5.16, show that
the torque on any steady current distribution (not just a square loop) in a uniform
field B is m × B.
Problem 6.3 Find the force of attraction between two magnetic dipoles, m1 and
m2, oriented as shown in Fig. 6.7, a distance r apart, (a) using Eq. 6.2, and (b) using
Eq. 6.3.

z
I
c
m1 m2

r c y
x

FIGURE 6.7 FIGURE 6.8

Problem 6.4 Derive Eq. 6.3. [Here‘s one way to do it: Assume the dipole is an in-
finitesimal square, of side ‹ (if it‘s not, chop it up into squares, and apply the argu-
ment to each one). Choose axes as shown in Fig. 6.8, and calculate ¸ F I (dl B)
=
along each of the four sides. Expand B in a Taylor series—on the right side, for in-
×
stance,
∂B
B = B(0, ‹, z) ∼
= B(0, 0, z) + ‹ .∂ y . .
(0,0,z)

For a more sophisticated method, see Prob. 6.22.]

Problem 6.5 A uniform current density J = J0 zˆ fills a slab straddling the yz plane,
from x = −a to x = +a. A magnetic dipole m = m 0 x̂ is situated at the origin.
(a) Find the force on the dipole, using Eq. 6.3.
(b) Do the same for a dipole pointing in the y direction: m = m 0 ŷ.
(c) In the electrostatic case, the expressions F = ∇ (p · E) and F = (p · ∇)E are
equivalent (prove it), but this is not the case for the magnetic analogs (explain
why). As an example, calculate (m · ∇)B for the configurations in (a) and (b).
270 Chapter 6 Magnetic Fields in Matter

Effect of a Magnetic Field on Atomic Orbits

Electrons not only spin; they also revolve around the nucleus—for simplicity, let‘s
assume the orbit is a circle of radius R (Fig. 6.9). Although technically this orbital
motion does not constitute a steady current, in practice the period T=2π R/v is
so short that unless you blink awfully fast, it‘s going to look like a steady current:
−e ev
I= = − 2π R.
T
(The minus sign accounts for the negative charge of the electron.) Accordingly,
the orbital dipole moment (Iπ R 2) is
1
m = − ev R ẑ. (6.4)
2
Like any other magnetic dipole, this one is subject to a torque (m × B) when
you turn on a magnetic field. But it‘s a lot harder to tilt the entire orbit than it is
the spin, so the orbital contribution to paramagnetism is small. There is, however,
a more significant effect on the orbital motion: The electron speeds up or slows
down, depending on the orientation of B. For whereas the centripetal acceleration
v 2 /R is ordinarily sustained by electrical forces alone,2
1 e2 v2
= me , (6.5)
4π‹0 R2 R
in the presence of a magnetic field there is an additional force,−e(v × B). For
the sake of argument, let‘s say that B is perpendicular to the plane of the orbit, as
shown in Fig. 6.10; then
1 e2 v̄ 2
4π‹0 R2 + ev¯ B = me R . (6.6)

Under these conditions, the new speed v¯ is greater than v:

me 2 me
ev̄ B = (v̄ − v 2 ) = (v̄ + v)(v̄ − v),
R R

z
v
R
e

FIGURE 6.9

2To avoid confusion with the magnetic dipole moment m, I‘ll write the electron mass with
subscript: m e .
 6.1 Magnetization 273

B
B B

e
R

e v

FIGURE 6.10

or, assuming the change Ov = v¯ − v is small,

eRB
Ov = . (6.7)
2me
When B is turned on, then, the electron speeds up.3
A change in orbital speed means a change in the dipole moment (Eq. 6.4):
1 e2 R 2
Om = − e(Ov)R zˆ = − B. (6.8)
2 4me
Notice that the change in m is opposite to the direction of B. (An electron circling
the other way would have a dipole moment pointing upward, but such an orbit
would be slowed down by the field, so the change is still opposite to B.) Ordi-
narily, the electron orbits are randomly oriented, and the orbital dipole moments
cancel out. But in the presence of a magnetic field, each atom picks up a little
―extra‖ dipole moment, and these increments are all antiparallel to the field. This
is the mechanism responsible for diamagnetism. It is a universal phenomenon,
affecting all atoms. However, it is typically much weaker than paramagnetism,
and is therefore observed mainly in atoms with even numbers of electrons, where
paramagnetism is usually absent.
In deriving Eq. 6.8, I assumed that the orbit remains circular, with its original
radius R. I cannot offer a justification for this at the present stage. If the atom
is stationary while the field is turned on, then my assumption can be proved—
this is not magnetostatics, however, and the details will have to await Chapter 7
(see Prob. 7.52). If the atom is moved into the field, the situation is enormously
more complicated. But never mind—I‘m only trying to give you a qualitative
account of diamagnetism. Assume, if you prefer, that the velocity remains the
same while the radius changes—the formula (Eq. 6.8) is altered (by a factor of 2),
but the qualitative conclusion is unaffected. The truth is that this classical model is
fundamentally flawed (diamagnetism is really a quantum phenomenon), so there‘s

3I said (Eq. 5.11) that magnetic fields do no work, and are incapable of speeding a particle up. I stand
by that. However, as we shall see in Chapter 7, a changing magnetic field induces an electric field, and
it is the latter that accelerates the electrons in this instance.
 6.1 Magnetization 273

not much point in refining the details.4 What is important is the empirical fact
that in diamagnetic materials the induced dipole moments point opposite to the
magnetic field.

Magnetization
In the presence of a magnetic field, matter becomes magnetized; that is, upon
microscopic examination, it will be found to contain many tiny dipoles, with a net
alignment along some direction. We have discussed two mechanisms that account
for this magnetic polarization: (1) paramagnetism (the dipoles associated with the
spins of unpaired electrons experience a torque tending to line them up parallel
to the field) and (2) diamagnetism (the orbital speed of the electrons is altered in
such a way as to change the orbital dipole moment in a direction opposite to the
field). Whatever the cause, we describe the state of magnetic polarization by the
vector quantity

M ≡ magnetic dipole moment per unit volume. (6.9)

M is called the magnetization; it plays a role analogous to the polarization
P in electrostatics. In the following section, we will not worry about how the
magnetization got there—it could be paramagnetism, diamagnetism, or even
ferromagnetism—we shall take M as given, and calculate the field this magneti-
zation itself produces.
Incidentally, it may have surprised you to learn that materials other than the
famous ferromagnetic trio (iron, nickel, and cobalt) are affected by a magnetic
field at all. You cannot, of course, pick up a piece of wood or aluminum with a
magnet. The reason is that diamagnetism and paramagnetism are extremely weak:
It takes a delicate experiment and a powerful magnet to detect them at all. If you
were to suspend a piece of paramagnetic material above a solenoid, as in Fig. 6.3,
the induced magnetization would be upward, and hence the force downward. By
contrast, the magnetization of a diamagnetic object would be downward and the
force upward. In general, when a sample is placed in a region of nonuniform field,
the paramagnet is attracted into the field, whereas the diamagnet is repelled away.
But the actual forces are pitifully weak—in a typical experimental arrangement
the force on a comparable sample of iron would be 10 4 or 105 times as great.
That‘s why it was reasonable for us to calculate the field inside a piece of copper
wire, say, in Chapter 5, without worrying about the effects of magnetization.5

4S. L. O‘Dell and R. K. P. Zia, Am. J. Phys. 54, 32, (1986); R. Peierls, Surprises in Theoretical Physics,

Section 4.3 (Princeton, N.J.: Princeton University Press, 1979); R. P. Feynman, R. B. Leighton, and
M. Sands, The Feynman Lectures on Physics, Vol. 2, Sec. 34–36 (New York: Addison-Wesley, 1966).
5 In 1997 Andre Geim managed to levitate a live frog (diamagnetic) for 30 minutes; he was awarded the

2000 Ig Nobel prize for this achievement, and later (2010) the Nobel prize for research on graphene.
See M. V. Berry and A. K. Geim, Eur. J. Phys. 18, 307 (1997) and Geim, Physics Today, September
1998, p. 36.
 6.1 Magnetization 273

Problem 6.6 Of the following materials, which would you expect to be paramag-
netic and which diamagnetic: aluminum, copper, copper chloride (CuCl2), carbon,
lead, nitrogen (N2), salt (NaCl), sodium, sulfur, water? (Actually, copper is slightly
diamagnetic; otherwise they‘re all what you‘d expect.)

THE FIELD OF A MAGNETIZED OBJECT

Bound Currents
Suppose we have a piece of magnetized material; the magnetic dipole moment per
unit volume, M, is given. What field does this object produce? Well, the vector
potential of a single dipole m is given by Eq. 5.85:

A(r) = μ0 m × rˆ. (6.10)

4π r2
In the magnetized object, each volume element dτ J carries a dipole moment
M dτ J , so the total vector potential is (Fig. 6.11)
∫
μ0 M(rJ ) × r̂ J
A(r) = dτ . (6.11)
4π r2
That does it, in principle. But, as in the electrical case (Sect. 4.2.1), the integral
can be cast in a more illuminating form by exploiting the identity
1 rˆ
∇J = 2 .
r r
With this,
∫ Σ . ΣΣ
μ0 1
A(r) = M(r ) × ∇
J J
dτ
4π r
.
Integrating by parts, Jusing product rule 7, gives
.∫ ∫ Σ Σ Σ
μ0 1 J M(rJ )
A(r) = [∇ × M(rJ )] dτ J − ∇ J × dτ J .
4π r r

m
d

FIGURE 6.11
6.2 The Field of a Magnetized Object 275

Problem 1.61(b) invites us to express the latter as a surface integral,

∫
μ0 1 J μ0 1
A(r) = [∇ × M(rJ )] dτ J + [M(rJ ) × daJ ]. (6.12)
4π r 4π r
The first term looks just like the potential of a volume current,

Jb = ∇ × M, (6.13)

while the second looks like the potential of a surface current,

Kb = M × n̂, (6.14)

where n̂ is the normal unit vector. With these definitions,

∫ J (rJ ) μ0 Kb (rJ )
μ0 b
dτ J + da J .
A(r) = 4π r 4π S r
(6.15)
V

What this means is that the potential (and hence also the field) of a magnetized
object is the same as would be produced by a volume current Jb = ∇ × M
throughout the material, plus a surface current K= b M n, on the boundary.
Instead of integrating the contributions of all the ˆ× infinitesimal dipoles, using
Eq. 6.11, we first determine the bound currents, and then find the field they
produce, in the same way we would calculate the field of any other volume and
surface currents. Notice the striking parallel with the electrical case: there the field
of a polarized object was the same as that of a bound volume charge ρb = −∇ · P
plus a bound surface charge σb = P · n̂.
Example 6.1. Find the magnetic field of a uniformly magnetized sphere.
Solution
Choosing the z axis along the direction of M (Fig. 6.12), we have

Jb = ∇ × M = 0, Kb = M × n̂ = M sin θ φ̂.

r

M
y


FIGURE 6.12
6.2 The Field of a Magnetized Object 276

Now, a rotating spherical shell, of uniform surface charge σ , corresponds to a

surface current density

K = σ v = σ ω R sin θ φ̂.
It follows, therefore, that the field of a uniformly magnetized sphere is identi-
cal to the field of a spinning spherical shell, with the identification σ Rω
→ M.
Referring back to Ex. 5.11, I conclude that
2
B=μ 0M, (6.16)
3
inside the sphere, while the field outside is the same as that of a perfect dipole,
4
m = π R3M.
3
Notice that the internal field is uniform, like the electric field inside a uniformly
polarized sphere (Eq. 4.14), although the actual formulas for the two cases are
. Σ
curiously different 32 in place of − 13 .6 The external fields are also analogous:

pure dipole in both instances.

Problem 6.7 An infinitely long circular cylinder carries a uniform magnetization

M parallel to its axis. Find the magnetic field (due to M) inside and outside the
cylinder.
Problem 6.8 A long circular cylinder of radius R carries a magnetization M =
ks 2 φ̂, where k is a constant, s is the distance from the axis, and φ̂ is the usual
azimuthal unit vector (Fig. 6.13). Find the magnetic field due to M, for points inside
and outside the cylinder.

z
R

a
a
y M
 w
x s

FIGURE 6.13 FIGURE 6.14

is no accident that the same factors appear in the ―contact‖ term for the fields of electric and
6 It

magnetic dipoles (Eqs. 3.106 and 5.94). In fact, one good way to model a perfect dipole is to take the
limit (R → 0) of a polarized/magnetized sphere.
6.2 The Field of a Magnetized Object 277

Problem 6.9 A short circular cylinder of radius a and length L carries a ―frozen-in‖
uniform magnetization M parallel to its axis. Find the bound current, and sketch the
magnetic field of the cylinder. (Make three sketches: one for L a, one for L a,
and one for L ≈ a.) Compare this bar magnet with the bar electret of Prob. 4.11.
Problem 6.10 An iron rod of length L and square cross section (side a) is given
a uniform longitudinal magnetization M, and then bent around into a circle with a
narrow gap (width w), as shown in Fig. 6.14. Find the magnetic field at the center
of the gap, assuming w a L. [Hint: treat it as the superposition of a complete
torus plus a square loop with reversed current.]

Physical Interpretation of Bound Currents

In the last section, we found that the field of a magnetized object is identical to the
field that would be produced by a certain distribution of ―bound‖ currents, Jb and
Kb. I want to show you how these bound currents arise physically. This will be a
heuristic argument—the rigorous derivation has already been given. Figure 6.15
depicts a thin slab of uniformly magnetized material, with the dipoles represented
by tiny current loops. Notice that all the ―internal‖ currents cancel: every time
there is one going to the right, a contiguous one is going to the left. However, at
the edge there is no adjacent loop to do the canceling. The whole thing, then, is
equivalent to a single ribbon of current I flowing around the boundary (Fig. 6.16).
What is this current, in terms of M? Say that each of the tiny loops has area a
and thickness t (Fig. 6.17). In terms of the magnetization M , its dipole moment

M
II
I I
I
I

FIGURE 6.15

I
M

n
I

FIGURE 6.16
6.2 The Field of a Magnetized Object 278

a t
I

FIGURE 6.17

is m = Mat . In terms of the circulating current I , however, m = Ia. Therefore

I =Mt , so the surface current is Kb= I/t = M . Using the outward-drawn unit
vector nˆ (Fig. 6.16), the direction of Kb is conveniently indicated by the cross
product:
Kb = M × n̂.
(This expression also records the fact that there is no current on the top or bottom
ˆ so the cross product vanishes.)
surface of the slab; here M is parallel to n,
This bound surface current is exactly what we obtained in Sect. 6.2.1. It is a
peculiar kind of current, in the sense that no single charge makes the whole trip—
on the contrary, each charge moves only in a tiny little loop within a single atom.
Nevertheless, the net effect is a macroscopic current flowing over the surface of
the magnetized object. We call it a ―bound‖ current to remind ourselves that every
charge is attached to a particular atom, but it‘s a perfectly genuine current, and it
produces a magnetic field in the same way any other current does.
When the magnetization is nonuniform, the internal currents no longer cancel.
Figure 6.18(a) shows two adjacent chunks of magnetized material, with a larger
arrow on the one to the right suggesting greater magnetization at that point. On
the surface where they join, there is a net current in the x direction, given by
∂ Mz
Ix = [M z(y + dy) − M z(y)] dz = ∂y dy dz.

The corresponding volume current density is therefore

∂ Mz
( Jb) x = .
∂y

z z
Mz( y + dy)

Mz( y) My( z + dz)

dz
dz
My( z)
dy dy
y y
x (a) x (b)

FIGURE 6.18
280 Chapter 6 Magnetic Fields in Matter

By the same token, a nonuniform magnetization in the y direction would con-

tribute an amount −∂ M y /∂ z (Fig. 6.18(b)), so
∂ Mz ∂ My
( Jb)x = ∂y − .
∂z
In general, then,
Jb = ∇ × M,
consistent, again, with the result of Sect. 6.2.1.
Incidentally, like any other steady current, Jb should obey the conservation law
5.33:
∇ · Jb = 0.
Does it? Yes, for the divergence of a curl is always zero.

The Magnetic Field Inside Matter

Like the electric field, the actual microscopic magnetic field inside matter fluc-
tuates wildly from point to point and instant to instant. When we speak of ―the‖
magnetic field in matter, we mean the macroscopic field: the average over regions
large enough to contain many atoms. (The magnetization M is ―smoothed out‖ in
the same sense.) It is this macroscopic field that one obtains when the methods of
Sect. 6.2.1 are applied to points inside magnetized material, as you can prove for
yourself in the following problem.

Problem 6.11 In Sect, 6.2.1, we began with the potential of a perfect dipole
(Eq. 6.10), whereas in fact we are dealing with physical dipoles. Show, by the
method of Sect. 4.2.3, that we nonetheless get the correct macroscopic field.

THE AUXILIARY FIELD H

Ampère’s Law in Magnetized Materials

In Sect. 6.2, we found that the effect of magnetization is to establish bound cur-
rents Jb = M within the material and Kb M n on the surface. The field
due to magnetization
ˆ∇× of the medium is just the field
= produced
× by these bound cur-
rents. We are now ready to put everything together: the field attributable to bound
currents, plus the field due to everything else—which I shall call the free current.
The free current might flow through wires imbedded in the magnetized substance
or, if the latter is a conductor, through the material itself. In any event, the total
current can be written as

J = Jb + J f . (6.17)
There is no new physics in Eq. 6.17; it is simply a convenience to separate the
current into these two parts, because they got there by quite different means: the
280 Chapter 6 Magnetic Fields in Matter

free current is there because somebody hooked up a wire to a battery—it involves

actual transport of charge; the bound current is there because of magnetization—it
results from the conspiracy of many aligned atomic dipoles.
In view of Eqs. 6.13 and 6.17, Ampère‘s law can be written
1
(∇ × B) = J = J f + Jb = J f + (∇ × M),
μ0
or, collecting together the two curls:
. Σ
1
∇ × B − M = fJ .
μ0
The quantity in parentheses is designated by the letter H:

1
H≡ B − M. (6.18)
μ0

In terms of H, then, Ampère‘s law reads

∇ ×H=Jf, (6.19)

or, in integral form,

H · dl = I fenc , (6.20)

where I fenc is the total free current passing through the Amperian loop.
H plays a role in magnetostatics analogous to D in electrostatics: Just as D
allowed us to write Gauss’s law in terms of the free charge alone, H permits us to
express Ampère’s law in terms of the free current alone—and free current is what
we control directly. Bound current, like bound charge, comes along for the ride—
the material gets magnetized, and this results in bound currents; we cannot turn
them on or off independently, as we can free currents. In applying Eq. 6.20, all
we need to worry about is the free current, which we know about because we put
it there. In particular, when symmetry permits, we can calculate H immediately
from Eq. 6.20 by the usual Ampère‘s law methods. (For example, Probs. 6.7 and
6.8 can be done in one line by noting that H = 0.)
Example 6.2. A long copper rod of radius R carries a uniformly distributed
(free) current I (Fig. 6.19). Find H inside and outside the rod.
Solution
Copper is weakly diamagnetic, so the dipoles will line up opposite to the field.
This results in a bound current running antiparallel to I , within the wire, and
parallel to I along the surface (Fig. 6.20). Just how great these bound currents will
282 Chapter 6 Magnetic Fields in Matter

B
M
H

Amperian loop I
s Jb Kb

FIGURE 6.19 FIGURE 6.20

be we are not yet in a position to say—but in order to calculate H, it is sufficient

to realize that all the currents are longitudinal, so B, M, and therefore also H, are
circumferential. Applying Eq. 6.20 to an Amperian loop of radius s < R,
πs2
H(2πs) = I fenc = I ,
π R2
so, inside the wire,
I
H
= s φˆ (s ≤ R). (6.21)
2π R2
Outside the wire
I
H
= φˆ (s ≥ R). (6.22)
2πs
In the latter region (as always, in empty space) M = 0, so
μ0 I ˆ
B = μ0 H = φ (s ≥ R),
2πs
the same as for a nonmagnetized wire (Ex. 5.7). Inside the wire B cannot be
determined at this stage, since we have no way of knowing M (though in practice
the magnetization in copper is so slight that for most purposes we can ignore it
altogether).

As it turns out, H is a more useful quantity than D. In the laboratory, you

will frequently hear people talking about H (more often even than B), but you
will never hear anyone speak of D (only E). The reason is this: To build an
282 Chapter 6 Magnetic Fields in Matter

electromagnet you run a certain (free) current through a coil. The current is the
thing you read on the dial, and this determines H (or at any rate, the line in-
tegral of H); B depends on the specific materials you used and even, if iron is
present, on the history of your magnet. On the other hand, if you want to set up
an electric field, you do not plaster a known free charge on the plates of a par-
allel plate capacitor; rather, you connect them to a battery of known voltage. It‘s
the potential difference you read on your dial, and that determines E (or rather,
the line integral of E); D depends on the details of the dielectric you‘re using. If
it were easy to measure charge, and hard to measure potential, then you‘d find
experimentalists talking about D instead of E. So the relative familiarity of H,
as contrasted with D, derives from purely practical considerations; theoretically,
they‘re on an equal footing.
Many authors call H, not B, the ―magnetic field.‖ Then they have to invent a
new word for B: the ―flux density,‖ or magnetic ―induction‖ (an absurd choice,
since that term already has at least two other meanings in electrodynamics). Any-
way, B is indisputably the fundamental quantity, so I shall continue to call it the
―magnetic field,‖ as everyone does in the spoken language. H has no sensible
name: just call it ―H.‖7

Problem 6.12 An infinitely long cylinder, of radius R, carries a ―frozen-in‖ magne-

tization, parallel to the axis,
M = ks ẑ,
where k is a constant and s is the distance from the axis; there is no free current
anywhere. Find the magnetic field inside and outside the cylinder by two different
methods:

(a) As in Sect. 6.2, locate all the bound currents, and calculate the field they
produce.
(b) Use Ampère‘s law (in the form of Eq. 6.20) to find H, and then get B from
Eq. 6.18. (Notice that the second method is much faster, and avoids any explicit
reference to the bound currents.)

Problem 6.13 Suppose the field inside a large piece of magnetic material is B0, so
that H0 = (1/μ0)B0 − M, where M is a ―frozen-in‖ magnetization.
(a) Now a small spherical cavity is hollowed out of the material (Fig. 6.21). Find
the field at the center of the cavity, in terms of B0 and M. Also find H at the
center of the cavity, in terms of H0 and M.
(b) Do the same for a long needle-shaped cavity running parallel to M.
(c) Do the same for a thin wafer-shaped cavity perpendicular to M.

7For those who disagree, I quote A. Sommerfeld‘s Electrodynamics (New York: Academic Press,
1952), p. 45: ―The unhappy term ‗magnetic field‘ for H should be avoided as far as possible. It seems
to us that this term has led into error none less than Maxwell himself . . . ‖
284 Chapter 6 Magnetic Fields in Matter

(a) Sphere (b) Needle (c) Wafer

FIGURE 6.21

Assume the cavities are small enough so M, B0, and H0 are essentially constant.
Compare Prob. 4.16. [Hint: Carving out a cavity is the same as superimposing an
object of the same shape but opposite magnetization.]

A Deceptive Parallel
Equation 6.19 looks just like Ampère‘s original law (Eq. 5.56), except that the
total current is replaced by the free current, and B is replaced by μ0H. As in the
case of D, however, I must warn you against reading too much into this corre-
spondence. It does not say that μ0H is ―just like B, only its source is J f instead of
J.‖ For the curl alone does not determine a vector field—you must also know the
divergence. And whereas ∇ ·B =0, the divergence of H is not, in general, zero.
In fact, from Eq. 6.18
∇ · H = −∇ · M. (6.23)
Only when the divergence of M vanishes is the parallel between B and μ0H
faithful.
If you think I‘m being pedantic, consider the example of the bar magnet—a
short cylinder of iron that carries a permanent uniform magnetization M parallel
to its axis. (See Probs. 6.9 and 6.14.) In this case there is no free current any-
where, and a naïve application of Eq. 6.20 might lead you to suppose that H = 0,
and hence that B =μ0M inside the magnet and B =0 outside, which is non-
sense. It is quite true that the curl of H vanishes everywhere, but the divergence
does not. (Can you see where ∇ ·M /= 0?) Advice: When you are asked to find
B or H in a problem involving magnetic materials, first look for symmetry. If the
problem exhibits cylindrical, plane, solenoidal, or toroidal symmetry, then you
can get H directly from Eq. 6.20 by the usual Ampère‘s law methods. (Evidently,
in such cases∇M· is automatically zero, since the free current alone determines
the answer.) If the requisite symmetry is absent, you‘ll have to think of another
284 Chapter 6 Magnetic Fields in Matter

approach, and in particular you must not assume that H is zero just because there
is no free current in sight.

Boundary Conditions
The magnetostatic boundary conditions of Sect. 5.4.2 can be rewritten in terms of
H and the free current. From Eq. 6.23 it follows that

Ha⊥bove − Hb⊥elow = −(Ma⊥bove − Mb⊥elow ), (6.24)

while Eq. 6.19 says
Hǁabove − Hǁbelow = K f × n̂. (6.25)
In the presence of materials, these are sometimes more useful than the correspond-
ing boundary conditions on B (Eqs. 5.74 and 5.76):

Ba⊥bove − Bb⊥elow = 0, (6.26)

and
Bǁabove − Bǁbelow = μ0 (K × n̂). (6.27)
You might want to check them, for Ex. 6.2 or Prob. 6.14.

Problem 6.14 For the bar magnet of Prob. 6.9, make careful sketches of M, B, and
H, assuming L is about 2a. Compare Prob. 4.17.
Problem 6.15 If J f = 0 everywhere, the curl of H vanishes (Eq. 6.19), and we can
express H as the gradient of a scalar potential W :
H = −∇W.
According to Eq. 6.23, then,
∇2 W = (∇ · M),
so W obeys Poisson‘s equation, with∇M · as the ―source.‖ This opens up all the
machinery of Chapter 3. As an example, find the field inside a uniformly magne-
∇ · M = 0 everywhere ex-
tized sphere (Ex. 6.1) by separation of variables. [Hint:
cept at the surface (r=R), so W satisfies Laplace‘s equation in the regions r < R
and r > R; use Eq. 3.65, and from Eq. 6.24 figure out the appropriate boundary
condition on W .]

LINEAR AND NONLINEAR MEDIA

Magnetic Susceptibility and Permeability

In paramagnetic and diamagnetic materials, the magnetization is sustained by the
field; when B is removed, M disappears. In fact, for most substances the mag-
netization is proportional to the field, provided the field is not too strong. For
6.4 Linear and Nonlinear Media 285

notational consistency with the electrical case (Eq. 4.30), I should express the
proportionality thus:
1
M = χ m B (incorrect!). (6.28)
μ0
But custom dictates that it be written in terms of H, instead of B:

M = χm H. (6.29)

The constant of proportionality χm is called the magnetic susceptibility; it is a

dimensionless quantity that varies from one substance to another—positive for
paramagnets and negative for diamagnets. Typical values are around 10 −5 (see
Table 6.1).
Materials that obey Eq. 6.29 are called linear media. In view of Eq. 6.18,
B = μ0(H + M) = μ0(1 + χm )H, (6.30)
for linear media. Thus B is also proportional to H:8
B = μH, (6.31)
where
μ ≡ μ0(1 + χm ). (6.32)
μ is called the permeability of the material.9 In a vacuum, where there is no
matter to magnetize, the susceptibility χm vanishes, and the permeability is μ0.
That‘s why μ0 is called the permeability of free space.

Material Susceptibility Material Susceptibility

Diamagnetic: Paramagnetic:
Bismuth −1.7 × 10−4 Oxygen (O2) 1.7 × 10−6
Gold −3.4 × 10−5 Sodium 8.5 × 10−6
Silver −2.4 × 10−5 Aluminum 2.2 × 10−5
Copper −9.7 × 10−6 Tungsten 7.0 × 10−5
Water −9.0 × 10−6 Platinum 2.7 × 10−4
Carbon Dioxide −1.1 × 10−8 Liquid Oxygen 3.9 × 10−3
(−200◦ C)
−9 −1
Hydrogen (H2) −2.1 × 10 Gadolinium 4.8 × 10
TABLE 6.1 Magnetic Susceptibilities (unless otherwise specified, values are for 1 atm,
20◦ C). Data from Handbook of Chemistry and Physics, 91st ed. (Boca Raton: CRC Press,
Inc., 2010) and other references.

8Physically, therefore, Eq. 6.28 would say exactly the same as Eq. 6.29, only the constant χm would
have a different value. Equation 6.29 is a little more convenient, because experimentalists find it
handier to work with H than B.
9If you factor out μ , what‘s left is called the relative permeability: μ 1 χ μ/μ . By the
0 r ≡m + 0 =
way, formulas for H in terms of B (Eq. 6.31, in the case of linear media) are called constitutive
relations, just like those for D in terms of E.
 6.4 Linear and Nonlinear Media 286

Example 6.3. An infinite solenoid (n turns per unit length, current I ) is filled
with linear material of susceptibility χm . Find the magnetic field inside the
solenoid.



FIGURE 6.22

Solution
Since B is due in part to bound currents (which we don‘t yet know), we cannot
compute it directly. However, this is one of those symmetrical cases in which we
can get H from the free current alone, using Ampère‘s law in the form of Eq. 6.20:
H = nI zˆ
(Fig. 6.22). According to Eq. 6.31, then,
B = μ0 (1 + χm )n I ẑ.
If the medium is paramagnetic, the field is slightly enhanced; if it‘s diamagnetic,
the field is somewhat reduced. This reflects the fact that the bound surface current
Kb = M × n̂ = χm (H × n̂) = χm n I φ̂
is in the same direction as I , in the former case (χm > 0), and opposite in the
latter (χm < 0).

You might suppose that linear media escape the defect in the parallel between
B and H: since M and H are now proportional to B, does it not follow that
their divergence, like B‘s, must always vanish? Unfortunately, it does not;10 at
the boundary between two materials of different permeability, the divergence of
M can actually be infinite. For instance, at the end of a cylinder of linear para-
magnetic material, M is zero on one side but not on the other. For the ―Gaussian
pillbox‖ shown in Fig. 6.23, M · da /= 0, and hence, by the divergence theorem,
∇ · M cannot vanish everywhere within it.
. Σ . Σ . Σ
10Formally, ∇ · H = ∇ ·
μ
1B = μ1 ∇ · B + B · ∇ μ1 = B · ∇ μ1 , so H is not divergenceless (in
general) at points where μ is changing.
 6.4 Linear and Nonlinear Media 287

Gaussian pillbox
Paramagnet
M=0
Vacuum M

FIGURE 6.23

Incidentally, the volume bound current density in a homogeneous linear mate-

rial is proportional to the free current density:
Jb = ∇ × M = ∇ × (χm H) = χm J f . (6.33)
In particular, unless free current actually flows through the material, all bound
current will be at the surface.

Problem 6.16 A coaxial cable consists of two very long cylindrical tubes, separated
by linear insulating material of magnetic susceptibility χm . A current I flows down
the inner conductor and returns along the outer one; in each case, the current dis-
tributes itself uniformly over the surface (Fig. 6.24). Find the magnetic field in the
region between the tubes. As a check, calculate the magnetization and the bound
currents, and confirm that (together, of course, with the free currents) they generate
the correct field.

I b
I a

FIGURE 6.24

Problem 6.17 A current I flows down a long straight wire of radius a. If the wire
is made of linear material (copper, say, or aluminum) with susceptibility χm , and
the current is distributed uniformly, what is the magnetic field a distance s from the
axis? Find all the bound currents. What is the net bound current flowing down the
wire?

! Problem 6.18 A sphere of linear magnetic material is placed in an otherwise uni-

form magnetic field B0. Find the new field inside the sphere. [Hint: See Prob. 6.15
or Prob. 4.23.]

Problem 6.19 On the basis of the naïve model presented in Sect. 6.1.3, estimate
the magnetic susceptibility of a diamagnetic metal such as copper. Compare your
answer with the empirical value in Table 6.1, and comment on any discrepancy.
6.4 Linear and Nonlinear Media 288

Ferromagnetism
In a linear medium, the alignment of atomic dipoles is maintained by a magnetic
field imposed from the outside. Ferromagnets—which are emphatically not lin-
ear11—require no external fields to sustain the magnetization; the alignment is
―frozen in.‖ Like paramagnetism, ferromagnetism involves the magnetic dipoles
associated with the spins of unpaired electrons. The new feature, which makes fer-
romagnetism so different from paramagnetism, is the interaction between nearby
dipoles: In a ferromagnet, each dipole “likes” to point in the same direction as its
neighbors. The reason for this preference is essentially quantum mechanical, and
I shall not endeavor to explain it here; it is enough to know that the correlation is
so strong as to align virtually 100% of the unpaired electron spins. If you could
somehow magnify a piece of iron and ―see‖ the individual dipoles as tiny arrows,
it would look something like Fig. 6.25, with all the spins pointing the same way.
But if that is true, why isn‘t every wrench and nail a powerful magnet? The
answer is that the alignment occurs in relatively small patches, called domains.
Each domain contains billions of dipoles, all lined up (these domains are actually
visible under a microscope, using suitable etching techniques—see Fig. 6.26), but
the domains themselves are randomly oriented. The household wrench contains an
enormous number of domains, and their magnetic fields cancel, so the wrench as
a whole is not magnetized. (Actually, the orientation of domains is not completely
random; within a given crystal, there may be some preferential alignment along
the crystal axes. But there will be just as many domains pointing one way as
the other, so there is still no large-scale magnetization. Moreover, the crystals
themselves are randomly oriented within any sizable chunk of metal.)
How, then, would you produce a permanent magnet, such as they sell in
toy stores? If you put a piece of iron into a strong magnetic field, the torque
Nm = B tends
× to align the dipoles parallel to the field. Since they like to stay
parallel to their neighbors, most of the dipoles will resist this torque. However,

FIGURE 6.25

11In this sense, it is misleading to speak of the susceptibility or permeability of a ferromagnet. The
terms are used for such materials, but they refer to the proportionality factor between a differential
increase in H and the resulting differential change in M (or B); moreover, they are not constants, but
functions of H.
6.4 Linear and Nonlinear Media 289

Ferromagnetic domains. (Photo courtesy of R. W. DeBlois)

FIGURE 6.26

at the boundary between two domains, there are competing neighbors, and the
torque will throw its weight on the side of the domain most nearly parallel to the
field; this domain will win some converts, at the expense of the less favorably ori-
ented one. The net effect of the magnetic field, then, is to move the domain bound-
aries. Domains parallel to the field grow, and the others shrink. If the field is strong
enough, one domain takes over entirely, and the iron is said to be saturated.
It turns out that this process (the shifting of domain boundaries in response to
an external field) is not entirely reversible: When the field is switched off, there
will be some return to randomly oriented domains, but it is far from complete—
there remains a preponderance of domains in the original direction. You now have
a permanent magnet.
A simple way to accomplish this, in practice, is to wrap a coil of wire around
the object to be magnetized (Fig. 6.27). Run a current I through the coil; this pro-
vides the external magnetic field (pointing to the left in the diagram). As you in-
crease the current, the field increases, the domain boundaries move, and the mag-
netization grows. Eventually, you reach the saturation point, with all the dipoles
aligned, and a further increase in current has no effect on M (Fig. 6.28, point b).
Now suppose you reduce the current. Instead of retracing the path back to
M = 0, there is only a partial return to randomly oriented domains; M decreases,
but even with the current off there is some residual magnetization (point c). The
wrench is now a permanent magnet. If you want to eliminate the remaining mag-
netization, you‘ll have to run a current backwards through the coil (a negative I ).
Now the external field points to the right, and as you increase I (negatively),
290 Chapter 6 Magnetic Fields in Matter

FIGURE 6.27

M drops down to zero (point d). If you turn I still higher, you soon reach sat-
uration in the other direction—all the dipoles now pointing to the right (e). At
this stage, switching off the current will leave the wrench with a permanent mag-
netization to the right (point f ). To complete the story, turn I on again in the
positive sense: M returns to zero (point g), and eventually to the forward satura-
tion point (b).
The path we have traced out is called a hysteresis loop. Notice that the mag-
netization of the wrench depends not only on the applied field (that is, on I ), but
also on its previous magnetic ―history.‖12 For instance, at three different times in
our experiment the current was zero (a, c, and f ), yet the magnetization was dif-
ferent for each of them. Actually, it is customary to draw hysteresis loops as plots
of B against H , rather than M against I . (If our coil is approximated by a long
solenoid, with n turns per unit length, then H=nI , so H and I are proportional.
Meanwhile, B = μ0(H M), + but in practice M is huge compared to H , so to all
intents and purposes B is proportional to M.)
To make the units consistent (teslas), I have plotted (μ0 H ) horizontally
(Fig. 6.29); notice, however, that the vertical scale is 104 times greater than the
horizontal one. Roughly speaking, μ0H is the field our coil would have produced
in the absence of any iron; B is what we actually got, and compared to μ0H, it is
gigantic. A little current goes a long way, when you have ferromagnetic materials

(Permanent (Saturation)
c b
Magnet)

d a g I

e f (Permanent
(Saturation) Magnet)

FIGURE 6.28
12Etymologically, the word hysteresis has nothing to do with the word history—nor with the word
hysteria. It derives from a Greek verb meaning ―lag behind.‖
290 Chapter 6 Magnetic Fields in Matter

B
1.5
1

0.5

15 10 5 5 10 15 20 (0H 10 4)

0.5
1
1.5

FIGURE 6.29

around. That‘s why anyone who wants to make a powerful electromagnet will
wrap the coil around an iron core. It doesn‘t take much of an external field to
move the domain boundaries, and when you do that, you have all the dipoles in
the iron working with you.
One final point about ferromagnetism: It all follows, remember, from the
fact that the dipoles within a given domain line up parallel to one another. Ran-
dom thermal motions compete with this ordering, but as long as the temperature
doesn‘t get too high, they cannot budge the dipoles out of line. It‘s not surprising,
though, that very high temperatures do destroy the alignment. What is surprising
is that this occurs at a precise temperature (770 ◦ C, for iron). Below this temper-
ature (called the Curie point), iron is ferromagnetic; above, it is paramagnetic.
The Curie point is rather like the boiling point or the freezing point in that there is
no gradual transition from ferro- to para-magnetic behavior, any more than there
is between water and ice. These abrupt changes in the properties of a substance,
occurring at sharply defined temperatures, are known in statistical mechanics as
phase transitions.

Problem 6.20 How would you go about demagnetizing a permanent magnet (such
as the wrench we have been discussing, at point c in the hysteresis loop)? That is,
how could you restore it to its original state, with M = 0 at I = 0?

Problem 6.21

(a) Show that the energy of a magnetic dipole in a magnetic field B is

U = −m · B. (6.34)

[Assume that the magnitude of the dipole moment is fixed, and all you have
to do is move it into place and rotate it into its final orientation. The energy re-
quired to keep the current flowing is a different problem, which we will confront
in Chapter 7.] Compare Eq. 4.6.
 6.4 Linear and Nonlinear Media 292

m1 m2
1 r 2

FIGURE 6.30

(b) Show that the interaction energy of two magnetic dipoles separated by a dis-
placement r is given by
μ0 1
U = [m · m2 − 3(m1 · r̂)(m2 · r̂)]. (6.35)
4π r 3 1
Compare Eq. 4.7.

(c) Express your answer to (b) in terms of the angles θ1 and θ2 in Fig. 6.30, and use
the result to find the stable configuration two dipoles would adopt if held a fixed
distance apart, but left free to rotate.

(d) Suppose you had a large collection of compass needles, mounted on pins at
regular intervals along a straight line. How would they point (assuming the
earth‘s magnetic field can be neglected)? [A rectangular array of compass nee-
dles aligns itself spontaneously, and this is sometimes used as a demonstration
of ―ferromagnetic‖ behavior on a large scale. It‘s a bit of a fraud, however, since
the mechanism here is purely classical, and much weaker than the quantum me-
chanical exchange forces that are actually responsible for ferromagnetism.13]

More Problems on Chapter 6

! Problem 6.22 In Prob. 6.4, you calculated the force on a dipole by ―brute force.‖
Here‘s a more elegant approach. First write B(r) as a Taylor expansion about the
center of the loop:
B(r) ∼= B(r0 ) + [(r − r0 ) · ∇0 ]B(r0 ),
where r0 is the position of the dipole and 0∇denotes differentiation with respect to
r0. Put this into the Lorentz force law (Eq. 5.16) to obtain

F = I dl × [(r · ∇0)B(r0)].

Or, numbering the Cartesian coordinates from 1 to 3:

. Σ
. 3
Σ Σ
Fi = I ‹ijk rl dl j ∇0l Bk (r0) ,
j,k,l=1

where ‹ ijk is the Levi-Civita symbol ( +1 if ijk = 123, 231, or 312; 1−if ijk =
132, 213, or 321; 0 otherwise), in terms of which the cross-product can be written
.3 ‹ A B . Use Eq. 1.108 to evaluate the integral. Note that
(A × B)i = j,k=1 ijk j k
. 3
‹ijk‹ljm = δil δkm − δimδkl ,
j =1

where δij is the Kronecker delta (Prob. 3.52).

13For an intriguing exception, see B. Parks, Am. J. Phys. 74, 351 (2006), Section II.
6.4 Linear and Nonlinear Media 293

z x

(a) (b)

FIGURE 6.31

Problem 6.23 A familiar toy consists of donut-shaped permanent magnets (magne-

tization parallel to the axis), which slide frictionlessly on a vertical rod (Fig. 6.31).
Treat the magnets as dipoles, with mass md and dipole moment m.

(a) If you put two back-to-back magnets on the rod, the upper one will ―float‖—the
magnetic force upward balancing the gravitational force downward. At what
height (z) does it float?

(b) If you now add a third magnet (parallel to the bottom one), what is the ratio of
the two heights? (Determine the actual number, to three significant digits.)
[Answer: (a) [3μ0m2/2πmd g]1/4; (b) 0.8501]
Problem 6.24 Imagine two charged magnetic dipoles (charge q, dipole moment m),
constrained to move on the z axis (same as Problem 6.23(a), but without gravity).
Electrically they repel, but magnetically (if both m‘s point in the z direction) they
attract.

(a) Find the equilibrium separation distance.

(b) What is the equilibrium separation for two electrons in this orientation.
[Answer: 4.72 × 10−13 m.]
(c) Does there exist, then, a stable bound state of two electrons?

Problem 6.25 Notice the following parallel:

.
∇ · D = 0, ∇ × E = 0, ‹0E = D − P, (no free charge);
∇ · B = 0, ∇ × H = 0, μ0H = B − μ0M, (no free current).
Thus, the transcription D→ B, E→ H, P →μ0M, ‹0 →μ0 turns an electrostatic
problem into an analogous magnetostatic one. Use this, together with your knowl-
edge of the electrostatic results, to rederive

(a) the magnetic field inside a uniformly magnetized sphere (Eq. 6.16);

(b) the magnetic field inside a sphere of linear magnetic material in an otherwise
uniform magnetic field (Prob. 6.18);
 6.4 Linear and Nonlinear Media 294

(c) the average magnetic field over a sphere, due to steady currents within the
sphere (Eq. 5.93).

Problem 6.26 Compare Eqs. 2.15, 4.9, and 6.11. Notice that if ρ, P, and M are
uniform, the same integral is involved in all three:
∫
r̂
dτ J .
r2
Therefore, if you happen to know the electric field of a uniformly charged object,
you can immediately write down the scalar potential of a uniformly polarized ob-
ject, and the vector potential of a uniformly magnetized object, of the same shape.
Use this observation to obtain V inside and outside a uniformly polarized sphere
(Ex. 4.2), and A inside and outside a uniformly magnetized sphere (Ex. 6.1).

1 B1

1
2

B2 2

FIGURE 6.32

Problem 6.27 At the interface between one linear magnetic material and another,
the magnetic field lines bend (Fig. 6.32). Show that tan θ2/ tan θ
=1 μ2/μ1, assum-
ing there is no free current at the boundary. Compare Eq. 4.68.

! Problem 6.28 A magnetic dipole m is imbedded at the center of a sphere (radius

R) of linear magnetic material (permeability μ). Show that the magnetic field inside
the sphere (0 < r ≤ R) is
. Σ
μ 1 2(μ0 − μ)m
[3(m · r̂)r̂ − m] + .
4π r 3 (2μ0 + μ)R3
What is the field outside the sphere?

Problem 6.29 You are asked to referee a grant application, which proposes to deter-
mine whether the magnetization of iron is due to ―Ampère‖ dipoles (current loops)
or ―Gilbert‖ dipoles (separated magnetic monopoles). The experiment will involve
a cylinder of iron (radius R and length L=10R), uniformly magnetized along the
direction of its axis. If the dipoles are Ampère-type, the magnetization is equivalent
to a surface bound current Kb = M φ̂; if they are Gilbert-type, the magnetization is
equivalent to surface monopole densities σ=b M ± at the two ends. Unfortunately,
these two configurations produce identical magnetic fields, at exterior points. How-
ever, the interior fields are radically different—in the first case B is in the same
6.4 Linear and Nonlinear Media 295

general direction as M, whereas in the second it is roughly opposite to M. The ap-

plicant proposes to measure this internal field by carving out a small cavity and
finding the torque on a tiny compass needle placed inside.
Assuming that the obvious technical difficulties can be overcome, and that the
question itself is worthy of study, would you advise funding this experiment? If so,
what shape cavity would you recommend? If not, what is wrong with the proposal?
[Hint: Refer to Probs. 4.11, 4.16, 6.9, and 6.13.
6.4 Linear and Nonlinear Media 296