Chapter 11

Basic Magnetism

There are no free magnetic charges in our part of the universe. The sources of magnetic
fields, down to the atomic scale, are electrical currents. A small current loop appears from
a distance as a magnetic dipole. Thus orbiting electrons give to an atom a magnetic dipole
moment associated with its orbital angular momentum. In addition, subatomic particles
also have magnetic dipole moments. The magnetic dipole moment of a nucleus comes in
part from the proton currents within it, and the magnetic dipole moment of a proton (and
other baryons, and mesons) comes in part from quark currents within it. In addition, the
electron and the up and down quarks also have an “intrinsic” magnetic moment associated
with their spin, although they are supposed to be point-like; the heavier leptons and quarks
have one too. In a naive classical picture, the electron continually goes in circles around
itself. The Dirac equation endows the electron with a spin S = 12 h̄ and a magnetic moment
µB = eh̄/2mc. Both arise from quantum relativistic effects that are built into Dirac’s theory.
Beyond that, quantum electrodynamics predicts that the intrinsic magnetic moment of the
electron is actually µ = 1.00116 µB , in stunning agreement with experiment. Generally,
the magnetic dipole moments of elementary particles are understood much better than their
masses.
To begin with, we discuss the magnetic field associated with steady currents, and we
also postulate the existence of point-like magnetic dipoles that are of pure quantum origin.
The “orbital” dipoles actually have atomic or nuclear sizes and can be described classically
as tiny current loops, except that the associated angular momentum is quantized in units of
h̄. The “intrinsic” moments of elementary particles may be truly point-like and can only be
understood in a relativistic quantum theory.

11.1 Steady currents

The current density J is related to the charge density ρ by the continuity equation
∂ρ
+∇·J=0 (11.1)
∂t

82
In a steady state ρ cannot depend on t and consequently ∇ · J = 0, at least in the region
under consideration. (There may be external current sources that are described as boundary
conditions on J). A current field, or any vector field, that has zero divergence is said to be
transverse. A reason for this name is that if we Fourier-expand
Z
d3 k
J(x) = exp(ik · x) J(k) (11.2)
(2π)3

we find that ∇ · J = 0 implies that k · J(k) = 0.

To find J within the region of interest, we also need an equation for ∇ × J or the
equivalent. For particles streaming in empty space, we can find J(x) by solving the equations
of motion of the individual particles, or the equivalent hydrodynamic equations for the
particle stream. For currents in material media, we can use the simple local relation (Ohm’s
law):
J(x) = σ(x)E(x) (11.3)
where σ is the electrical conductivity (not to be confused with the surface charge density!).
This relation holds in a conductor, which is a medium where a steady current flows indefi-
nitely under the action of a steady field. The resulting equations,

∇ × E = 0, ∇ · J = 0, J = σE, (11.4)

can be used to find J and E in conductors in a similar way as

∇ × E = 0, ∇ · D = ρ, D = E (11.5)

are used in to find D and E in insulators. To be complete, the set (11.4) must be supple-
mented in two ways:

• There must be sources, or boundary conditions that are in effect sources. Most com-
monly the normal component Jn is assigned on part of the outer boundary to describe
current entering and exiting through “leads”.

• At the interface between different conductors Jn and Ek are continuous, and corre-
spondingly En and Jk are discontinuous. The discontinuity of En means that a surface
charge accumulates at the interface. There can be regions of empty space where σ = 0,
hence J = 0; in these regions ∇ · J = 0 is replaced by ∇ · E = ρ/0 .

There is then a whole new set of boundary value problems dealing with current dis-
tribution in conductors, and an even wider set where both conductors and dielectrics are
present. All these problems have practical importance, and some of the solutions obtained
earlier can simply be transcribed. For instance, two-dimensional electrostatic problems be-
come problems of conduction in a thin metal sheet or other conducting layer.

83
11.1.1 Conductance and resistance
The simplest and most common problem is that of a wire or rod of length d and constant
cross section A. The current is I = JA and the electric field is E = V /d, where V is the
potential difference between the ends of the wire. Thus I = (σA/d) V and the conductance
is σA/d. This formula is analogous to C = A/d for the capacitance of a parallel-plate
capacitor. The inverse of the conductance is the resistance
V d ρd
R≡ = = (11.6)
I σA A
where we have defined the resistivity ρ = 1/σ. In the SI system the unit of resistance is
the ohm : 1 ohm = 1 V/A. Resistivity is commonly expressed in ohm · cm conductivity in
(ohm · cm)−1 . Representative values of σ are 64 × 104 for Cu and around 1 (but strongly
dependent on doping) for Si, both at 0◦ C.

11.2 Currents and magnetic forces

Experimentally, the basic fact is that there is a force between two current-carrying wires.
Ampère found that for two long parallel wires, separated by a distance d, the force per unit
length of wire is
dF µ 0 I1 I2
=− (11.7)
dl 2π d
where the − sign indicates that the force is attractive when the currents flow in the same
direction. Ampère also found that two long current-carrying wires at right angles exert no
force on each other, quite unlike the Coulomb case. By fiddling with two current loops 1
and 2 he found that in general
µ0 I1 I2 I I x1 − x2
Fon 1 by 2 = − dl1 · dl2 (11.8)
4π |x1 − x2 |3

It is not obvious how this action-at-a-distance law can be written in terms of a force field. A
clue came from earlier observations by Biot and Savart who had seen that a compass needle
is deflected by a nearby current.

11.3 The magnetic field

It was known since Gilbert that two magnetized bars, or needles, exert on each other forces
and torques entirely analogous to those between two electric dipoles. That suggests that a
magnetized thin bar or needle has a magnetic dipole moment m and generates a magnetic
field !
µ0 m · x
B(x) = −∇ (11.9)
4π |x|3

84
(when the bar is at x = 0). And the energy of a dipole m in the field B is −m · B. These
relations can be used to define the standard unit dipole and the units of B, in the same
way as for Coulomb’s law. However, these definitions have now been replaced by others, as
discussed below.
Comparing the forces exerted by currents with those exerted by magnetic bars, Biot
and Savart found that at a distance R from a long straight wire carrying a current I there
is a magnetic field of magnitude
µ0 I
B= (11.10)
2πR
and direction given by the right-hand rule (Jackson Fig. 5.2).
For a general current loop the Biot-Savart law is

µ0 I2 I dl2 × (x1 − x2 )
B(x1 ) = (11.11)
4π |x1 − x2 |3

and this suggests that one could write the Ampère force law as
µ 0 I1 I
Fon 1 by 2 = dl1 ×B(x1 ) (11.12)
4π
or more explicitly
µ 0 I1 I2 I I dl2 × (x1 − x2 )
Fon 1 by 2 = dl1 × (11.13)
4π |x1 − x2 |3
It is a simple matter to get back Eq. (11.8) from this by using

dl1 × [dl2 × (x1 − x2 )] = − (dl1 ·dl2 ) (x1 − x2 ) + dl2 [dl1 · (x1 − x2 )] (11.14)

and noting that I I

x1 − x2 1
dl1 · 3 = − dl1 · ∇ =0 (11.15)
|x1 − x2 | |x1 − x2 |
More generally, the magnetic force acting on a current distribution J is
Z
F= J(x) × B(x) d3 x (11.16)

In the SI system, the dimensions of B are M T −1 C −1 and the unit of B is the Tesla
(T):
1 T = 104 gauss = 10 kilogauss. (11.17)
The Earth’s field on its surface is around 0.5 gauss, but interstellar fields are of the order of
10−5 gauss. The magnet at the high B/T facility in the microkelvin building is 20 T, and
the continuous magnets at the National High Magnetic Field Laboratory are 20–30 T. Some
neutron stars are thought to produce fields of 104 T or more.

85
11.4 Differential equations for the B field
The basic equation (11.11) for the magnetic field for any current distribution J(x) is

µ0 Z x − x0 3 0
B(x) = J(x0 ) × dx (11.18)
4π |x − x0 |3
µ0 Z J(x0 ) 3 0
= ∇× d x. (11.19)
4π |x − x0 |

Since B can be written as the curl of some vector function, then it follows that ∇ · B = 0.
By taking the curl of the above equation again, we obtain
Z ! Z Z !
J(x0 ) 3 0 J(x0 ) 3 0 2 J(x0 ) 3 0
∇× ∇× d x = ∇ ∇· d x − ∇ d x. (11.20)
|x − x0 | |x − x0 | |x − x0 |

But Z
J(x0 ) 3 0 Z 0 1
∇· d x = J(x ) · ∇ d3 x0 . (11.21)
|x − x0 | |x − x0 |
Replace ∇(1/|x − x0 |) with −∇0 (1/|x − x0 |) and integrate by parts,
Z Z
1 ∇0 · J(x0 ) 3 0
J(x0 ) · ∇ d3 0
x = d x = 0. (11.22)
|x − x0 | |x − x0 |

Hence we are left with

Z Z !
J(x0 ) 3 0 1
∇× ∇× d x = − J(x0 )∇2 d3 x0 = 4πJ(x), (11.23)
|x − x |
0 |x − x0 |

where we have used ∇2 (1/|x − x0 |) = −4πδ(x − x0 ). Therefore the magnetic field satisfies
Ampére’s Law,
∇ × B = µ0 J. (11.24)
By applying Stokes’ theorem to this equation, we obtain Ampére’s Law in integral form,
I Z
B·dl = µ0 J · n da = µ0 I. (11.25)
C S

11.5 The vector potential

Just as the equation ∇ × E = 0 has the general solution E = −∇Φ, so the equation ∇ · B =
0 has the general solution
B=∇×A (11.26)
where A is the vector potential. The analog of ∇2 Φ = −ρ/0 is

∇× (∇ × A) = µ0 J (11.27)

86
or
∇ (∇ · A) − ∇2 A = µ0 J. (11.28)
This equation can be further simplified (see next section) but in any case A, being a vector,
is not as convenient to work with as Φ. Boundary value problems in magnetostatics are
harder than those in electrostatics.
The equations for the B field in a current-free region

∇ · B = 0, (11.29)

∇ × B = 0, (11.30)
are the same as those for the E field in a charge-free region. Thus in many problems we can
introduce a magnetic scalar potential by

B = −∇ΦM (11.31)

and reduce the problem to solving

∇2 ΦM = 0 (11.32)
subject to the appropriate boundary conditions. This is generally much easier than solving
for A. However, some physicists regard the use of ΦM as wrong in principle and even immoral.

11.5.1 Gauge invariance

The vector potential is not unique, because for any Λ(x) we obtain the same B from A and
from
A0 = A + ∇Λ (11.33)
This freedom of adding an arbitrary gradient to A is called gauge invariance. It has a deep
meaning that can only be fully appreciated in quantum theory. It allows us to simplify the
equation for A by imposing one additional condition that can in principle be satisfied by an
appropriate choice of Λ. For instance, in the infinite domain the exact expression
Z
µ0 J(x0 ) 3 0
B(x) = ∇× dx (11.34)
4π |x − x0 |

suggests taking for A the particular choice

µ0 Z J(x0 ) 3 0
AC (x) = dx (11.35)
4π |x − x0 |

which is the direct analog of the Coulomb potential,

1 Z ρ(x0 ) 3 0
Φ(x) = d x. (11.36)
4π0 |x − x0 |

87
It is easily checked that the explicit solution (11.35) obeys the condition

∇ · AC = 0. (11.37)

More generally, gauge invariance allows us to impose ∇ · AC = 0 as an additional condition,

thus restricting A to the Coulomb gauge (which in the time-independent case is the same as
another popular gauge, the Lorentz gauge). In this gauge Eq. (11.28) reduces to

∇2 AC = −µ0 J, (11.38)

which is the close analog of the Poisson equation ∇2 Φ = −ρ/0 . In general the choice of
a suitable gauge condition can be advantageous. For instance, in Cartesian coordinates
∇2 A = −µ0 J gives three separate Poisson equations for Ax , Ay , Az . One can get general
solutions for each component and then impose ∇ · A = 0. Often there are shortcuts.

88