Magnetism in Matter: 13.1 Macroscopic Equations
Magnetism in Matter: 13.1 Macroscopic Equations
Magnetism in Matter: 13.1 Macroscopic Equations
Magnetism in matter
µ0 Z M(x0 ) × (x − x0 ) 3 0 µ0 Z 1
A(x) = 3 dx = M(x0 ) × ∇0 d3 x0 (13.1)
4π 0
|x − x | 4π |x − x0 |
µ0 Z ∇0 × M(x0 ) 3 0
A(x) = dx (13.2)
4π |x − x0 |
JM = ∇ × M (13.3)
∇ · B = 0, (13.4)
∇ × B = µ0 (J + ∇ × M) (13.5)
97
The macroscopic equations are obtained by a suitable average of these, and have
the same form. From now on in this lecture, M and B are averaged quantities. It is also
convenient to introduce the field
H = B/µ0 − M (13.6)
so that the macroscopic equations become
∇ · B = 0, (13.7)
∇×H = J (13.8)
in analogy with
∇·D = ρ (13.9)
∇×E = 0 (13.10)
H is called the magnetic field and B is called the magnetic flux density. The trouble with
these names is that they suggest that the fundamental field is H, in analogy to electrostatics,
where the fundamental field is E, while actually the fundamental field is B.
As in the electrical case, the boundary conditions at the interface of two media are
that Bn is continuous and Hk is continuous in the absence of a surface current density.
A major difference from the electric case is that many materials exhibit a spontaneous
magnetization and do not have a simple local relation between M and H that is analogous
to P = χ(e) E. If there is a relation
M = χ(m) H (13.11)
where χ(m) is the magnetic susceptibility, then one defines the magnetic permeability µ =
µ0 (1 + χ(m) ) and obtains
B = µH (13.12)
in analogy with D = E. Because this relation makes life simple, one tries to use it as much
as possible, allowing µ to depend on H and even on the history of the sample.
F = qv × B (13.13)
is such that F · v = 0. Thus the energy of any particle is not changed when a magnetic field
is switched on and the thermal distribution, which depends only on energy, is unchanged.
98
More formally, the Hamiltonian of a particle of mass M, classically or quantally, is
1
H= (p − qA(x))2 + V (x) (13.14)
2M
All thermodynamic results are derivable from the partition function Z. Classically
Z
Zcl = d3 p d3 x exp(−H/kB T ) (13.15)
The integral over momenta can be performed by changing variable from p to p − qA(x),
leaving a Zcl that does not depend on A:
Z
3/2
Zcl = (2πM kB T ) d3 x exp (−V (x)/kB T ) (13.16)
Quantum theory is essential to describe magnetism in the real world. To see how
things work out, we take a constant field in the z direction. The vector potential depends
on the gauge. A popular gauge choice in quantum theory is the Landau gauge, where
Ax = − By Ay = 0 Az = 0 (13.17)
but in our case it is more elegant to use the symmetrical gauge, where
1
A= B×x (13.18)
2
Then
p2 q q2B 2 2
H= − B·x×p+ (x + y 2 ) + V(x) (13.19)
2M 2M c 8M c2
Let us consider separately the two field-dependent terms in H, first for atoms, then for
molecules and solids
99
if we understand that R R
D E r2 ρ(x) d3 x r2 ρ(x) d3 x
2
r = R = (13.23)
ρ(x) d3 x Z
Equating the energy change to − 12 B· m(dia) we obtain the induced dipole moment
Ze2 hr2 i
m(dia) = − B (13.24)
6Me c2
We should use Blocal instead of B, but the difference turns out to be insignificant. With N
atoms per unit volume, the diamagnetic susceptibility is then
N Ze2 hr2 i
χ(dia) = − (13.25)
6Me c2
For atomic hydrogen hr2 i = 3a20 and recalling χ(el) = (9/2)a30 we find −χ(dia) /χ(el) =
2
(1/9)e2 /Me c2 a0 = (1/9) (e2 /h̄c) ' 5.92 · 10−6 . The ratio e2 /Me c2 a0 is also equal to hv 2 i /c2 ,
in agreement with the earlier remark that magnetism is a relativistic effect. In general, the
magnitude of χ(dia) is Z · 10−6 per mole.
The paramagnetic term can be written in various ways:
q q q
− B·x×p=− B·L=− B(xpy − ypx ) (13.26)
2M c 2M c 2M c
Classically, we compute Zcl by integrating over the direction and the magnitude of L.
This is equivalent to integrating over all values of p and gives a result that does not depend
on B, as one can double-check. In the real world, both the direction and the magnitude of
L are quantized, and there is also the spin. The full paramagnetic term is
e
− B· (L + g0 S) (13.27)
2Me c
for each electron, where g0 = 2 · 1.00116. The individual electron’s L and S combine to form
a total L and S, and these in turn add up to a total angular momentum J = L + S (not to be
confused with the current). In an atom or ion only one value of J contributes significantly
to the quantum partition function Z, and to first order in B the paramagnetic interaction
is (not obviously) equivalent to
eg
− B·J (13.28)
2Me c
where, taking g0 = 2, the Landé g factor is given by
" #
3 1 S(S + 1) − L(L + 1)
g= + (13.29)
2 2 J(J + 1)
100
interacting with the field. For small values of µH/kB T, we can treat the problem classically
and we find, as in the electric case,
µ2
hµz i = H (13.31)
3kB T
This leads to Curie’s law for the paramagnetic susceptibility
!2
(para) N J(J + 1) egh̄ C
χ = = (13.32)
3kB T 2Me c T
where of course e2 /2a0 = 13.6 eV is the ground state energy, and kB T ' 1/40 at room
temperature. In general χ(para) is larger than χ(dia) by roughly the ratio of the ground state
energy to kB T, or about 500 at room temperature. Thus when J 6= 0 gases and insulators are
paramagnetic, but paramagnetism is a weak effect. At low T or high field the magnetization
is a non-linear function of H and “saturates” at Msat = N µ, as discussed more fully below.
101
13.2.3 Ferromagnetism
It would appear from the above that magnetism in matter is always weak, but this is not true.
Electrostatic forces between electrons, coupled with the Pauli exclusion principle, can favor
a parallel alignment of the spins of nearby atoms. This exchange interaction is described
by −J µi · µj for two nearby spins. When all the spins are aligned, it causes an effective
molecular field Nn J µ to act on each spin. Here Nn is the number of nearest neighbors.
When the spins are only partially aligned, the mean molecular field is Nn J hµz i , where hµz i
is computed from the self-consistent equation
µ2
hµz i = (H + Nn J hµz i) (13.34)
3kB T
This equation leads to the Curie-Weiss law
C
χ(m) = (13.35)
T − Tc
showing that the susceptibility blows up at a critical temperature Tc . For T < Tc spontaneous
magnetization occurs.
A spontaneously magnetized piece of material may or may not have a net overall
magnetic moment. Generally the direction of magnetization varies in the different magnetic
domains within the material. In a soft ferromagnet the domains reorient easily and there is
still effectively a linear relation between B and H with a very large effective χ. In a hard
ferromagnet the domains do not reorient easily and the phenomenon of hysteresis occurs.
102