Notes on Magnetism for Solid State Physics 1

Version: Semester 2 AY 2012-2013

Leek Meng Lee email: mlleek@ntu.edu.eg
Contents
1 Basic Concepts and Quantities 2
2 Overview and Classication 3
3 Bohr-van Leeuwen Theorem 4
4 Diamagnetism 5
4.1 General: Atom in a Magnetic Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
4.2 Diamagnetism of core electrons: Larmor or Langevin Diamagnetism . . . . . . . . . . . . 6
4.3 Diamagnetism of conduction electrons: Landau Diamagnetism . . . . . . . . . . . . . . . 7
5 Paramagnetism 9
5.1 Atomic Paramagnetism: Curie or Langevin Paramagnetism . . . . . . . . . . . . . . . . . 9
5.2 Paramagnetism of Conduction Electrons: Pauli Paramagnetism . . . . . . . . . . . . . . . 11
5.3 Van Vleck Paramagnetism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
5.4 Isentropic Demagnetization or Adiabatic Demagnetization . . . . . . . . . . . . . . . . . . 14
6 Collective Magnetism 15
6.1 Direct Exchange Interaction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
6.2 Ferromagnetic Order . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
6.2.1 Weiss Mean Field theory of Ferromagnetism (Simplest theory of ferromagnetism) . 19
6.2.2 Magnons (Quantized Spin Waves) . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
6.2.3 Ferromagnetic Domains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
[Why do domains form?] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
[Types of Domain Walls] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
[Estimation of a 180
o
Blochs Walls energy:] . . . . . . . . . . . . . . . . . . . . . 27
[Magnetization and Hysteresis] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
6.3 Antiferromagnetic and Ferrimagnetic Order . . . . . . . . . . . . . . . . . . . . . . . . . . 29
6.3.1 Weiss Mean Field Theory of Antiferromagnetism . . . . . . . . . . . . . . . . . . . 29
7 Tutorial Problems 31
8 Tutorial Solutions 33
1
1 Basic Concepts and Quantities
Here, we briey review some concepts from electromagnetism.
1. Magnetization

M:

M = lim
V 0
1
V
over V

i
m
i
(1)
where m
i
is the magnetic moment of the ith atom and V is a small volume. Thus magnetization
is also the magnetic dipole moment per unit volume.
2.

H-eld:
The magnetization current is

J
M
=

M and the total

B-eld is

B =
0
_

J +

J
M
_
(2)
where

J is the ordinary current. Dene

H-eld as

H =
1

B

M =

H =

J (3)
3. Magnetic susceptibility :
Assume the constitutive equation

M =

H then

B =
0
(

H +

M) =
0
(1 +)

H (4)
4. Energy of magnetic moment:
Figure 1: A current loop and the area element vector.
d m = Id

A (5)
in standard notation
= d = Id

A (6)
= I
_
d

A (7)
E =

B from electromagnetism (8)
The quantum Hamiltonian will be inspired from that expression.
5. Bohr magneton
B
:
Consider the hydrogen in the n = 1 ground state. In Bohr-Sommerfeld quantization, m
e
vr = h.
magnetic moment: = I A = r
2
I (9)
= r
2
e
t
(10)
= r
2
e
2r/v
(11)
2
Figure 2: The Hydrogen atom.
=
evr
2
(12)
| use Bohr-Sommerfeld quantization rule (13)
=
eh
2m
e
(14)
| dene the Bohr magneton
B
=
eh
2m
e
(15)
=
B
(16)
2 Overview and Classication
Magnetic Phenomena can be classied into 3 main groups:
1. Diamagnetism
susceptibility
dia
< 0
external eld induces magnetic dipoles which orientate themselves antiparallel to the eld
following Lenzs law.
Thus it is a property of all materials.
However it is a weak eect that is usually overwhelmed by other magnetic eects.
Superconductors are perfect diamagnets, i.e. the external eld is completely negated in the
superconductor, or
dia
= 1.
2. Paramagnetism
susceptibility
para
> 0
permanent magnetic dipoles
Localized moments due to partially lled inner electron shells (Curie, Langevin Param-
agnetism).
Itinerant moments due to free conduction electrons (Pauli Paramagnetism).
Collective or Ordered Magnetism
susceptibility is a complicated function = (T, H, history of material)
This type of magnetism is due to (quantum mechanical) exchange interaction between
the permanent magnetic dipoles.
There is a critical temperature T
C
and below T
C
, there is spontaneous magnetization.
3 main types: Ferromagnetism, Ferrimagnetism and Antiferromagnetism.
3
3 Bohr-van Leeuwen Theorem
The Bohr-van Leeuwen theorem is
In a classical system, there is no thermal equilibrium magnetization.
Proof:
Assume a 3D solid with N identical ions with magnetic moment m each. The magnetization is

M =
N
V
m (17)
where m is done using the classical statistical average,
m =
_
dx
1
dx
3N
dp
1
dp
3N
m e
H
_
dx
1
dx
3N
dp
1
dp
3N
e
H
(18)
where =
1
k
B
T
(k
B
is the Boltzmann constant and T is the thermodynamic temperature) and from
E =

B we can write
=
E

B
= m =
H

B
(19)
where H is the classical Hamiltonian. Then we can write
m =
_
dx
1
dx
3N
dp
1
dp
3N
_

B
_
e
H
_
dx
1
dx
3N
dp
1
dp
3N
e
H
(20)
| dene the classical partition function Z =
1
N!h
3N
_
dx
1
dx
3N
dp
1
dp
3N
e
H
(21)
| note
Z

B
=
1
N!h
3N
_
dx
1
dx
3N
dp
1
dp
3N
_

B
_
e
H
so we can write (22)
=
1
Z
Z

B
(23)
The classical Hamiltonian with magnetic eld

B =

A is in the form
H =
1
2m
3N

i=1
_
p
i
+e

A
i
_
2
+ other terms without

A (24)
Z =
1
N!h
3N
_
dx
1
dx
3N
dp
1
dp
3N
e

2m

3N
i=1
( p
i
+e

A
i
)
2
+
(25)
| change variables u
i
= p
i
+e

A
i
and note the integration limits for p
i
goes from to +(26)
=
1
N!h
3N
_
dx
1
dx
3N
du
1
du
3N
e

2m

3N
i=1
u
2
i
+
(27)
which gives the result that Z is independent of

B. In fact, Z is the same as that without the magnetic
eld. Thus
Z

B
= 0 and m = 0. PROVED.
We provide a cartoon picture of the theorem. See gure 3.
4
Figure 3: The cartoon picture of Bohr-van Leeuwen theorem. The closed orbits and skipping orbits
contribute opposite magnetic moments and so they cancel out.
4 Diamagnetism
4.1 General: Atom in a Magnetic Field
Assume there are Z (core) electrons in the atom in a magnetic eld

B =

A, the quantum Hamiltonian

is (where V is the atomic potential)
H =
Z

i=1
1
2m
e
_
p
i
+e

A
i
_
2
+V (28)
=
Z

i=1
1
2m
e
_
p
2
i
+e( p
i

A
i
+

A
i
p
i
) +e
2

A
i

A
i
_
+V (29)
| choose Coulomb gauge

A = 0 (30)
| writing

A(r) =
1
2
(

B r) will satisfy both

A = 0 and

B =

A (31)
=
Z

i=1
1
2m
e
_
p
2
i
+e( p
i

A
i
+

A
i
p
i
)e
2
_
1
2
(

B r
i
)
_
2
_
+V (32)
| note that in position representation, using

A
i
= 0, (33)
| we can write p
i

A
i
+

A
i
p
i
ih

i
(

A
i
) +

A
i
(ih

i
) = 2ih

A
i

i
2

A
i
p
i
(34)
| then substitute

A
i
=
1
2
(

B r
i
) (35)
=
Z

i=1
1
2m
e
_
p
2
i
+e(

B r
i
) p
i
+
e
2
4
(

B r
i
)
2
_
+V (36)
| write (

B r
i
) p
i
= (r
i
p
i
)

B which is the triple product identity (37)
| and note that r
i
p
i
= h

L
i
where

L
i
is the dimensionless orbital angular momentum (38)
=
Z

i=1
1
2m
e
_
p
2
i
+eh

L
i

B +
e
2
4
(

B r
i
)
2
_
+V (39)
| include spin of the electrons interacting with the magnetic eld H
spin
=
B
g
0

S

B (40)
| where g
0
2 and
B
=
eh
2m
e
is the Bohr magneton (41)
5
=
Z

i=1
p
2
i
2m
e
+V +
Z

i=1

B
_

L
i
+g
0

S
i
_

B
. .
paramagnetism
+
Z

i=1
e
2
8m
e
(

B r
i
)
2
. .
diamagnetism
(42)
We said diamagnetism is a phenomena in all materials but the Hamiltonian seems to indicate that
paramagnetism and diamagnetism are universal phenomena. Actually, the paramagnetism term vanishes
for certain states of the atom. We can see that by rst rewriting the paramagnetic Hamiltonian. (we
consider only 1 term in the sum and drop the index i for simplicity)
H
para
=
B
(

L +g
0

S)

B (43)
=
B

B

J
1

J
2
_

L +g
0

J

S
_
(44)
| note (

L)
2
=

J
2
+

L
2
2

L where

J =

L +

S and

J

L commutes with

L

J (45)
| a similar expression is used for

J

S (46)
=
B

B

J
1
2

J
2
_

J
2
+

L
2
(

L)
2
+g
0
(

J
2
+

S
2
(

J

S)
2
)
_
(47)
| use

J =

L +

S (48)
=

B

B

J
2

J
2
_

J
2
+

L
2

S
2
+g
0
(

J
2
+

S
2

L
2
)
_
(49)
| recall that the square of (dimensionless) angular momentum operators (50)
| are:

J
2
= J(J + 1),

L
2
= L(L + 1) and

S
2
= S(S + 1) (51)
| this is written in atomic physics notation (52)
=
B

B

J
_
1 +g
0
2
+
g
0
1
2
_
S(S + 1) L(L + 1)
J(J + 1)
__
(53)
| recall that g
0
2 (54)
=
B

B

J
_
1 +
J(J + 1) L(L + 1) +S(S + 1)
2J(J + 1)
_
(55)
= g
e

B

B

J (56)
where g
e
is called the Lande g-factor. And so in a state |J = 0 (which is determined by Hunds
rules in atomic physics), there is no paramagnetism and thus diamagnetism is truly universal.
4.2 Diamagnetism of core electrons: Larmor or Langevin Diamagnetism
We will consider the state |J = 0 since in other states, the presence of paramagnetism will overwhelm dia-
magnetism. So now H
dia
is the perturbation Hamiltonian. Recall from rst order Rayleigh Schrodinger
time independent perturbation theory, the energy correction is,
E
(1)
n
= n
(0)
|H
int
|n
(0)
(57)
E
(1)
0
=
_
0

i=1
e
2
8m
e
_

B r
i
_
2

0
_
(58)
| where we wrote |J = 0 = |0 (59)
| assume the magnetic eld is in the z-axis

B = (0, 0, B) (60)
=
Z

i=1
e
2
8m
e
0 |(Bx
i
y
i
B) z (Bx
i
y
i
B) z| 0 (61)
=
Z

i=1
e
2
B
2
8m
e
0|x
2
i
+y
2
i
|0 (62)
6
| assume the state is spherically symmetric which is reasonable for J = 0 (63)
| then the expectation values x
2
i
= y
2
i
=
1
3
r
2
i
(64)
=
e
2
B
2
12m
e
Z

i=1
0|r
2
i
|0 (65)
Recall the magnetic moment is =
E

B
, so the moment of this single atom is
=
E
(1)
0
B
=
e
2
B
6m
e
Z

i=1
0|r
2
i
|0 (66)
Assume a 3D solid of volume V with N identical such atoms, the magnetization is (recall

M =
N
V
m),
M =
N
V
=
N
V
e
2
B
6m
e
Z

i=1
0|r
2
i
|0 (67)
Recall

M =

H and

H =
1

B

M. Since M << H, we can approximate H
1

0
B, thus the
susceptibility is,
=
M
H


0
M
B
(68)
=
N
V
e
2

0
6m
e
Z

i=1
0|r
2
i
|0 (69)
A few comments are in order:
A classical derivation gives a very similar expression but Bohr-van Leeuwen theorem tells us that
the classical expression is not the whole story.
The susceptibility seems to be temperature independent but thats because the derivation is not a
statistical mechanical one. Experimentally, it turns out that the temperature dependence is weak.
The negative sign in conrms that it is an eect that follows Lenzs law.
Remember that any other magnetic eect is larger than atomic diamagnetism (except in a super-
conductor).
4.3 Diamagnetism of conduction electrons: Landau Diamagnetism
In metals where there is a sea of mobile electrons, the application of a magnetic eld results in a
diamagnetism eect. The assumption is that there are Z mobile free electrons, the Hamiltonian is the
same as before (without the atomic potential V
i
),
H =
Z

i=1
1
2m
e
_
p
i
+e

A
i
_
2
(70)
and it turns to be exactly solvable. We take the magnetic eld to be in the z-direction,

B = (0, 0, B).
We choose

A = (0, Bx, 0) and it satises

A =

B and

A (Coulomb gauge).
H =
Z

i=1
1
2m
e
(p
ix
x +p
iy
y +p
iz
z +eBx y) (p
ix
x +p
iy
y +p
iz
z +eBx y) (71)
7
=
Z

i=1
1
2m
e
_
p
2
ix
+p
2
iz
+ (p
iy
+eBx)
2
_
(72)
| we can rewrite it to make it look more familiar, dene x
i0
=
p
iy
eB
and
c
=
eB
m
e
(73)
=
Z

i=1
_
p
2
ix
2m
e
+
1
2
m
e

c
(x
i
x
i0
)
2
+
p
2
iz
2m
e
_
(74)
The (time independent) Schrodinger equation is H = E and for position representation p
ix
ih

x
i
,
p
iy
ih

y
i
and p
iz
ih

z
i
. Considering only 1 electron (thus dropping index i),
_

h
2
2m
e

2
x
2
+
1
2
m
e

2
c
_
x
ih
eB

y
_
2

h
2
2m
e

2
z
2
_
= E (75)
write in the form = e
ikzz
e
ikyy
(x) | (76)
_

h
2
2m
e

2
x
2
+
1
2
m
e

2
c
_
x +
hk
y
eB
_
2
+
h
2
k
2
z
2m
e
_
(x) = E(x) (77)
_

h
2
2m
e

2
x
2
+
1
2
m
e

2
c
_
x +
hk
y
eB
_
2
_
(x) =
_
E
h
2
k
2
z
2m
e
_
(x) (78)
This is simply a (shifted potential) harmonic oscillator problem. Hence the energy eigenvalues are
E
h
2
k
2
z
2m
e
=
_
n +
1
2
_
h
c
(79)
E =
_
n +
1
2
_
h
c
+
h
2
k
2
z
2m
e
(80)
Thus the electron is quantized in the x-y plane perpendicular to the eld and it is a free particle in
the z-direction which is the direction of the eld.
Now we will only outline the statistical mechanics calculation leading to the susceptibility. The
(quantum) partition function for a canonical ensemble is calculated using
Z =
_
dk
z

n
e

E
k
B
T
=
_
dk
z

n
e

1
k
B
T

(n+
1
2
)hc+
h
2
k
2
z
2me

(81)
Then the Helmholtz free energy F is given by
F = k
B
T lnZ (82)
The magnetization is obtained from
M =
F
B
=
F

0
H
(83)
Finally, the magnetic susceptibility is obtained from
=
_
M
H
_
H=0
(84)
The result for Landau diamagnetism is

Landau
=
N
2V

2
B

F
(85)
where
F
is the Fermi energy.
8
5 Paramagnetism
5.1 Atomic Paramagnetism: Curie or Langevin Paramagnetism
Consider an atom with electrons in states giving a net angular momentum J (integer or half integer).
1
The paramagnetic Hamiltonian is
H
para
= g
e

B

B

J (86)
Take

B = B z so

B

J = BJ
z
and denoting atomic states as |J, m
J
, the energy eigenvalues are
H
para
|J, m
J
= g
e

B
BJ
z
|J, m
J
= g
e

B
Bm
J
|J, m
J
(87)
Note that there is no h as J
z
is dened to be dimensionless.
The calculation leading to the susceptibility is done using statistical mechanics. The (quantum)
partition function for a canonical ensemble is
Z =
J

m
J
=J
e

E
k
B
T
(88)
=
J

m
J
=J
e

g
e

B
Bm
J
k
B
T
(89)
| note that we can rename the summation to absorb the minus sign (90)
| and let x =
g
e

B
B
k
B
T
(91)
=
J

m
J
=J
e
xm
J
(92)
= e
xJ
+e
x(J1)
+. . . +e
x(J1)
+e
xJ
(93)
= e
xJ
_
1 +e
x
+e
2x
+. . . +e
x(2J1)
+e
x2J
_
(94)
| which is a nite geometric series with rst term = 1, common factor = e
x
(95)
| so use S
N
=
N

k=0
r
k
=
1 r
N+1
1 r
and here there are 2J + 1 terms so N + 1 = 2J + 1 (96)
= e
xJ
1 e
x(2J+1)
1 e
x
(97)
=
e
xJ
e
x(J+1)
1 e
x
(98)
=
e
xJ
e
x/2
e
x(J+1)
e
x/2
e
x/2
e
x/2
(99)
=
e

x
2
(2J+1)
e
x
2
(2J+1)
e

x
2
e
x
2
(100)
| use sinh z =
1
2
(e
z
e
z
) (101)
=
sinh
_
x
2
(2J + 1)
_
sinh
_
x
2
_ (102)
Magnetization is calculated from
M =
F
B
=

B
(k
B
T ln Z) (103)
1
This is determined using Hunds Rules in atomic physics.
9
= k
B
T
1
Z
Z
B
(104)
| but recall x =
g
e

B
B
k
B
T
=
x
B
=
g
e

B
k
B
T
(105)
= g
e

B
1
Z
Z
x
(106)
= g
e

B
1
Z
sinh
_
x
2
_
sinh
_
x
2
(2J + 1)
_
_
2J+1
2
cosh
_
x
2
(2J + 1)
_
sinh
_
x
2
_
sinh
_
x
2
(2J + 1)
_
cosh
_
x
2
_
2 sinh
2
_
x
2
_
_
(107)
= g
e

B
J
_
2J + 1
2J
coth
_
2J + 1
2J
xJ
_

1
2J
coth
_
xJ
2J
__
(108)
| dene the Brillouin function B
J
(y) =
2J + 1
2J
coth
_
2J + 1
2J
y
_

1
2J
coth
_
y
2J
_
(109)
= g
e

B
JB
J
(xJ) (110)
The susceptibility can be calculated using =
_
M
H
_
H=0
=
0
_
M
B
_
B=0
but this will make the make
the algebra very messy. We will calculate 2 specic cases only.
Case 1: Smallest angular momentum value of J =
1
2
.
B
J
(xJ)|
J=1/2
=
2
_
1
2
_
+ 1
2
_
1
2
_ coth
_
2
_
1
2
_
+ 1
2
_
1
2
_
_

1
2
_
1
2
_ coth
_
x
_
1
2
_
2
_
1
2
_
_
(111)
= 2 coth x coth
_
x
2
_
(112)
| use the addition identity coth x =
coth
2
_
x
2
_
+ 1
2 coth
_
x
2
_ (113)
=
1
coth
_
x
2
_ (114)
= tanh
_
x
2
_
(115)
M = g
e

B
J tanh
_
x
2
_
(116)
=
0
_
M
B
_
B=0
=
0
_
M
x
x
B
_
B = 0 or x = 0
(117)
=
0
_
g
e

B
1
2
2
2
_
x
2
_
g
e

B
k
B
T
_
x=0
(118)
=
g
2
e

2
B
4k
B
T
(119)
Case 2: x << 1 which is the usual experimental situation.
We Taylor expand B
J
(xJ) to rst order,
B
J
(xJ) =
2J + 1
2J
coth
_
2J + 1
2J
xJ
_

1
2J
coth
_
xJ
2J
_
(120)
| recall the Taylor series of coth y =
1
y
+
y
3
+ for y << 1 (121)

2J + 1
2J
_
2
(2J + 1)x
+
(2J + 1)x
2 3
_

1
2J
_
2
x
+
x
2 3
_
(122)
10
=
1
Jx
+
(2J + 1)
2
x
12J

1
Jx

x
12J
(123)
=
4J
2
+ 4J
12J
x (124)
=
J + 1
3
x (125)
M = g
e

B
J
J + 1
3
x (126)
=
0
_
M
x
x
B
_
B = 0 or x = 0
(127)
=
0
_
g
e

B
J(J + 1)
3
g
e

B
k
B
T
_
x=0
(128)
=

0
g
2
e

2
B
J(J + 1)
3k
B
T
(129)
This relation of
1
T
is known as the Curies law of paramagnetism.
5.2 Paramagnetism of Conduction Electrons: Pauli Paramagnetism
For conduction electrons, they contribute a paramagnetic moment because each electron has spin
1
2
. For
simplicity, the derivation will be done assuming T = 0.
We split the density of states of conduction electrons into spin up and spin down components.
(E) =

(E) +

(E) (130)
The magnetic eld is assumed to be in the positive z-direction

B = B z and m
s
= +
1
2
(parallel
to

B) and m
s
=
1
2
(antiparallel to

B).
When there is no magnetic eld, there is an equal number of -electrons as -electrons and

(E) =

(E) and there is zero net magnetization.

Figure 4: The density of states for free electrons separated into spin up and spin down parts. There
are equal numbers of spin up and spin down electrons on average. The electrons are lled to
F
, the
Fermi energy.
On application of the magnetic eld

B = B z, and considering only spin, the Hamiltonian of the
problem is the paramagnetic Hamiltonian H
para
= g
0

S

B.
Spin up electrons: H
para
| = 2
B

S

B

m
s
= +
1
2
_
= 2
B
BS
z

m
s
= +
1
2
_
=
B
B

m
s
= +
1
2
_
Spin down electrons: H
para
| = 2
B

S

B

m
s
=
1
2
_
= 2
B
BS
z

m
s
=
1
2
_
=
B
B

m
s
=
1
2
_
11
Thus spin up electrons gain
B
B energy and spin down electrons lose
B
B energy. The density of
states are modied,

(E)

(E
B
B) and

(E)

(E +
B
B) (131)
Figure 5 shows the situation clearly.
Figure 5: The density of states for free electrons separated into spin up and spin down parts under
the inuence of a magnetic eld in the spin up direction. Spin up and spin down electrons gains
B
B
and loses
B
B respectively. Then the system reaches new thermal equilibrium at
F
.
There is now excess spin down electrons and this causes a net magnetization.
M =

B
V
(N

) =
B
(n

) (132)
The number densities of up and down electrons are denoted by n

and n

respectively. We need to
evaluate the number densities.
n

=
_
+

dE

(E
B
B)f
FD
(E) (133)
| where f
FD
is the Fermi-Dirac distribution (134)
=
_
+

B
B
dE

(E
B
B)f
FD
(E) (135)
| make a change of variables E

= E
B
B (136)
=
_

0
dE

(E

)f
FD
(E

+
B
B) (137)
| since
B
B <<
F
, we Taylor expand f
FD
(E

+
B
B) f
FD
(E

) +
B
B
f
FD
(E

)
E

(138)

_

0
dE

(E

)
_
f
FD
(E

) +
B
B
f
FD
(E

)
E

_
(139)
Similarly,
n

_

0
dE

(E

)
_
f
FD
(E

)
B
B
f
FD
(E

)
E

_
(140)
The magnetization is
M =
B
(n

) (141)
=
B
__

0
dE

(E

(E

))f
FD
(E

)
B
B
_

0
dE

(E

) +

(E

))
f
FD
(E

)
E

_
(142)
12
| recall that

(E) =

(E) and

(E) +

(E) = (E) (143)

=
2
B
B
_

0
dE

(E

)
f
FD
(E

)
E

(144)
We can calculate the susceptibility now.

Pauli
=
0
_
M
B
_
B=0
(145)
=
0

2
B
_

0
dE

(E

)
f
FD
(E

)
E

(146)
| at low temperature, we approximate f
FD
(E) (
F
E) which is the step function (147)
| thus
f
FD
(E

)
E

(
F
E

) = (
F
E

) and evaluate the integral (148)

=
0

2
B
(
F
) (149)
| recall in free electron theory (
F
) =
3
2
N
V
1

F
(150)
=
3N
2V

2
B

F
(151)
Notice that there is a relationship between
Landau
and
Pauli
:

Landau
=
1
3

Pauli
(152)
Thus the total susceptibility of a metal in a magnetic eld is

metal
=
Landau
+
Pauli
+
Larmor
+
Curie
(153)
=
1
3

Pauli
+
Pauli
+
Larmor
+
Curie
(154)
=
2
3

Pauli
+
Larmor
+
Curie
(155)
It turns out that the dominant contribution to
metal
depends very much on the material.
5.3 Van Vleck Paramagnetism
Recall in the discussion of diamagnetism, in the state |J = 0 = |0, the paramagnetic Hamiltonian gives
no contribution in rst order perturbation, i.e.
For H
int
=
Z

i=1

B
(

L
i
+g
0

S
i
)

B +
e
2
8m
e
(

B r
i
)
2
(156)
=
Z

i=1
g
e

B

B

J
i
+
e
2
8m
e
(

B r
i
)
2
(157)
E
(1)
0
= 0|H
int
|0 (158)
=
Z

i=1
g
e

B
0|

B

J
i
|0 +
e
2
8m
e
0|(

B r
i
)
2
|0 (159)
| and the rst term is zero (160)
=
e
2
B
2
12m
e
Z

i=1
0|r
2
i
|0 (161)
13
Van Vleck paramagnetism is simply the second order contribution from H
para
=

Z
i=1
g
e

B

B

J
i
.
Thus recall the second order perturbation theory energy correction expression:
E
(2)
n
=

m(=n)
|n|H
int
|m|
2
E
n
E
m
(162)
| in this case, the state for correction is |n = |0 and H
int
= H
para
(163)
E
(2)
0
=

m(=0)

_
0

Z
i=1
g
e

B

B

J
i

m
_

2
E
0
E
m
(164)
Note that the state |J = 0 = |0 is usually the lowest energy state (atomic physics), thus E
m
E
0
> 0
and E
(2)
0
< 0.
Again take

B = (0, 0, B) so

B

J
i
= BJ
iz
.
E
(2)
0
=

m(=0)
Z

i=1
g
2
e

2
B
B
2
|0|J
iz
|m|
2
E
0
E
m
(165)
The moment of this single atom is
=
E
(2)
0
B
=

m(=0)
Z

i=1
g
2
e

2
B
B
|0|J
iz
|m|
2
E
m
E
0
(166)
The magnetization of N identical atoms in volume V is, M =
N
V
. The susceptibility is,


0
M
B
=
N
V

0
B
(167)
=
N
V

0

m(=0)
Z

i=1
g
2
e

2
B
|0|J
iz
|m|
2
E
m
E
0
(168)
This positive susceptibility is called Van Vleck paramagnetism and it is very small. It is temperature
independent because this is a quantum mechanical derivation and not a statistical mechanical derivation.
It turns out, from a more detailed derivation, that the above expression for is essentially correct and
is temperature independent.
5.4 Isentropic Demagnetization or Adiabatic Demagnetization
This is a technique to cool paramagnetic materials. Recall the expression of magnetization in atomic
paramagnetism,
M = g
e

B
J(J + 1)
3
x = g
2
e

2
B
J(J + 1)
3
B
k
B
T
(169)
Thus increasing the B-eld increases the magnetization and increasing T decreases M, which is
compatible with common sense.
The required concepts in entropy are:
Boltzmann denition of entropy: S = k
B
ln where is the number of (micro)arrangements.
At low temperatures, all the magnetic moments align with the magnetic eld and there is only 1
arrangement, = 1, therefore S = 0.
At high temperatures or zero magnetic eld, the system is disordered and in state J, there are
dim(m
J
) = 2J + 1 possibilities for each atom. For N identical atoms, = (2J + 1)
N
giving
S = k
B
ln(2J + 1)
N
= Nk
B
ln(2J + 1).
14
The 2 steps of isentropic cooling are:
1. Isothermal magnetization:
Field is applied to magnetize the material. The temperature of the sample is kept constant by a
helium bath that removes heat from the sample. As seen from above, the entropy decreases on
magnetization.
2. Adiabatic Demagnetization:
The sample is now isolated (adiabatic) and the magnetic eld is slowly reduced to zero (quasi-
static). This demagnetization increases the entropy of magnetic moments. But because it is an
adiabatic process (Q = 0), the entropy change is zero, thus there is a compensating decrease in
entropy elsewhere. This decrease in entropy is in the phonons and hence the material cools.
Figure 6: The entropy-temperature S T diagram showing the 2 steps of isentropic cooling.
6 Collective Magnetism
Notice that we could understand the phenomena of diamagnetism and paramagnetism without assuming
any explicit interaction among the magnetic moments (or among the electrons that gave rise to these
magnetic moments).
For collective magnetism, there is interaction between the magnetic moments and this give rise to
cooperative or collective phenomena (ferromagnetism, ferrimagnetism and antiferromagnetism). These
interactions are called under the title of exchange interactions. The signature of cooperative phe-
nomena is a spontaneous ordering of magnetic moments below a critical temperature. In ferro- and
ferrimagnetism, this is called the Curie temperature T
C
and in antiferromagnetism, this is called the
Neel temperature T
N
.
We make a broad classication rst:
Insulators
magnetic moments are from partially lled electron shells
can be ferro-, ferri- or antiferromagnetic
well described by Heisenberg model (see later)
Metals
Band magnetism (Beyond the scope here.)
15
conduction electrons are responsible for magnetism
exchange interaction causes a spin-dependent band shift (for T < T
C
) and thus a partic-
ular spin orientation is preferred
simplest model is the Hubbard model
Localized magnetism
magnetic moments are not from conduction electrons
exchange interaction is between localized electrons that contributes magnetic moments
and itinerant conduction electrons
6.1 Direct Exchange Interaction
The physics behind direct exchange interaction is extremely simple: Coulomb interaction + Paulis
exclusion principle.
We will illustrate this using the standard Heitler-London treatment in molecular physics. We will
also derive the Heisenberg model.
Figure 7: Hydrogen molecule consisting of 2 hydrogen atoms.
Consider the model of a hydrogen molecule (or 2 hydrogen atoms). The Hamiltonians of the 2
separate atoms (in position representation) are,
First atom: H
a
0
(1) =
h
2
2m
e

2
r
a1

e
2
4
0
r
a1
(170)
Second atom: H
b
0
(2) =
h
2
2m
e

2
r
b2

e
2
4
0
r
b2
(171)
Molecule: H = H
a
0
(1) +H
b
0
(2)
e
2
4
0
_
1
r
a2
+
1
r
b1

1
r
12

1
R
_
(172)
The Hamiltonian of the molecule gives an solvable Schrodinger equation. Thus we shall make an
approximation using physically reasonable wavefunctions.
Exchange Symmetry of 2 quantum particles: (Quantum Mechanics 3)
The 2-boson wavefunction obeys the following under exchange:

boson
(1, 2)
12

boson
(2, 1) =
boson
(1, 2) (173)
The 2-fermion wavefunction obeys the following under exchange:

fermion
(1, 2)
12

fermion
(2, 1) =
fermion
(1, 2) (174)
16
Now we make the guess for the spatial part of the 2-electron wavefunction.

+
=
1

2
_

a
(1)
b
(2) +
a
(2)
b
(1)
_
(175)

=
1

2
_

a
(1)
b
(2)
a
(2)
b
(1)
_
(176)
It appears that
+
does not obey the 2-fermion exchange symmetry and indeed it does not, but we
keep
+
anyway.
For the spin part, we recall the results of the addition of the two spin-
1
2
angular momenta.
s
1
=
1
2
, s
2
=
1
2
=s = 0, 1 (177)

m
s
1
=
1
2
, m
s
2
=
1
2
_
= |
1

m
s
1
=
1
2
, m
s
2
=
1
2
_
= |
1

m
s
1
=
1
2
, m
s
2
=
1
2
_
= |
1

m
s
1
=
1
2
, m
s
2
=
1
2
_
= |
1

_
|s = 0, m
s
= 0 = |0, 0 =
1

2
(|
1

2
|
1

2
) =
S
(1, 2)
|s = 1, m
s
= 1 = |1, 1 = |
1

2
=
T
(1, 2)
|s = 1, m
s
= 0 =
1

2
(|
1

2
+|
1

2
) =
T
(1, 2)
|s = 1, m
s
= 1 = |
1

2
=
T
(1, 2)
where
S
(1, 2) is the singlet state and
S
(1, 2)
12

S
(2, 1) =
S
(1, 2). And
T
(1, 2) are the triplet
spin states with
T
(1, 2)
12

T
(2, 1) =
T
(1, 2).
Now we can construct the full spatial + spin wavefunctions obeying 2-fermion exchange symmetry.
Recall that
symmetric antisymmetric = antisymmetric (178)
symmetric symmetric = symmetric (179)
antisymmetric antisymmetric = symmetric (180)

S
=
+

S
(181)
=
1

2
_

a
(1)
b
(2) +
a
(2)
b
(1)
_

S
(1, 2) (182)

T
=

T
(183)
=
1

2
_

a
(1)
b
(2)
a
(2)
b
(1)
_

T
(1, 2) (184)
and
S
and
T
obey the exchange symmetry, i.e.
S
(1, 2)
12

S
(2, 1) =
S
(1, 2) and
T
(1, 2)
12

T
(2, 1) =
T
(1, 2).
The energy expectation values are, (note that these are not eigenvalues!)
E
S
=
_
d1d2

S
H
S
(185)
E
T
=
_
d1d2

T
H
T
(186)
Note the dierence between the 2 energies are,
E
S
E
T
=
_
d1d2 (

S
H
S

T
H
T
) (187)
| recall that
S
and
T
are orthogonal to each other (188)
| and expand out the spatial parts (189)
=
_
d1d2
a
(1)
b
(2)H
a
(2)
b
(1) +
_
d1d2
a
(2)
b
(1)H
a
(1)
b
(2) (190)
J (191)
Now we want to make an eective Hamiltonian that treats the wavefunctions
S
and
T
as eigenstates
and E
S
and E
T
as the respective eigenvalues.
17
We rst need to note this manipulation:
Consider,
_

S
1
+

S
2
_
2
=

S
2
1
+

S
2
2
+ 2

S
1

S
2
(192)

S
1

S
2
=
1
2
_
_

S
1
+

S
2
_
2

S
2
1


S
2
2
_
(193)
| note that

S
1
+

S
2
=

S (194)
=
1
2
_

S
2

S
2
1


S
2
2
_
(195)
| recall that the

S
2
= s(s + 1) (recall that they are dimensionless)(196)
=
1
2
[s(s + 1) s
1
(s
1
+ 1) s
2
(s
2
+ 1)] (197)
| and for s
1
=
1
2
, s
2
=
1
2
we have s = 0 and s = 1 (198)
=
1
2
_
s(s + 1)
3
2
_
(199)
=
_

3
4
for s = 0
1
4
for s=1
(200)
Thus if we dene the eective Hamiltonian,
H
e
=
1
4
(E
S
+ 3E
T
) (E
S
E
T
)

S
1

S
2
(201)
it has 2 eigenvalues,
for s = 0, H
e
|s = 0 =
_
1
4
(E
S
+ 3E
T
) (E
S
E
T
)
_

3
4
__
|s = 0 = E
S
|s = 0 (202)
for s = 1, H
e
|s = 1 =
_
1
4
(E
S
+ 3E
T
) (E
S
E
T
)
_
1
4
__
|s = 1 = E
T
|s = 1 (203)
Using the notation of exchange integral J,
H
e
=
1
4
(E
S
+ 3E
T
) J

S
1

S
2
(204)
The rst term is a constant and can always be absorbed by a shift. This gives the Heisenberg
Hamiltonian for a 2-spin system,
H
Heisenberg
J

S
1

S
2
(205)
A few comments are in order:
What we have done is to show that Coulomb interaction (not explicitly shown here actually) and
Paulis Exclusion principle (via exchange symmetry) can give rise to singlet spin states (preferring
opposite spins) and triplet spin states (preferring parallel spins) having dierent energy values. The
singlet state turns out to be lower in energy and thus in this case, the opposite spin arrangement
is preferred.
This is not really a derivation of Heisenberg Hamiltonian. This is just a justication that even
without the presence of a magnetic eld, other interactions can still bring about a preferred spin
arrangement.
For nearest-neighbour interaction in a multispin system, we can write the Heisenberg Hamiltonian
as
H
Heisenberg
=

i,j
J
ij

S
i

S
j
(206)
18
where i, j means over nearest neighbours and the exchange integral J
ij
is usually taken to be
a parameter.
The comparison between the dierent models are:
Heisenberg model: H =

i,j
J
ij
_
S
x
i
S
x
j
+S
y
i
S
y
j
+S
z
i
S
z
j
_
(207)
Ising model: H =

i,j
J
ij
S
z
i
S
z
j
(208)
XY model: H =

i,j
J
ij
_
S
x
i
S
x
j
+S
y
i
S
y
j
_
(209)
The other types of exchange interactions are called indirect exchange interactions. The dierent
classes of indirect exchange interactions are:
Superexchange interaction
This exchange interaction is mediated by a non-magnetic ion in-between the magnetic ions.
RKKY exchange interaction
This is devised by Ruderman, Kittel, Kasuya and Yosida. Exchange is mediated by valence
electrons and thus the exchange integral is distance dependent.
Double exchange
Magnetic ions can have dierent oxidation states and electron hopping between ions results
in an extended ferromagnetic (i.e. parallel spins) exchange interaction.
6.2 Ferromagnetic Order
Ferromagnetism refers to spontaneous magnetization, without an applied eld, such that the magnetic
moments all point in the same direction. Beyond the so-called Curie temperature T
C
, the interaction
breaks down and the material becomes paramagnetic.
There are 2 aspects of ferromagnetism:
1. within a domain, the magnetic moments are all aligned
2. domains interact to produce other magnetic eects such as hysteresis
We will discuss the rst aspect: interactions within a domain.
6.2.1 Weiss Mean Field theory of Ferromagnetism (Simplest theory of ferromagnetism)
A theory of ferromagnetism (and collective magnetism) must be able to at least show a phase transition
between paramagnetism and ferromagnetism. The transition temperature should be calculable from the
theory.
In Weiss Mean eld theory, he took the paramagnetic Hamiltonian H
para
= g
e

B

B

J and wrote

B

B+

B
MF
where

B is the external eld and

B
MF
is some kind of internal mean eld (or exchange
interaction or molecular eld).
He further assumed that the mean eld is proportional to the magnetization.

B
MF
=

M (210)
where is a parameter and

M is the magnetization. The Hamiltonian for the theory is then
H
Weiss
=
Z

i=1
g
e

B

J
i
(

B +

B
MF
) =
Z

i=1
g
e

B

J
i
(

B +

M) (211)
19
Since this Hamiltonian is so similar to the one used for Curie Paramagnetism, we can immediately
carry the results over. We carry over the magnetization expression
M = g
e

B
JB
J
(xJ) (212)
and B
J
(xJ) is the Brillouin function with the modication:
x =
g
e

B
(B +M)
k
B
T
(213)
For external eld B = 0,
M = g
e

B
JB
J
_
g
e

B
M
k
B
T
J
_
(214)
Thus solving for M is not trivial but can be done graphically. Let the equations be written as 2
equations
y = M (215)
y = g
e

B
JB
J
_
g
e

B
M
k
B
T
J
_
(216)
The graphical solution of M will mean nding the intersection points of these 2 graphs.
M = 0 is always a solution but it is trivial. We shall now discusss the non-trivial solutions. We see
that for dierent values of temperatures y = g
e

B
JB
J
(. . .) can change and we have 3 possible cases:
Figure 8: The 3 cases of the solutions of M.
1. For case (a):
The only solution for M = 0 which is the paramagnetism situation (recall that there is no external
eld). So it represents T > T
C
.
20
2. For case (c):
The non-trivial solution is marked by the cross. This non-zero M represents a spontaneous mag-
netization (ferromagnetism) and this occurs for T < T
C
.
3. For case (b):
The second intersection point is about to appear. Thus it makes sense to dene this temperature
as the transition temperature T
C
. We can actually use this denition to solve for T
C
. The condition
is that: for small M, the gradient of y = g
e

B
JB
J
(. . .) should be 1, therefore (write T = T
C
),
d
dM
g
e

B
JB
J
(. . .)

M=0
= 1 (217)
d
dM
g
e

B
JB
J
_
g
e

B
M
k
B
T
C
J
_

M=0
= 1 (218)
since M is small, we recall the expansion B
J
(xJ)
J + 1
3
x | (219)
g
e

B
J
d
dM
_
J + 1
3
g
e

B
M
k
B
T
C
_
= 1 (220)
g
2
e

2
B
J(J + 1)
3k
B
T
C
= 1 (221)
T
C
=
g
2
e

2
B
J(J + 1)
3k
B
(222)
Next, we work out the behaviour of M near T
C
. We recall the expression of magnetization M.
M = g
e

B
JB
J
_
g
e

B
M
k
B
T
J
_
(223)
| substitute
g
e

B
J
k
B
=
3T
C
g
e

B
(J + 1)
(224)
= g
e

B
JB
J
_
M
T
3T
C
g
e

B
(J + 1)
_
(225)
| since M is small near T
C
, we can expand B
J
as a Taylor series (226)
| this time we need to retain one more order: coth y =
1
y
+
y
3

y
3
45
+ (227)
| we already know the rst 2 terms give B
J
(xJ)
J + 1
3
x (228)
= g
e

B
J
_
J + 1
3J
M
T
3T
C
g
e

B
(J + 1)

2J + 1
2J
_
2J + 1
2J
_
3
1
45
_
M
T
3T
C
g
e

B
(J + 1)
_
3
(229)
+
1
2J
_
1
2J
_
3
1
45
_
M
T
3T
C
g
e

B
(J + 1)
_
3
_
(230)
| remove one common factor of M (231)
1 =
T
C
T
_
1 +
1 (2J + 1)
4
45(2J)
4
_
T
C
T
_
2
27M
2
J
g
2
e

2
B
(J + 1)
3
_
(232)
T
T
C
1 =
1 (2J + 1)
4
45(2J)
4
27J
g
2
e

2
B
(J + 1)
3
_
T
C
T
_
2
M
2
(233)
T
T
C
1 =
3(2J
2
+ 2J + 1)
10g
2
e

2
B
J
2
(J + 1)
2
_
T
C
T
_
2
M
2
(234)
M
2
=
10g
2
e

2
B
J
2
(J + 1)
2
3(2J
2
+ 2J + 1)
_
1
T
T
C
__
T
T
C
_
2
(235)
21
M
2

_
T
T
C
_
2
_
1
T
T
C
_
(236)
| for T T
C
, we can set
_
T
T
C
_
2
1 (237)
M
_
1
T
T
C
(238)
Figure 9: The relationship between magnetization M near T
C
as given by Weiss mean eld theory.
Finally we can discuss the susceptibility of the ferromagnet in Weiss Mean eld theory. We need to
have a (small) magnetic eld in the system. Recall the magnetization expression with an external
eld.
M = g
e

B
JB
J
(xJ) (239)
| with modication x =
g
e

B
(B +M)
k
B
T
and write
g
e

B
k
B
=
3T
C
g
e

B
J(J + 1)
(240)
= g
e

B
JB
J
_
3T
C
g
e

B
J(J + 1)
B +M
T
J
_
(241)
| at T > T
C
, the external eld causes a small magnetization (242)
| thus recall the expansion B
J
(xJ)
J + 1
3
x =
J + 1
3J
xJ (243)
g
e

B
J
_
J + 1
3J
3T
C
g
e

B
J(J + 1)
B +M
T
J
_
(244)
=
T
C

B +M
T
(245)
M =
T
C
T
B

_
1
T
C
T
_ (246)

0
M
B
= =

0
T
C
T

_
1
T
C
T
_ (247)
=

0
T
C

1
T T
C
(248)
The expression
1
TT
C
is the famous Curie-Weiss Law. Do not confuse this with the Curies Law
of paramagnetism.
A few comments are now in order:
22
Recall that Weiss mean eld theory is based on the assumption that there is some internal eld
creating interactions between the magnetic moments. Weiss had no idea of the exchange interaction!
Thus it is already remarkable that it contains the phase transition between paramagnetism and
ferromagnetism. Measuring T
C
will give a t for .
Further analysis shows the failures of Weiss theory:
For T 0, the expression of magnetization from Weiss theory and that from experiments do
not agree.
The T T
C
expression obtained earlier: M
_
1
T
T
C
also does not agree with the experi-
mental expression: M
_
1
T
T
C
_
1/3
.
The specic heat for T > T
C
predicted by Weiss theory is zero and it is completely incorrect.
Experimental results seem to indicate a logarithmic singularity as T T
C
for the specic
heat. Weiss theory could not even be used to provide any explanation for that.
Weiss Mean eld theory can be derived from Heisenberg Hamiltonian by the following argument:
Consider the Heisenberg Hamiltonian with nearest neighbour (nn) and next nearest neighbour
(nnn) interactions, H
Heisenberg
= J

S
i

S
i+1
J

S
i

S
i+2
. Then we take average of the nn
and nnn spins to get H
Heisenberg
J

S
i

S
i+1
J

S
i

S
i+2
and the average spin

S
is proportional to the magnetization. Thus, H
Heisenberg
and H
Weiss
have a similar form.
6.2.2 Magnons (Quantized Spin Waves)
At low temperatures, a low energy excitation called a spin wave propagates among the interacting spins.
Quantized spin waves are called magnons. This is extremely similar to phonons which are quantized
lattice waves.
We avoid any semiclassical derivations as this is really a quantum mechanical eect. This is described
by the Heisenberg model and we shall give a simplied derivation of spin waves.
We consider a 1D chain of N interacting spins
2
. Thus each spin has 2 nearest neighbours except for
the rst and last spin. The Heisenberg Hamiltonian can be written as
H
Heisenberg
=
N1

i=1
J
i,i+1

S
i

S
i+1
(249)
| take J
i,i+1
= J
i+1,i
= J (250)
= J
N1

i=1

S
i

S
i+1
(251)
= J
N1

i=1
S
z
i
S
z
i+1
+S
x
i
S
x
i+1
+S
y
i
S
y
i+1
(252)
| write S
x
and S
y
in terms of raising and lowering operators (253)
| thus recall S
x
i
=
1
2
(S
+
i
+S

i
) and S
y
i
=
1
2i
(S
+
i
S

i
) (254)
= J
N1

i=1
S
z
i
S
z
i+1
+
1
2
(S
+
i
+S

i
)
1
2
(S
+
i+1
+S

i+1
) +
1
2i
(S
+
i
S

i
)
1
2i
(S
+
i+1
S

i+1
)(255)
= J
N1

i=1
S
z
i
S
z
i+1
+
1
2
(S
+
i
S

i+1
+S

i
S
+
i+1
) (256)
2
Strictly speaking, we require translational invariance in this derivation. We can achieve this by either taking N
or by joining the head and tail of the spin chain together (Periodic Boundary condition). We neglect this detail in this
derivation and a better derivation will be done in the tutorial.
23
Now we specialise to spin-
1
2
for a simplied discussion thus there are only 2 states: m
s
=
1
2
up and
m
s
=
1
2
down. The T = 0 ground state of the ferromagnet can be written as
|0 = |
1

N
(257)
The lowest excited state is one spin ip at site j.
|1
j
= |
1

j1

j+1

N
(258)
The total spin of the system changed by 1 (hence the notation). Thus this is a bosonic excitation.
We want to nd the energy eigenvalue of such 1-spin-ip states.
First consider,
H
Heisenberg
|0 =
_
J
N1

i=1
S
z
i
S
z
i+1
+
1
2
(S
+
i
S

i+1
) +S

i
S
+
i+1
_
|
1

N
(259)
| note that S
+
i
|
i
= 0 and S
z
i
|
i
=
1
2
|
i
(260)
= J
N1

i=1
S
z
i
S
z
i+1
|
1

N
(261)
= J
N 1
4
|0 (262)
And also,
H
Heisenberg
|1
j
=
_
J
N1

i=1
S
z
i
S
z
i+1
+
1
2
(S
+
i
S

i+1
+S

i
S
+
i+1
)
_
|
1

j

N
(263)
| note the only relation we need is S
+
j
|
j
= |
j
(264)
= J
N1

i=1
_
S
z
i
S
z
i+1

1
2
(|
1

j

j+1

N
+|
1

j1

j

N
)
_
(265)
= J
_
(S
z
1
S
z
2
+ +S
z
j1
S
z
j
+S
z
j
S
z
j+1
+ +S
z
N1
S
z
N
)|1
j
+
1
2
(|1
j+1
+|1
j1
)
_
| note that S
z
j
|
1

j

N
=
1
2
|
1

j

N
(266)
= J
_
N 3
4

1
4

1
4
_
|1
j

1
2
J(|1
j+1
+|1
j1
) (267)
= J
N 5
4
|1
j

1
2
J(|1
j+1
+|1
j1
) (268)
Now consider the plane wave state |q =
1

j
e
iqR
j
|1
j
(this is inspired from phonons)
H
Heisenberg
|q = H
Heisenberg
1

j
e
iqR
j
|1
j
(269)
| where R
j
is the distance from the rst spin to j-th spin (270)
| let a be the spacing between adjacent spins, so R
j
= ja (271)
=
1

j
e
iqR
j
_
J
N 5
4
|1
j

1
2
J(|1
j+1
+|1
j1
)
_
(272)
=
1

j
_
J
N 5
4
e
iqR
j
|1
j

1
2
J(e
iqR
j
e
iqa
e
iqa
|1
j+1
+e
iqR
j
e
iqa
e
iqa
|1
j1
)
_
24
| note R
j
+a = R
j+1
, R
j
a = R
j1
and
1

j
e
iqR
j1
|1
j1
= |q (273)
| the last relation requires translational invariance which we ignore here (274)
=
_
J
N 5
4

1
2
J(e
iqa
+e
iqa
)
_
|q (275)
=
_
J
N 1
4
+J(1 cos(qa))
_
|q (276)
Thus |q is an eigenstate of H
Heisenberg
. The state |q is a sum of all 1-spin ip states |1
j
with phase
factors e
iqR
j
, thus |q is a plane wave of 1-spin ips.
If we shift up the Heisenberg Hamiltonian by J
N1
4
, we have,
H

Heisenberg
= H
Heisenberg
+J
N 1
4
(277)
then, H

Heisenberg
|0 = 0|0 (278)
H

Heisenberg
|q = J(1 cos(qa))|q (279)
Thus the dispersion of a plane wave of 1-spin ips or spin wave is E(q) = J(1 cos(qa)). (See
gure)
We will not carry on to quantize the spin waves into magnons as this will involve second quantization
notation and Holstein-Primako transformation which is beyond the scope of the module.
Figure 10: The dispersion relation of Ferromagnetic spin waves
Figure 11: The semiclassical picture of spin waves. Do not take it too seriously.
6.2.3 Ferromagnetic Domains
[Why do domains form?] There are 4 types of energy competing in the ferromagnet. This results
in why the ferromagnet is not made of a single domain of aligned spins but domains of aligned spins in
dierent directions. Such a structure of domains actually results in a lower energy for the material.
25
1. Exchange Energy:
This short-ranged interaction was modelled by the Heisenberg Hamiltonian H = J

S
i

S
j
and
parallel alignment lowers the energy.
2. Magnetostatic Energy:
The simplest way to describe this long-ranged energy is to think that this energy is used to main-
tain the external magnetic eld of the magnetized material. If we break the material, making 2
dipoles, and anti-align them, the external eld (and thus the energy) will be greatly reduced.
3
This favours anti-alignment and the formation of domains with anti-aligned moments.
Figure 12: The formation of domains (from left to right) resulting in a reduction of magnetostatic energy.
The rightmost diagram shows closure domains.
3. Anisotropy Energy:
If we only consider the rst 2 types of energies, then the most favourable condition is that the
magnetic moments gradually ip over from one end of the material to the other end, i.e. the width
of the domain wall is the width of the material! Due to the crystalline structure of the material,
there are certain directions that cost less energy for the magnetic moments to align with (easy
axes). The hard axes are dened oppositely. The width of the domain would be a balance of
exchange and anisotropy energies as most of the rotated spins would not be along the easy axis
and this costs anisotropy energy.
4. Magnetoelastic energy:
Change in magnetization results in (small) elastic distortions of the material. The larger the
domains, the more elastic energy is needed. Thus the formation of smaller domains (such as
closure domains) is favoured by magnetoelastic energy.
[Types of Domain Walls]
Bloch Wall: magnetization rotates in a plane parallel to the plane of the domain wall. See Figure
13(a).
Neel Wall: magnetization rotates in a plane perpendicular to the plane of the domain wall. See
Figure 13(b).
90
o
or 180
o
change in magnetic moments between adjacent domains. See Figure 14.
3
One can also think of the fact that a broken magnet will also tend to anti-align.
26
Figure 13: (a) Bloch Wall and (b) Neel Wall.
Figure 14: (a) 180
0
domain wall and (b) 90
0
domain wall.
[Estimation of a 180
o
Blochs Walls energy:] Assume the change in moments occurs over N
sites and thus each spin rotates

N
radians with respect to its neighbour. The exchange energy E
ex
is
estimated using the Heisenberg Hamiltonian.
H
Heisenberg
= J

S
i

S
i+1
(280)
roughly speaking
= E = J

i
S
i
S
i+1
cos (281)
= JNS
2
cos
_

N
_
(282)
| assume the angle is small and expand cos x 1
x
2
2
(283)
= JNS
2
+JNS
2
_

N
_
2

1
2
(284)
The rst term is for parallel alignment. The exchange energy is thus the second term
E
ex

1
2
JNS
2
_

N
_
2
(285)
For the anisotropy energy density, we assume a simple form.
E = K sin
2
K > 0 (286)
where K is an anisotropy constant and this model would mean that each spin prefers = 0 or = .
Thus the total anisotropy energy is
E
ani
=
N

i=1
K sin
2

i
(287)
| assume a large number of spins and take continuum limit (288)
=
N

_

0
K sin
2
d (289)
| the factor
N

is needed since =

N
(290)
=
N

K
_

0
1
2

1
2
cos 2d (291)
27
=
N

2
(292)
=
NK
2
(293)
Total energy in the domain wall:
E
total
= E
ex
+E
ani
=
JS
2

2
2N
+
NK
2
(294)
Thus this expression is in line with the earlier discussion that exchange energy favours a wide domain
wall N. We minimise E
total
with respect to N,
dE
total
dN
= 0 =
JS
2

2
2N
2
+
K
2
= 0 (295)
N = S
_
J
K
(296)
Thus the width of the domain wall depends on the ratio between the exchange constant and the
anisotropy constant.
[Magnetization and Hysteresis] The most obvious consequence is that the Curie temperature of
iron T
C
1000K so why isnt iron magnetized at room temperature? The answer is that each domain
is magnetized in one direction and these directions are quite random so there is no net magnetization.
The other consequence is that for soft magnetic materials, it only requires a small magnetic eld
(as low as 10
6
T) to attain saturation magnetization. The short answer is that since there is already
alignment within a domain, only a small magnetic eld is needed to align the domains.
We now consider the process of magnetization of a ferromagnet with domains.
Figure 15: The so-called hysteresis loop. The 7 parts in the hysteresis loop are discussed in the text. M
s
- saturation magnetization, M
r
- remanent magnetization and H
c
- coercive eld.
1. We assume there is no net magnetization in the material to begin with. Magnetization M increases
with applied eld H. There is a linear region where the process is reversible: removing the eld
28
results in no magnetization. Increasing the eld beyond the linear region brings the magnetization
to a maximum saturated value M
s
. Microscopically, domains aligned parallel to the eld will grow
in size and this is reversible for small domain wall motion. Saturation magnetization is reached
when domains that are aligned to H are at their largest size.
2. From M
s
, when the external eld is decreased to zero, there is a remanent magnetization M
r
.
Microscopically, domains that are aligned to H decrease in size when H is decreased. However,
domain walls may get pinned by strains and impurities resulting in net magnetization even when
H = 0.
3. An opposite eld is applied to kill M
r
. That amount of opposite eld is called H
c
, the coercive
eld. Microscopically, H
c
forces the pinned domain walls to move so that net magnetization can
be zero. The size of H
c
thus depends on the material preparation and properties.
4. A large enough opposite eld will cause an opposite saturation magnatization.
5. This is the repeat of process 2.
6. This is the repeat of process 3.
7. This is the repeat of process 4.
6.3 Antiferromagnetic and Ferrimagnetic Order
Antiferromagnetism and ferrimagnetism are simply cases where adjacent spins prefer to anti-align with
each other. In antiferromagnetism, the antiparallel moments are of the same magnitude and they cancel
to give no net magnetization. In ferrimagnetism, the antiparallel moments are unequal in magnitude
and this results in a net magnetization.
We will discuss Weiss mean eld theory for antiferromagnetism. Weiss mean eld theory for ferri-
magnetism is set as a tutorial problem.
We should also note that the Heisenberg model can be modied to describe antiferromagnetism and
ferrimagnetism. To favour anti-alignment, the exchange energy simply needs to be negative.
H
Heisenberg
= J

i=1

S
i

S
i+1
= |J|

i=1

S
i

S
i+1
J < 0 (297)
6.3.1 Weiss Mean Field Theory of Antiferromagnetism
We assume only nearest neighbour (nn) interactions so we have 2 kinds of eective elds.
B

= ||M

+B and B

= ||M

+B (298)
where
Term ||M

is the mean eld as seen by the up-spins. Since we assume nn interactions, up-spins
sees a down-spin magnetization. The opposite case is for the term ||M

. We also assumed
that there is only one mean eld constant . is negative because the exchange interaction is
negative.
Term B is the external eld.
The Hamiltonian is then
H
Weiss
=
Z

i=1
g
e

B

J
i

+
Z

i=1
g
e

B

J
i

(299)
29
Thus we can again carry over the Curie Paramagnetism expression, (set B = 0)
M

= g
e

B
JB
J
_

g
e

B
||M

k
B
T
J
_
(300)
M

= g
e

B
JB
J
_

g
e

B
||M

k
B
T
J
_
(301)
But in an antiferromagnet below T
C
(and B = 0), M

= M

, so we write |M

| = |M

| = M. We
get 2 equations that are exactly the same
M = g
e

B
JB
J
_
g
e

B
||M
k
B
T
J
_
(302)
Since this expression is so similar to the one in ferromagnetism, we simply carry over the results from
ferromagnetism and replace with ||. First the transition temperature,
4
Neel Temperature: T
N
=
g
2
e

2
B
||J(J + 1)
3k
B
(303)
Next, the susceptibility for T > T
N
, (put back the external eld B)
M

= g
e

B
JB
J
_
g
e

B
(B ||M

)
k
B
T
J
_
(304)
| write
g
e

B
k
B
=
3T
N
g
e

B
||J(J + 1)
and expand B
J
(xJ)
J + 1
3J
xJ (305)
g
e

B
J
_
J + 1
3J
3T
N
g
e

B
||J(J + 1)
B ||M

T
J
_
(306)
=
T
N
||
B ||M

T
(307)
Similarly, the other equation becomes
M

T
N
||
B ||M

T
(308)
Solving these 2 equations give,
M

= M

=
BT
N
||
1
T +T
N
(309)
= =

0
M
B
=

0
(M

+M

)
B
=
2
0
T
N
||
1
T +T
N
(310)
A more realistic calculation will be to include mean elds from next nearest neighbours (nnn) but it
will not be carried out here.
4
Neel temperature ts the experimental values badly actually.
30
7 Tutorial Problems
1. Numerical comparison of Diamagnetism and Paramagnetism
Estimate the diamagnetic orbital susceptibility of a gas of hydrogen atoms (with number density
1 10
20
m
3
) in ground state (n = 0), and compare this with the paramagnetic spin susceptibility
at 100K.
2. Exchange Interaction is not magnetic dipole-dipole interaction
Estimate the ratio of the exchange and dipolar coupling of two adjacent Fe atoms in metallic Fe.
(The exchange constant in Fe can be crudely estimated by setting it equal to k
B
T
C
where T
C
is
the Curie temperature. For Fe, T
C
= 1043K.) The dipolar energy is the usual expression from
electromagnetism.
E =

0
4r
3
_

1

2

3
r
2
(
1
r)(
2
r)
_
(311)
3. Corrected Derivation of Ferromagnetic Spin Waves
By imposing periodic boundary condition on the chain
1

N
, we need to add one term to
H
Heisenberg
. Show that this modication gives a more satisfying derivation of ferromagnetic spin
wave dispersion, i.e. we get
H
Heisenberg
|0 =
JN
4
, H
Heisenberg
|q =
_

JN
4
+J(1 cos(qa))
_
|q (312)
4. Ground state of an antiferromagnet is a dicult problem
Consider the system of 4 spin-
1
2
particles with antiferromagnetic arrangement |0 = |
1

4
.
This is commonly called the Neel state. The theoretical model is the antiferromagnetic Heisenberg
Hamiltonian H
Heisenberg
= |J|

4
i=1

S
i

S
i+1
. Use periodic boundary conditions where

S
5
=

S
1
.
(a) Show that |0 is not an eignestate of H
Heisenberg
. Then show that the expectation value
0|H
Heisenberg
|0 = |J|.
(b) The (unnormalised) ground state state is
|0

= (|
1

2
|
1

2
) (|
3

4
|
3

4
) (313)
= 2|
1

4
|
1

4
|
1

4
+ 2|
1

4
|
1

4
|
1

(314)
Check that indeed this is an eigenstate with eigenvalue 2|J|, i.e. H
Heisenberg
|0

= 2|J||0

. Do
not tackle the 16-dimensional problem directly.
5. Weiss Mean Field theory for Ferrimagnetism
Consider the following model of ferrimagnetism:
Here, we include next neighbours (nnn) interactions. Denote as the mean eld constant associated
with exchange constant J

= J

and

) as the mean eld constant associated with exchange

constant J

(J

).
(a) Write down the 2 eective eld including external eld B.
(b) Write down the Hamiltonian and magnetization expressions M

, M

. Note that the Lande

g-factor and angular momentum J of the 2 sublattices may not be the same.
(c) For T < T
C
, and no external eld,
31
i. Solve for the critical transition temperature T
C
by requiring non-trivial solution for mag-
netizations M

, M

.
ii. Give a reason why the positive term is chosen to be T
C
.
(d) For T > T
C
and a small external eld,
i. Solve for the susceptibility =

0
M
B
=

0
B
(M

+M

).
(e) Treat some special cases: (in each case write down T
C
and )
i. Ferrimagnetism with nn mean eld.
ii. Antiferromagnetism with nnn mean eld.
iii. Antiferromagnetism with nn mean eld.
iv. Ferromagnetism with nn mean eld.
32