Chapter 1

Introduction to Magnetism

Magnetic properties originate in the spin degrees of freedom of electrons and their
associated motion in solids. We first describe the microscopic magnetic moments of
electrons, and the formation of atomic magnetic moments due to strong Coulomb
interactions in an atom with unfilled shell. Atomic magnetic moments change their
nature when the atoms form a solid. The key to understanding the behavior of mag-
netic moments in solids is the degree of electron localization. We briefly introduce
the concept of the metal and the insulator (i.e., the Mott insulator). Electrons in the
latter are localized on each atom, so that their magnetic properties are described
by the atomic magnetic moments and the magnetic interactions between them. We
present in this chapter only a minimal discussion on the magnetism in insulators.
On the other hand, electrons move from site to site over the entire crystal in metals,
so that the magnetic properties are connected to the whole degrees of freedom of
correlated electrons. Needless to say, the main theme of this book is the magnetism
of metals and alloys. In the last section, we elucidate various magnetic structures in
solids to provide a basic knowledge on magnetism.

1.1 Magnetic Moments

Magnetic moment M in the electromagnetics is defined by the torque N on a magnet
under the magnetic field H as

N ≡ M × H. (1.1)

It originates in electrons in the magnet. According to the classical electromagnetics,

the coil-type local current i(r) due to electron motion causes a magnetic moment.
It is given in the CGS-Gauss unit as [1]

1
ML = r × i(r) d 3 x. (1.2)
2c

Y. Kakehashi, Modern Theory of Magnetism in Metals and Alloys, 1

Springer Series in Solid-State Sciences 175, DOI 10.1007/978-3-642-33401-6_1,
© Springer-Verlag Berlin Heidelberg 2012
2 1 Introduction to Magnetism

Here c is the speed of light. i(r) is the current density of electrons, given by i(r) =

− i ev i δ(r − r i ) with v i being the velocity of the electron and −e the charge of
the electron at r i . Substituting the expressions into (1.2), we obtain
 e
ML = − li . (1.3)
2me c
i

Here me is the mass of an electron and l i denotes the angular momentum of elec-
tron i. Equation (1.3) indicates that an electron with angular momentum l generates
a magnetic moment,

e
ml = − l. (1.4)
2me c

This is called the orbital magnetic moment for an electron.

According to the quantum mechanics, an electron has its own magnetic moment
called the spin magnetic moment [2]. It is given by

e
ms = −ge s. (1.5)
2me c

Here s is the spin angular momentum of an electron with spin s = 1/2. The constant
ge = 2.0023 is referred as the g-value of electron. The deviation from 2 is caused by
the interaction with electro-magnetic fields. The spin magnetic moment was found
first experimentally by Stern and Gerlach in 1922, and was established theoretically
in 1928 by Dirac in his relativistic theory of electrons [3]. In the following we as-
sume that ge = 2 for simplicity.
As seen from (1.5), the spin of an electron leads to the magnetic moment

e
μB = = 0.9274 × 10−20 emu. (1.6)
2me c

Here  is the Planck constant divided by 2π . The magnetic moment μB denotes the
Bohr magneton, and is often used as a unit of the magnetic moment in atomic scale.
Equations (1.4) and (1.5) indicate that an electron has the following magnetic
moment in the atomic scale.

m = ml + ms = −(l + 2s)μB , (1.7)

and the total magnetic moment is given by

 
M =− (l i + 2s i )μB . (1.8)
i
1.2 Basic Hamiltonian 3

Here the angular momenta l and s are measured in units of  = 1. The l i and s i
at the right-hand-side (r.h.s.) of the above equation should be regarded as quantum
mechanical operators, and ∼ means a quantum mechanical expectation value. In
the following, we omit the minus sign at the r.h.s. of (1.8) for convenience bearing
in mind that the real magnetic moments are opposite in direction.
It should be noted that the nuclei in the magnet also have magnetic moments
because they have their own spins. The size of their magnetic moments is however
characterized by the nuclear magneton μN , which is defined by

e
μN = . (1.9)
2mp c

Here mp denotes the mass of proton. The nuclear magneton μN is only 1/1800 of the
Bohr magneton μB because μN /μB = me /mp . Therefore the magnetism in solids
is dominated by the magnetic moments of electrons. In the following equations, we
adopt a unit of μB = 1 for simplicity.

1.2 Basic Hamiltonian

Magnetic moments and magnetic properties are governed by the Hamiltonian of
the system. The basic Hamiltonian of electrons in solids is given as follows in the
second quantization.
  
1
H= ψ † (r) − ∇ 2 + vN (r) ψ(r) dr
2

1 1
+ ψ † (r)ψ † r  ψ r  ψ(r) dr dr 
2 |r − r  |

1 1
+ ψ † (r) ∇vN (r) × (−i∇) · σ ψ(r) dr
2 2

− ψ † (r)(l + 2s)ψ(r) dr · H . (1.10)

Here we have adopted the units me = 1,  = 1, and e = 1 for simplicity. ψ(r) =

(ψ↑ (r), ψ↓ (r)) is the electron field operator. 
vN (r) is the one electron potential en-
ergy due to nuclei, and is given by vN (r) = − i Zi /|r − R i |. Zi is the atomic num-
ber of the atom at position R i . H denotes the magnetic field, and σ = (σx , σy , σz )
denote the Pauli spin matrices given by
     
0 1 0 −i 1 0
σx = , σy = , σz = . (1.11)
1 0 i 0 0 −1

The first term in the Hamiltonian (1.10) consists of the kinetic energy and the
attractive potential due to nuclei, and describes the independent motion of electrons.
4 1 Introduction to Magnetism

The second term denotes the electron–electron interaction, the third term denotes the
spin–orbit interaction, and the last term is the Zeeman interaction due to magnetic
field. Note that in the above expression we have omitted the diamagnetic term which
is proportional to the square of the magnetic field [2], because we do not consider
the diamagnetism in this book.
In solids, we may express the field operators by means of a basis set of wave-
functions {ϕi (r)} as follows.

ψσ (r) = aiσ ϕi (r). (1.12)
i

Here ϕi (r) are orthonormal basis functions. The suffix i stands for a set of the
atomic position i (quantum number n) and orbital L (momentum k) when we adopt
atomic orbitals (one-electron eigen functions in solids) as the basis functions.
Using the orthonormal basis set, we can express the Hamiltonian (1.10) as fol-
lows.
 † 1  † †
H= εij aiσ aj σ + Vij kl aiσ aj σ  akσ  alσ + HSO + HZeeman . (1.13)
2 
ij σ ij klσ σ

Here εij (Vij kl ) are the matrix elements for the independent electron (the Coulomb
interaction) part of the Hamiltonian (1.10). They are given by
  
∗ 1 2
εij = dr ϕi (r) − ∇ + vN (r) ϕj (r), (1.14)
2
 ϕi∗ (r)ϕj∗ (r  )ϕk (r  )ϕl (r)
Vij kl = dr dr  . (1.15)
|r − r  |
The third and last terms at the r.h.s. of (1.13) are the spin–orbit interaction term
and the Zeeman term, respectively.
 †
HSO = ζiαj γ aiα aj γ , (1.16)
iαj γ
 †
HZeeman = − (l + 2s)iαj γ aiα aj γ · H . (1.17)
iαj γ

Here the spin–orbit interaction matrix elements ζiαj γ are given by


1
ζiαj γ = dr ϕi∗ (r) ∇v(r) × (−i∇) ϕj (r) · (s)αγ . (1.18)
2
In molecules and solids, atomic orbitals are often useful as the basis functions.
(atom)
ϕi (r) = φinl |r − R i | Ylm (r − R i ). (1.19)

Here φinl (|r −R i |) is a radial wave function for an atom located at R i , the subscripts
n and l denote the principal quantum number and the azimuthal quantum number,
1.2 Basic Hamiltonian 5

respectively. Ylm (r − R i ) is the spherical harmonics located at atom R i , and m

denotes the magnetic quantum number.
In the application to solids, it is convenient to make use of real functions {Plν }
instead of complex basis functions {Ylm }. The former called the cubic harmonics
are constructed by linear combinations of spherical harmonics. Defining the orbital
L = (l, ν), the atomic functions for solids are written as follows.

ϕinL (r) = φinl |r − R i | Plν (r − R i ). (1.20)

The cubic harmonics are constructed to yield the irreducible representation for
the cubic symmetry of the point group in crystal. The s (l = 0) function belonging
to the A1g representation for the cubic point symmetry is a constant function.

1
Pa1g (r) = Y00 = √ . (1.21)
4π

The p (l = 1) functions belonging to the T1u representation are defined as follows.

1 3 x
Pt1u α (r) = √ (−Y11 + Y1,−1 ) = , (1.22)
2 4π r
i 3 y
Pt1u β (r) = √ (Y11 + Y1,−1 ) = , (1.23)
2 4π r
3 z
Pt1u γ (r) = Y10 = . (1.24)
4π r
The d (l = 2) orbitals form the Eg (dγ ) and T2g (dε) representations. The d
functions belonging to the Eg representation are given by

5 1 3z2 − r 2
Pegu (r) = Y20 = , (1.25)
4π 2 r2
√
1 5 3 x2 − y2
Pegv (r) = √ (Y22 + Y2,−2 ) = . (1.26)
2 4π 2 r2

The remaining d (l = 2) functions belonging to the T2g representation are given by

i 5 √ yz
Pt2gξ (r) = √ (Y21 + Y2,−1 ) = 3 2, (1.27)
2 4π r
1 5 √ zx
Pt2gη (r) = √ (−Y21 + Y2,−1 ) = 3 2, (1.28)
2 4π r
i 5 √ xy
Pt2gζ (r) = √ (−Y22 + Y2,−2 ) = 3 2. (1.29)
2 4π r
6 1 Introduction to Magnetism

Fig. 1.1 Atomic orbitals from s to d symmetry. The sign of wave functions is shown by + and −
in the figure

The angular dependence of the cubic harmonics is shown in Fig. 1.1. The spatial
distribution of the atomic wave functions governs the electron hoppings in solids,
thus determining the electronic and magnetic properties of solids.

1.3 Formation of Atomic Moments

The electronic structure and related properties of an atom may be understood by an

independent-electron picture. The spin–orbit interaction in the Hamiltonian (1.13) is
smaller than the Coulomb interaction in the ‘light’ elements such as the 3d transition
metals. We first neglect the former, and treat the latter by means of the following
approximation.
† †  †   
aiσ aj σ  akσ  alσ = aiσ alσ aj†σ  akσ  + aiσ†
alσ aj†σ  akσ 
 †   
− aiσ akσ  aj†σ  alσ − aiσ†
akσ  aj†σ  alσ
 † † 
− aiσ aj σ  akσ  alσ . (1.30)

Here the average   is taken with respect to an independent particle system which
will be chosen later. Note that the two particle operator has been decoupled so that
(1.30) exactly holds true when we take the average. This is called the Hartree–Fock
approximation.
In the Hartree–Fock approximation, the Hamiltonian (1.13) is reduced to
 † 1   †  
H= hij σ aiσ aj σ − (Vij kl − Vij lk δσ σ  ) aiσ alσ aj†σ  akσ  . (1.31)
2 
ij σ ij klσ σ
1.3 Formation of Atomic Moments 7

Here hij σ is given as follows.

   
1 n(r  )
hij σ = ϕiσ †
(r) − ∇ 2 + vN (r) + dr  ϕj σ (r)
2 |r − r  |

ψ † (r  )ψσ (r)
− dr dr  ϕiσ†
(r) σ ϕj σ r  . (1.32)
|r − r  |
The Hartree–Fock one-electron wave functions {ϕiσ (r)} are determined by solv-
ing the following self-consistent equations.
   
1 2  n(r )
− ∇ + vN (r) + dr ϕiσ (r)
2 |r − r  |
  ϕj∗σ (r  )ϕiσ (r  )
− nj σ  dr  ϕj σ (r) = εiσ ϕiσ (r). (1.33)
|r − r  |
j

Here εiσ is the Hartree–Fock energy eigen value. When we apply the Hartree–Fock
independent particle Hamiltonian in the  average  , nj σ  is given by the Fermi dis-
tribution function f (εj σ ) and n(r) = j σ f (εj σ )ϕj∗σ (r)ϕj σ (r) denotes a charge
density in the Hartree–Fock approximation. The third term at the l.h.s. of (1.33)
gives the electrostatic potential due to electrons. The last term also originates in
the Coulomb interaction, but the wave functions have been exchanged due to the
anti-symmetric property of the Slater determinant. It is referred as the exchange po-
tential. The Hartree–Fock wave function is the best Slater determinant at the ground
state according to the variational principle [4].
Taking the Hartree–Fock wavefunctions as the basis functions, one can express
the Hartree–Fock Hamiltonian (1.31) as follows.
 1 
H= εiσ niσ − (Vijj i − Vij ij δσ σ  )niσ nj σ  . (1.34)
2 
iσ ij σ σ

Here Vijj i (Vij ij ) is known as the Coulomb integral (exchange integral). (Note that
Vij kl should more precisely be Viσj σ  kσ  lσ .) The second term at the r.h.s. is to elim-
inate the double counting of Coulomb interaction in the Hartree–Fock one-electron
energy.
The exchange potential in the Hartree–Fock equation (1.33) is nonlocal. Slater
proposed an approximate local exchange potential, using the free-electron wave
functions. It is given as follows [4].
 1/3
3
vexσ (r) = −3 nσ (r)1/3 , (1.35)
4π

where nσ (r) is the electron density with spin σ .

The effective potential in the Hartree–Fock self-consistent equation is spherical
in the atomic system when we apply the potential (1.35). We can express the atomic
8 1 Introduction to Magnetism

wave function for an electron as

ϕnlmσ (r) = φnlσ (r)Ylm (r̂). (1.36)

Here φnlσ (r) is the radial wave function, and Ylm (r̂) is the spherical harmonics. n,
l, and m denote the principal quantum number, the azimuthal quantum number, and
the magnetic quantum number, respectively. Orbitals l = 0, 1, 2, 3, . . . are called s,
p, d, f, . . . , respectively. The nl shells are therefore written as ns, np, nd, nf, etc. In
each shell, there are 2l + 1 degenerate orbitals.
With the use of the Hartree–Fock atomic orbitals, a many electron state of an
atom is expressed in the form of the Slater determinant. The ground-state electronic
structure of an atom is constructed according to the Pauli principle. For example,
in the case of Ar, we have the ground state 1s2 2s2 2p6 3s2 3p6 . Since the outermost
electron shell of Ar is closed, the magnetic moment of Ar vanishes.
For an atom with an unfilled shell, the magnetic moment may appear. Since the
total number of electrons (N ), the total spin (S), and the total orbital moment (L) of
an atom commute with the Hamiltonian, the eigen function Ψ should be specified by
the set (NLMSMs ): Ψ (N LMSMs ), where L (S) and M (Ms ) denote the magnitude
and z component of orbital moment L (spin S), respectively. The eigen energy,
on the other hand, should depend only on N , L, and S because of the rotational
symmetry: EA (NLS).
The ground state of electrons in an atom should be obtained by minimizing the
Coulomb energy. The Coulomb interaction of the Hamiltonian for the unfilled shell
is written as follows according to (1.13).
  1
 
HCoulomb = Uνν nν↑ nν↓ + Uνν  − Jνν  nν nν  − 2 Jνν  s ν · s ν  .
ν 
2 
ν>ν ν>ν
(1.37)

Here ν denotes an orbital lm (m = −l . . . l). nν (s ν ) denotes the charge (spin) density

operator for electrons in the orbital ν. We have taken into account the Hartree–Fock
Coulomb and exchange interactions, Uνν  = Vνν  ν  ν and Jνν  = Vνν  νν  , and omitted
the other types of interactions. The first term at the r.h.s. of (1.37) denotes the intra-
orbital Coulomb interaction, the second term expresses the inter-orbital interaction,
and the third term the exchange interaction between the spins on different orbitals.
Note that Uνν > Uνν  (ν = ν  ) > Jνν  > 0. Typical values of Coulomb interactions
are Uνν ≈ 20 eV, Uνν − Uνν  ≈ 2 eV, and Jνν  ≈ 0.9 eV for 3d transition-metal
elements [5]. Note that N stands for the total number of electrons in the unfilled
shell when the Hamiltonian (1.37) is applied.
The intra-orbital Coulomb interaction (i.e., the first term at the r.h.s. of (1.37))
acts to reduce the doubly occupied states on the same orbital so that it creates active
spins on the orbitals of an unfilled shell. The third term in (1.37) aligns the active
spins on the orbitals because Jνν  > 0. These effects suggest that the magnitude of
the total spin S is maximized at the ground state. This is known as Hund’s first rule,
and therefore Jνν  are often called the Hund-rule couplings. On the other hand, the
1.3 Formation of Atomic Moments 9

second term in (1.37) is the energy associated with the configuration of electrons
on different orbitals. One might expect that such an interaction energy is reduced
when electrons move around the nucleus in the same direction avoiding each other.
Thus we expect that the magnitude of the total angular momentum L should be
maximized at the ground state under the maximum magnitude of total spin. This is
referred as Hund’s second rule. The first and second Hund rules are verified by the
full Hartree–Fock numerical calculations.
The maximum S and L at the ground state are obtained by using the Pauli
principle and the Hund rule, especially from the conditions S = Ms (Max) and
L = M(Max) under the maximum S that corresponds to the state Ψ (N LLSS):
⎧N
⎪
⎨ for N ≤ 2l + 1,
S= 2 (1.38)
⎩ 4l + 2 − N for N > 2l + 1,
⎪
2

1
N (2l + 1 − N ) for N ≤ 2l + 1,
L = 21 (1.39)
2 (N − 2l − 1)(4l + 2 − N ) for N > 2l + 1.

The ground state Ψ (NLMSMs ) is (2L + 1)(2S + 1)-fold degenerate. The mul-
tiplet is written as 2S+1 LJ where L takes S, P, D, F, G, H, . . . according to
L = 0, 1, 2, 3, 4, 5, . . . , and J (= |L ± S|) is the total angular momentum. These
properties are summarized in Fig. 1.2 for 3d transition metal atoms.
The value of J at the ground state is determined by the spin–orbit interaction.
The spin–orbit interaction (1.16) in the atomic system can be written as

HSO = ζnl (l)νν  · (s)αγ aνα
†
aν  γ . (1.40)
ναν  γ

Here
 ∞
1 
ζnl = φnl (r)2 dV r dr > 0. (1.41)
2 0 dr

Assume that the ground state is determined by the Hund rule, and the temperature
is such that the corresponding thermal energy is much lower than that of the first ex-
cited state of the Coulomb interaction, i.e., the Hund-rule coupling JH : kB T  JH .
Here JH is an average value of {Jνν  }, and kB denotes Boltzmann’s constant. More-
over the spin–orbit interaction energy is much smaller than JH in 3d transition
metals. In this case, we can neglect all the excited state of the Coulomb interac-
tions, and limit the states to the (2L + 1)(2S + 1) dimensional Hund-rule subspace:
{Ψ (NLMSMs )}. We can then verify the following relation in the subspace.
   
Ψ (NLMSMs )HSO Ψ N LM  SMs
   
= λ(NLS) Ψ (N LMSMs )L · S Ψ N LM  SMs . (1.42)
10 1 Introduction to Magnetism

Fig. 1.2 The ground state of atoms with 3d unfilled shell. N and mz denote the d electron number
and the orbital magnetic quantum number, respectively. Electron configuration for each N in the
upper row of the figure shows the state Ψ (NLLSS) leading to the total spin S and the orbital
angular momentum L at the ground state. J denotes the total angular momentum. The ground
state multiplets (GS) are expressed as 2S+1 LJ . The experimental values of the spin–orbit coupling
constant λ are given. Examples of the 3d transition metal ions are also given in the bottom row

Thus we have an effective Hamiltonian for the spin–orbit interaction in the Hund-
rule subspace as follows.

HSO = λ(NLS)L · S. (1.43)

The coefficient λ(NLS) is obtained by comparing the diagonal matrix element of

HSO for the state Ψ (N LLSS) obtained from (1.40) with that of the r.h.s. of (1.43).
⎧
⎪ ζnl
⎨ for N ≤ 2l + 1,
λ(NLS) = N (1.44)
⎪
⎩− ζnl
for N > 2l + 1.
4l + 2 − N
Because ζnl > 0, λ(NLS) > 0 for N ≤ 2l + 1 and λ(N LS) < 0 for N > 2l + 1.
Thus, J = |L − S| is realized for N ≤ 2l + 1, and J = |L + S| for N > 2l + 1 at the
ground state among possible states J = |L + S|, |L + S − 1|, . . . , |L − S|. Therefore
J = |L − S| is the ground-state in the light transition-metal elements and the light
1.4 Metal and Insulator in Solids 11

rare-earth elements, and J = |L + S| is the ground state in the heavy transition-

metal elements and the heavy rare-earth elements. Note that the typical value of the
spin–orbit coupling is only one tenth of the Hund rule coupling (JH ∼ 1 eV) in the
case of 3d transition metal ions as shown in Fig. 1.2. When the spin–orbit coupling
is significant as in the rare-earth atoms, the ground-state should be (2J + 1)-fold
degenerate instead of (2S + 1)(2L + 1) fold.
Ions such as Ti2+ , V3+ , and Cr4+ , for example, have two electrons in the d shell
(see the N = 2 column in Fig. 1.2). The Hund rule tells us that the total spin and
orbital angular momenta at the ground state are S = 1 and L = 3, respectively. Be-
cause the spin–orbit coupling λ > 0 in the light atoms, the total angular momentum
at the ground state is given by J = 3 − 1. Thus we have the ground-state multiplet
3 F according to the notation 2S+1 L . In the same way, the ion Co2+ has 7 elec-
2 J
trons in the d shell. According to the Hund rule, we find the ground-state spin and
orbital momenta S = 3/2 and L = 3, respectively. Because the spin–orbit coupling
λ < 0 in the heavy atoms, the total angular momentum is given by J = 9/2. Thus
we obtain the ground-state multiplet 4 F9/2 for Co2+ ion.
When the external magnetic field H is applied to the atoms, the energy due to the
magnetic field is given by −gJ μB Jz H . Here gJ is Landé’s g factor and is given by
gJ = 3/2 + [S(S + 1) − L(L + 1)]/2J (J + 1). The susceptibility which is defined
by the induced magnetization divided by H is then given by

gJ2 μ2B J (J + 1)
χ= . (1.45)
3kB T
The above temperature dependence is referred as the Curie law, and C ≡
gJ2 μ2B J (J + 1)/3kB is called the Curie constant. The observation of the Curie law
in the susceptibility measurement is regarded as an indication of the existence of
local magnetic moments.

1.4 Metal and Insulator in Solids

When atoms form a solid, electrons in the outermost shell of atoms start to hop from
atom to atom via overlap between atomic orbitals. The question then is whether
electrons tend to remain in each atom or to itinerate in the solid. In the case of the
former, atomic magnetic moments are well defined in solids and their properties are
expected to be explained from the atomic point of view. In the case of the latter,
magnetic properties may be quite different from those expected from atomic ones
and one has to start from the itinerant limit in order to explain their magnetism.
Whether electrons in solids are movable or not is governed by the detailed bal-
ance between the energy gain due to electron hopping and the loss of Coulomb in-
teraction. The simplest example may be the case of the hydrogen molecule H2 . Ac-
cording to the molecular orbital picture, the 1s atomic orbitals split into the bonding
state with energy ε0 − |t| and the anti-bonding state with energy ε0 + |t| when hy-
drogen atoms form a molecule. Here ε0 denotes the atomic ground state, and t is the
12 1 Introduction to Magnetism

electron hopping integral which is defined by atomic potential v(r) and wave func-
tion ϕ(r) as t = ϕ(r − R)v(r − R)ϕ(r) dr. The bonding orbital is occupied by
two electrons according to the Pauli principle. It leads to the total energy gain 2|t|.
This is the covalent bond in the hydrogen molecule. In this state, electrons hop from
one atom to another, and thus they are movable. The covalent bond is stabilized by
the kinetic energy gain of independent electrons. It consists of the polarized state in
which the 1s orbital of an atom is doubly occupied and the orbital of another atom
is unoccupied, and the neutral state in which each atomic orbital is occupied by an
electron.
When the Coulomb interaction between electrons are taken into account, the
covalent bonding state is not necessarily stable because it contains the polariza-
tion state with double occupancy on an atomic orbital. Assume that the loss of
the intraatomic Coulomb interaction energy in the covalent bonding state is given
by U , and consider the neutral-atom state as the state in which each electron is
localized on an atom. The total energy in the covalent bonding state is given by
E(covalent) = 2ε0 − 2|t| + U , while the energy in the neutral atom state is given
by E(neutral) = 2ε0 . Thus, the neutral atom state is realized when the following
condition is satisfied.

2|t| < U. (1.46)

This condition is satisfied when the interatomic distance goes to infinity because |t|
goes to zero. Thus the neutral atom state in which electrons are localized on each
atom and each atom has a well defined local magnetic moment s = 1/2 is realized
in the atomic limit.
In solids, we expect the same behavior as found in the hydrogen molecule. Let us
consider the behavior of electrons when atoms form a solid. Electrons in solids move
in a potential obtained by a superposition of the atomic potentials i v(r − R i ).
Here v(r − R i ) denotes the atomic potential on site i. When we adopt the atomic
orbitals {ϕiν (r − R i )} as the basis functions and assume the orthogonality between
the orbitals, the Hamiltonian for electrons on the outermost shells in solids may be
obtained from (1.13) as

H = H0 + HCoulomb , (1.47)
  †
H0 = εiν niν + tiνj ν  aiνσ aj ν  σ , (1.48)
iν iνj ν 
   1

HCoulomb = Uiνν niν↑ niν↓ + Uiνν  − Jiνν  niν niν 
2
iν i ν>ν 

−2 Jiνν  s iν · s iν  . (1.49)
i ν>ν 

Here εiν is the atomic level for the orbital ν of site i, and tiνj ν  the transfer integral
between the orbital ν at site i and the orbital ν  at site j . The latter is expressed in
1.4 Metal and Insulator in Solids 13

the two-center approximation as


∗
tiνj ν  = ϕiν (r − R i )v(r − R i )ϕiν  (r − R j ) dr. (1.50)

Note that we have taken into account in HCoulomb only the intra-atomic Coulomb
interactions and omitted the inter-site Coulomb interaction contributions for sim-
plicity.
It is not easy to treat the electrons in solids described by the Hamiltonian (1.47)
because both the electron hoppings and the Coulomb repulsions have to be taken
into consideration. In order to discuss both the itinerant and localized behaviors of
electrons in solids, we can consider a simpler Hamiltonian consisting of one orbital
per site as follows.
  †

H= ε0 niσ + tij aiσ aj σ + U ni↑ ni↓ . (1.51)
iσ ij σ i

Here ε0 , tij , and U denote the atomic level, the transfer integral between sites i
and j , and the intra-atomic Coulomb interaction energy, respectively. The Hamilto-
nian (1.51) is known as the Hubbard model, and was proposed by Gutzwiller and
Hubbard independently [6–10].
The Hubbard model (1.51) is the simplest Hamiltonian which describes the
motion of interacting electrons in solids. Nevertheless it describes the localiza-
tion of electrons as well as their itinerant behavior in solids. Let us consider the
atomic limit of the model. For an atom, we have 4 atomic states: the empty state
(n↑ = 0, n↓ = 0), the single electron states (n↑ = 1, n↓ = 0) and (n↑ = 0, n↓ = 1),
and the doubly occupied state (n↑ = 1, n↓ = 1). Associated energies are given as 0,
ε0 , ε0 , and 2ε0 + U , respectively. An unfilled shell with spin S = 1/2 appears only
for the n = 1 state. Note that there is no Hund’s rule arrangement for spin in this
case.
In the atomic limit for a solid where tij = 0, electron number ni on eachatom is
no longer constant, though the total number of electrons N is given; N = iσ niσ .
The eigenstates are given by |Ψ  = |{niσ }, i.e., a set of the electron numbers with
spin σ on site i. The eigenenergy for the state is given by

E {niσ } = (ε0 ni + U ni↑ ni↓ ). (1.52)
i

It should be noted that the energy of the system increases by U when the number
of doubly occupied states D is increased by one. The ground-state energy E0 of the
atomic limit is obtained by minimizing the energy with respect to the number of
double occupancy in solid. Assume that the number of lattice points is given by L.
When N < L, the ground-state energy is obtained as E0 = ε0 N by choosing D = 0.
Magnetic moments on the sites with an electron are active in this case as shown in
Fig. 1.3. Because there are L!/N !(L − N )! electron configurations on the L lattice
points, the ground state is [2N L!/N !(L − N )!]-fold degenerate.
14 1 Introduction to Magnetism

Fig. 1.3 Electron configurations for less than half filling (upper figure) and for more than half
filling (lower figure) in the atomic limit

At half filling, all the atoms are occupied by an electron so that spin de-
grees of freedom by 2N remain; the degenerated wave functions are given by
|1s1z 1s2z 1s3z . . .  when the wave function |{niσ } is written as |n1 s1z n2 s2z n3 s3z . . . 
by using the charge ni = ni↑ + ni↓ and the spin siz = (ni↑ − ni↓ )/2.
When the electron number N is larger than L, it is no longer possible to keep
D = 0; the minimum value of D is given by D = N − L. The ground state energy is
then given by E0 = ε0 N +U (N −L). The ground state is [22L−N L!/(2L−N )!(N −
L)!]-fold degenerate because there are L!/(2L − N )!(N − L)! configurations for
choosing 2L − N(< L) sites with the single electron from L lattice sites and there
are 22L−N spin degrees of freedom for each configuration. Note that the spins on
2L − N sites are active in this case.
Let us consider the case that electron hopping tij is small but finite. When N = L,
electrons are mobile at T = 0 because electrons can move from site to site without
increasing the double occupation number D. Thus the system is a metal. However,
at half-filling, electron hopping creates a doubly occupied state, thus creating a finite
excitation energy by the amount of Coulomb interaction energy U . Therefore elec-
trons cannot move under the infinitesimal electric field. We have then an insulator
with local magnetic moments at each site in this case.
When there is no Coulomb interaction (U = 0), on the other hand, electrons are
generally itinerant. The Hamiltonian is given as
 †
H= (H 0 )ij aiσ aj σ . (1.53)
ij σ

Here (H 0 )ij = ε0 δij + tij (1 − δij ). The noninteracting Hamiltonian is diagonalized

  †
by a unitary transformation aiσ = k akσ i|k (aiσ †
= k akσ i|k∗ ) so that

H= εk nkσ , (1.54)
kσ

where εk = ij k|i(H 0 )ij j |k is an eigen value for the tight-binding one-electron
Hamiltonian matrix H 0 , and the set {j |k} (j = 1, . . . , L) is the eigen vector for εk .
1.4 Metal and Insulator in Solids 15

The eigen states for noninteracting Hamiltonian H are given by |Ψ  = |{nkσ },
i.e., a set of electrons with momentum k and spin σ . The eigenenergy is given by

E {nkσ } = εk nkσ , (1.55)
kσ

where the c-number nkσ takes on the value of 0 or 1. The ground state is obtained
by putting electrons on the energy levels from the bottom to the Fermi level εF
according to the Pauli principle,
ε 

k <εF
†
|φ0  = akσ |0, (1.56)
kσ

so that the ground-state energy is given by

ε
k <εF
E0 {nkσ } = εk . (1.57)
kσ

Alternatively, defining the density of states per atom per spin as

1
ρ(ε) = δ(ε − εk ), (1.58)
L
k

we can express the ground-state energy per atom as

 εF
E0 = 2 ερ(ε) dε. (1.59)
−∞

A non-interacting electron system is in general metallic unless the electrons in

the atom form a closed shell. The electrons in such systems are mobile. This is
because one can add an electron at the energy level just above the Fermi level by
applying infinitesimal electric field. Note that spins of itinerant electrons are also
mobile. We may expect that there is a transition from metal to insulator at half
filling when the intra-atomic Coulomb interaction is increased. Assume that there
is a band for a non-interacting system whose band width is W . The center of the
gravity of the noninteracting band is assumed to be located at ε0 . When the Coulomb
interaction U is increased, each atom tends to be occupied by one electron, and
electron hopping to neighboring sites tends to be suppressed in order to reduce the
on-site Coulomb interaction energy. In the strongly correlated region, an electron
should have a potential ε0 + U on a site having an opposite-spin electron because
of the increment of the Coulomb interaction energy due to double occupation, while
an electron has a potential ε0 on an empty site. We then expect one more band with
the band width of order of W around ε0 + U . The density of states as excitation
spectrum is expected to split into two bands at Uc ∼ W (see Fig. 1.4). The formation
16 1 Introduction to Magnetism

Fig. 1.4 The upper and

lower Hubbard bands created
by on-site Coulomb
interaction U

Fig. 1.5 NaCl type structure

in NiO

of a gap at the Fermi level implies the existence of an insulator. The insulating state
therefore may be realized by the electron correlations when

U > W. (1.60)

This is Hubbard’s alloy analogy picture to the metal–insulator transition [9, 10].
The metal–insulator transition due to electron correlations as mentioned above is
commonly known as the Mott transition. The split bands are named the upper and
lower Hubbard bands, respectively. The insulator caused by the electron correlations
is referred as the Mott insulator.
The concept of the Mott insulator was first proposed by Mott [11]. He consid-
ered the case of NiO. NiO has the NaCl structure in which Ni–O chain network is
formed along [100] direction (see Fig. 1.5). The electronic configuration of the Ni28
(O8 ) atom is given by 1s2 2s2 2p6 3s2 3p6 3d8 4s2 (1s2 2s2 2p4 ). The oxygen atoms are
considered to form a closed shell in the compound taking electrons from Ni atoms,
so that we have

Ni2+ : 1s2 2s2 2p6 3s2 3p6 3d8 O2− : 1s2 2s2 2p6 .

In this case, the Fermi level should be on the d bands in the crystalline system, and
we can expect a metal because the 5-fold d bands overlap each other in general (see
Fig. 1.6).
1.4 Metal and Insulator in Solids 17

Fig. 1.6 Density of states for

NiO in the paramagnetic state
obtained by the band
calculation [12]. The vertical
dashed line shows the Fermi
level

Experimental results however indicate that NiO is an insulator. To clarify the

insulating behavior, Mott considered the electron hopping on the NiO network. Let
us estimate an excitation energy Eg when an electron moves to the neighboring
Ni2+ site. This process is similar to an excitation of the H2 molecule to a polarized
state.
Ni2+ O2− 2
−→ Ni3+ O2− + Ni+ O2− .
 
Assume that the intra-atomic Coulomb interaction is given by i (νσ,ν  σ  ) U ·
niνσ niν  σ  for simplicity. There an electron is coupled to the remaining 7 electrons
via the on-site Coulomb interaction. After the hopping, the electron is coupled to 8
electrons at the neighboring site, so that the Coulomb interaction is increased by U .
On the other hand, the kinetic energy gain due to electron hopping at Ni3+ and Ni2+
sites may be given by

2z|t| ∼ W, (1.61)

because electrons in surrounding Ni sites can enter into the Ni3+ ions and the elec-
trons in Ni2+ ions can jump to the surrounding Ni sites. Here z is the number of Ni
nearest neighbors, and t is an effective electron hopping between Ni atoms. Note
that |t| corresponds to an energy gain for the formation of the bonding state due to
electron hopping in the H2 molecule, while U corresponds to the Coulomb energy
loss in the polarized state. The excitation energy of an electron hopping is therefore
given by

Eg ∼ U − W. (1.62)

Thus, we again have the same criterion for the formation of the Mott insulator (1.60)
from the condition Eg > 0.
The magnetic materials in the insulator are considered to be the Mott-type insu-
lator in which the atomic magnetic moments are built up in the unfilled shell. CrO2
and CrBr2 are typical ferromagnetic insulators, while MnO, FeO, CoO, and NiO
18 1 Introduction to Magnetism

are known as antiferromagnetic insulators in which the atomic magnetic moments

change their direction alternatively from site to site.
On the other hand, there are many magnetic metals where the polarized electrons
are mobile. These materials are referred as the metallic magnets or itinerant electron
magnets. The magnetic properties of these systems are known as the metallic mag-
netism or itinerant magnetism. Fe, Co, and Ni are typical examples of the metallic
ferromagnet, while Cr and Mn alloys show the antiferromagnetism. We describe in
this book the itinerant magnetism of metals and alloys.

1.5 Quenching of Orbital Magnetic Moments

The expectation value of the total orbital magnetic moment M L  = L is usu-
ally quenched in the crystalline system due to the crystalline potential of electrons.
Let us consider the case where the spin–orbit interactions are negligible as in the
3d transition metals, and that the magnetic field is not applied. Moreover we may
assume in solids that the ground state is nondegenerate. One can then choose the
ground-state wave function to be real; Ψ ∗ = Ψ .
The angular momentum L is real because it is a physical quantity:

L = L∗ . (1.63)

Using the relation Ψ ∗ = Ψ , the r.h.s. of the above equation is written as

L∗ = −Ψ |L|Ψ . (1.64)

Therefore (1.63) indicates that

L = 0. (1.65)

Thus, one can expect in metals with sufficiently small spin–orbit interactions that
only spin magnetic moments contribute to the magnetization.

M = 2S. (1.66)

This is referred as the quenching of orbital magnetic moments.

In the atomic system, electrons are bound by the nuclei, and the orbital angular
momentum is finite in general according to Hund’s second rule. In the crystalline
system, the Coulomb energy gain leading to Hund’s second rule is not expected be-
cause electrons can hop to the neighboring sites to reduce the Coulomb energy, and
electrons can move turning around to right and left (see Fig. 1.7). It may lead to
L = 0. This is a physical reason for the quenching of orbital moments in solids. In
the 3d transition metals and alloys which we will discuss, the orbital magnetic mo-
ments are usually quenched. In the system with a large spin–orbit coupling such as
the rare-earth metals or the system with a strong magnetic field, the orbital moments
remain.
1.6 Heisenberg Model 19

Fig. 1.7 Concept of orbital

magnetic moments in atom
(L = 0: left) and solids
(L = 0: right)

1.6 Heisenberg Model

In the insulator, magnetic behaviors are often described by a simple Hamiltonian

known as the Heisenberg model. We briefly discuss the Hamiltonian and related
properties of the insulator magnetism in this section.
Let us consider the hydrogen molecule. We assume one orbital for each hydrogen
atom, ϕ1 for atom 1 and ϕ2 for atom 2, and neglect the overlap integrals for simplic-
ity, i.e., ϕ1 |ϕ2  = 0. The Hamiltonian for the hydrogen molecule is obtained from
(1.13) as we derived (1.47).
    
† 1
H= ε0 niσ + tij aiσ aj σ + U ni↑ ni↓ + K − J n1 n2 − 2J s 1 · s 2 .
2
iσ ij σ i
(1.67)

Here K and J denote the intersite Coulomb and exchange integrals given by

ϕ1∗ (r)ϕ2∗ (r  )ϕ2 (r  )ϕ1 (r)
K= dr dr  , (1.68)
|r − r  |

ϕ1∗ (r)ϕ2∗ (r  )ϕ1 (r  )ϕ2 (r)
J= dr dr  . (1.69)
|r − r  |

In the atomic subspace where n1 = n2 = 1 and the electron hopping is sup-

pressed, the Hamiltonian (1.67) reduces as follows.

1
H = 2ε0 + K − J − 2J s 1 · s 2 . (1.70)
2
It is diagonalized with the use of the total spin S (in the representation of S and
Sz ) as
 
H = 2ε0 + K − J S(S + 1) − 1 . (1.71)

Therefore the hydrogen molecule has the energy 2ε0 + K − J for the triplet state
and 2ε0 + K + J for the singlet state when the inter-site atomic distance is large.
This is the Heitler–London theory for the hydrogen molecule.
The last term in (1.70) denotes an interaction between the atomic spins s 1 and s 2 .
In general, the following Hamiltonian showing the interactions between atomic
20 1 Introduction to Magnetism

spins {S i } is known as the Heisenberg model.


H =− Jij S 1 · S 2 . (1.72)
(i,j )

The interaction Jij = 2J in the Heitler–London theory is the inter-site exchange

integral due to the Coulomb interaction. This coupling in the Heisenberg model
is referred as the direct exchange interaction. The direct exchange interaction is
generally positive in sign (i.e., ferromagnetic).
Another type of the Heisenberg model with effective exchange coupling is pos-
sible in the insulator. In order to derive the effective Hamiltonian, we first derive
the perturbative effective Hamiltonian from a more general point of view [13]. The
effective Hamiltonian is constructed such that it yields the same energy eigen values
in a subspace as in the original Hamiltonian acting on the full Hilbert space.
 P spanned by a set of states {|s}, and introduce a
Let us consider a subspace
projection operator P = s |ss| which projects any state onto subspace P. The
projection operator which chooses the complementary subspace Q is defined by
Q = 1 − P . Note that P 2 = P , Q2 = Q, and P Q = QP = 0.
The eigen value problem of the original Hamiltonian H is written as

(H − E)ψ = 0. (1.73)

Inserting the identity 1 = P + Q into the above equation as (P + Q)(H − E)(P +

Q)ψ = 0 and multiplying P from the l.h.s. of the equation, we obtain
 
(P H P − E)P + P H Q ψ = 0. (1.74)

In the same way, we can derive the following equation

 
QH P + (QH Q − E)Q ψ = 0. (1.75)

In (1.74) and (1.75), the wave function ψ is separated into two components be-
longing to different subspaces, P ψ and Qψ. From both equations, one can elim-
inate the wave function Qψ that leads to the eigenvalue equation in subspace P,
namely,
 
P H P − P H Q(QH Q − E)−1 QH P P ψ = EP ψ. (1.76)

The above equation indicates that the following Hamiltonian Hp (E) acting on the
wave function P ψ in the subspace P is regarded as an effective Hamiltonian which
yields the same eigen value E as in the original Hamiltonian H , where

Hp (E) = P H P − P H Q(QH Q − E)−1 QH P . (1.77)

The effective Hamiltonian depends on the energy which should be obtained self-
consistently. Taking the subspace spanned by a subset of the eigenfunctions of an
1.6 Heisenberg Model 21

unperturbed Hamiltonian, one can derive the effective Hamiltonian which is cor-
rect up to the second order of the interaction Hamiltonian. Assume that the Hamil-
tonian consists of a noninteracting part H0 and an interaction part HI such that
H = H0 + HI . The eigenvalues {Ek0 } and eigenfunctions {φk } for H0 are assumed to
be known. We define the subspace P spanned by a part of {|φk }. The projection op-

erator is then defined by P = P μ |φμ φμ |. We have the relation QH P = QHI P ,
and thus obtain the effective Hamiltonian which is correct up to the second order
in interaction, after having replaced the Hamiltonian H and the energy E in the
denominator of (1.77) with the zeroth-order ones, H0 and Ek0 , i.e.,

−1
Hp = P H P − P HI Q QH0 Q − Ek0 QHI P . (1.78)

The effective Hamiltonian depends on the eigen state k explicitly. Therefore we

rewrite the second term as follows.
P 
 Q
−1 −1
QH0 Q − Ek0 QHI P φk = Eν0 − Eμ0 |φν φν |HI |φμ φμ |φk .
μ ν
(1.79)

Thus the perturbative Hamiltonian (1.78) is expressed as follows.

−1
Hp = P H P − P HI Q EQ
0
− EP0 QHI P . (1.80)

Here EP0 (EQ0 ) denotes the eigenvalues belonging to subspace P (Q). The eigen

value equation for the perturbative effective Hamiltonian is then given by

Hp P ψk = Ek P ψk . (1.81)

We can derive the effective Heisenberg model in the strong Coulomb interac-
tion regime at half-filling using the formula (1.80). Let us consider the Hubbard
model (1.51). As found in Sect. 1.4, the ground state of the atomic limit at half-
filling is specified by {ni = 1} states. The ground state is 2L -fold degenerate, L
being the number of sites. We start from the ground state in the atomic limit at
half filling, i.e., the {ni = 1} states, and define the subspace P from them. The
complementary subspace Q consists of the subspace containing empty states and
the subspace containing the doubly occupied states. In the strongly correlated re-
gion, the electron hopping rate |tij | is much smaller than the Coulomb interac-
tion strength U , so that the Hamiltonian H is separated into the atomic state
   †
H0 = iσ ε0 niσ + i U ni↑ ni↓ and the ‘interaction’ term HI = ij σ tij aiσ aj σ .
The first term at the r.h.s. of (1.80) is given by P Lε0 P . In the calculations of
 † †
the second term, we have QHI P = ij σ tij aiσ aj σ P because aiσ aj σ P belong to
the space Q. Moreover, EQ 0 − E 0 = U when it is operated on the single doubly-
P
occupied state QHI P . Thus we obtain the expression for the second term at the r.h.s.
22 1 Introduction to Magnetism

of (1.80) as follows.
−1
 tlk tij † †
0
P HI Q EQ − EP0 QHI P = P alσ  akσ  aiσ aj σ P . (1.82)

U
ij klσ σ

Among the summation at the r.h.s. of (1.82) only the k = i and l = j terms
remain. Rearranging the creation and annihilation operators, we obtain the relation.
 1
aj†σ  aiσ  aiσ
†
aj σ = nj − ni nj − 2s i · s j . (1.83)

2
σσ

Substituting the above relation into (1.82), we obtain the following effective Hamil-
tonian.
  
1  |tij |2 4|tij |2
Hp = P L ε0 − + si · sj P . (1.84)
2 U U
j (i,j )

Apart from the constant term, we arrive at the following effective Heisenberg model
in the strong Coulomb interaction regime at half filling, which operates on the sub-
space P.

H =− Jij s 1 · s 2 . (1.85)
(i,j )

Here
4|tij |2
Jij = − . (1.86)
U
The effective exchange integral (1.86) is known as the super exchange interaction
because it is caused by a virtual exchange of electrons belonging to different atoms
via transfer integrals. The super exchange interaction is antiferromagnetic, while the
direct exchange interaction (1.69) is ferromagnetic. Note that the subspace on which
the Hamiltonian (1.85) acts is given by {P |n1 s1z , n2 s2z , . . .} = {|1s1z , 1s2z , . . .}.
This can be simplified as {|s1z , s2z , . . .}. In many magnetic insulators, anions are
often located between the magnetic ions. In such a case, the super exchange interac-
tion becomes dominant. The Heisenberg model (1.85) is not justified in the metallic
system where the band width W ∼ U . Nevertheless, we often find the same type of
the intersite magnetic interactions, which are useful for qualitative understanding of
the magnetism (see Sects. 6.4 and 8.3 for examples).

1.7 Magnetic Structure

Magnetic materials have their microscopic structure for arrangement of atomic mag-
netic moments. This is called the magnetic structure. The structure characterizes the
1.7 Magnetic Structure 23

magnetic property as well as the electronic structure. In this section, we introduce

the Fourier representation of the magnetic structure and explain briefly the magnetic
structures in solids.
The magnetic structure is characterized by the magnetic moment density which
is given by
 
M(r) = ψ † (r)mψ(r) . (1.87)

Here ψ(r) = (ψ↑ (r), ψ↓ (r)) is the field operator of electrons. m denotes the mag-
netic moment for an electron as given by (1.7).
In a crystalline system, we may divide the space into the Wigner–Seitz cells for
each atom. Then we can define a magnetic moment of atom i as follows.

mi = M(r) d 3 x. (1.88)
i

Here the integration is over the Wigner–Seitz cell belonging to atom i. In a non-
crystalline system we may adopt the Voronoi polyhedra instead of the Wigner–Seitz
cells.
In the case of the crystalline system, the atomic position R l is expressed by using
the primitive translational vectors a, b, c as

R l = l1 a + l2 b + l3 c. (1.89)

The atomic magnetic moment is expanded with use of the Fourier lattice series as
follows.

ml = m(q)eiq·R l . (1.90)
q

Here q is the wave vector in the first Brillouin zone. The Fourier components of the
magnetic moments are given as
1
m(q) = ml e−iq·Rl , (1.91)
L
l

L being the number of lattice points. Note that m(−q) = m(q)∗ because ml are
real. Moreover we assumed that there is only one atom per unit cell. When there are
more than one atom per unit cell, we have to add the atomic position in a unit cell
ηλ to (1.89). Accordingly, the magnetic moments (1.90) and (1.91) are specified by
(λ)
the type of atom λ as ml (m(λ) (q)). Hereafter we consider the crystalline system
with one magnetic atom per unit cell.
The microscopic magnetic structure of magnetic materials is specified by a set of
{ml } or {m(q)}. The simplest structure is that all of the atomic magnetic moments
have the same direction. It is realized when all the Fourier components vanish except
m(q = 0) = m, and is known as the ferromagnetic structure (see Fig. 1.8). Typical
transition metals such as Fe, Co, and Ni exhibit the ferromagnetism.
24 1 Introduction to Magnetism

Fig. 1.8 Ferromagnetic

arrangement

Fig. 1.9 Antiferromagnetic

arrangement

The antiferromagnetic structure (AF) is defined by alternative arrangement of the

atomic magnetic moments along a direction, e.g., the z direction (see Fig. 1.9). It is
specified by a set, m(±Q) = mk exp (±iα) and m(q = ±Q) = 0, where Q consists
of half a reciprocal lattice vectors, and k denotes a unit vector along the z axis, and
α denotes a phase of the structure. In the case α = 0, we have

ml = mk cos(Q · R l ). (1.92)

When the position R l moves along Q vector, Q · R l changes by π . Accordingly,

the magnetic moment ml changes its sign.
The AF structure on the simple cubic lattice as shown in Fig. 1.10(a) is expressed
by the wave vector Q = (K 1 + K 2 + K 3 )/2 = (1, 1, 1)π/a, a being the lattice con-
stant. The AF structure on the body-centered cubic lattice as shown in Fig. 1.10(b)
is expressed by the wave vector Q = (−K 1 + K 2 + K 3 )/2 = (0, 0, 1)2π/a. In the
case of the face-centered cubic lattice (fcc) structure, two kinds of AF structures are
known. The AF structure of the first-kind as shown in Fig. 1.10(c) is described by a
wave vector Q = (K 2 + K 3 )/2 = (0, 0, 1)2π/a. The atomic moment alternatively
changes the direction with a translation by (0, 0, 1)a/2. It is also possible to change
direction alternatively along 1 1 1 axis. This is referred as the AF structure of the
second-kind, and is characterized by Q = (K 1 + K 2 + K 3 )/2 = (1, 1, 1)π/a (see
Fig. 1.11). Cr with 1 at% Mn on the bcc lattice shows the AF structure, and γ -Mn
on the fcc lattice shows the AF structure of the first-kind according to the neutron
diffraction experiments.
It should be noted that the network consisting of the up-spin atoms (or the down-
spin atoms) forms a lattice referred as the sublattice. There are two sublattices in
the antiferromagnetic structures shown in Fig. 1.10. Each sublattice forms the fcc
lattice with the lattice constant 2a in the case of the sc structure (Fig. 1.10(a)), the
sc lattice with the lattice constant a in the case of the bcc structure
√ (Fig. 1.10(b)),
and the simple tetragonal structure with the lattice constants a/ 2 and a in the case
of the fcc structure (Fig. 1.10(c)), respectively.
The sinusoidal spin density wave (SDW) structure is expressed by

ml = mk sin(Q · R l + α), (1.93)

where Q vector is neither equal to 0 nor to the AF wave vectors. The magnetic
moment sinusoidally changes with a period λ = 2π/|Q| along the Q direction (see
1.7 Magnetic Structure 25

Fig. 1.10 Antiferromagnetic arrangements on the simple cubic lattice (a), body-centered cubic
lattice (b), and face-centered cubic lattice (c). The lattice constants are shown by a

Fig. 1.11 The

antiferromagnetic structure of
the second-kind on the
face-centered cubic lattice

Fig. 1.12 Spin density wave

structure

Fig. 1.12). Note that the period is not necessarily commensurate with the lattice
spacing in general. Cr is a well-known example of the SDW.
It is also possible that the magnetic moment rotates with a translation (see
Fig. 1.13). This is known as the helical structure, and is expressed as
 
ml = m e1 cos(Q · R l + α) + e2 sin(Q · R l + α) . (1.94)

Here e1 and e2 are the unit vectors being orthogonal to each other. The magnetic
moment with amplitude m rotates on the e1 –e2 plane with the translation along the
Q vector.
The above expression (1.94) can also be written as

ml = m(Q)eiQ·R l + m(Q)∗ e−iQ·R l , (1.95)

with
m
m(Q) = (e1 − ie2 )eiα . (1.96)
2
It indicates that the helical structure is specified by the Q vector and the condition

m(Q)2 = 0. (1.97)
26 1 Introduction to Magnetism

Fig. 1.13 Helical structure (left) and conical structure (right)

Fig. 1.14 2Q multiple-

spin-density wave structure

It is known from the neutron experiments that Au2 Mn, MnO2 , CrO2 , Eu, Tb, Dy,
Ho etc. show the helical structure.
So far the magnetic structures are characterized by only one Q vector. One can
also consider the helical type structure with the bulk magnetization as follows.

ml = mz k + m i cos(Q · R l + α) + j sin(Q · R l + α) . (1.98)

The above expression is alternatively written as

ml = mz k + m(Q)eiQ·R l + m(Q)∗ e−iQ·R l , (1.99)

with

m(Q)2 = 0. (1.100)

This is referred as the conical magnetic structure (see Fig. 1.13). The magnetic mo-
ment rotates for example on the x–y plane when the moment translates along the
direction of Q.
1.7 Magnetic Structure 27

Fig. 1.15 3Q multiple-

spin-density wave structure

One can also consider a magnetic structure consisting of two wave vectors, Q1
and Q2 . For example, we can consider a structure such as

ml = m(Q1 )eiQ1 ·Rl + m(Q2 )eiQ2 ·R l + c. c. (1.101)

Assuming that m(Q1 ) = mx i/2 and m(Q2 ) = mx j /2, we have

ml = mx i cos(Q1 · R l ) + my j cos(Q2 · R l ). (1.102)

This is known as a double Q multiple SDW (2Q-MSDW) (see Fig. 1.14). In the
case of the fcc lattice, we may consider the wave vectors Q1 = (1, 0, 0)2π/a and
Q2 = (0, 1, 0)2π/a. Then the x(y) component changes the sign with a translation
R l = (1, 0, 0)a/2 (R 2 = (0, 1, 0)a/2) as shown in Fig. 1.14.
We can also consider the magnetic structure consisting of three wave vectors
Q1 , Q2 and Q3 , which is the so-called triple Q multiple SDW (3Q-MSDW). For
example, we have a 3Q-MSDW such that

ml = mx i cos(Q1 · R l ) + my j cos(Q2 · R l ) + mz k cos(Q3 · R l ), (1.103)

with Q1 = (1, 0, 0)2π/a, Q2 = (0, 1, 0)2π/a, and Q3 = (0, 0, 1)2π/a. Each com-
ponent changes sign after the translation by a/2 along the same direction as shown
in Fig. 1.15.
In the substitutional disordered alloys, we have more complicated structures
which cannot be described by a small number of wave vectors. When we take a
configurational average of the atomic magnetic moments, we can define the aver-
age magnetization [ml ]c . Here [∼]c denotes the configurational average. In some

Fig. 1.16 Spin glass arrangement

28 1 Introduction to Magnetism

cases, an ordered state with randomly oriented magnetic moments referred as the
spin glass (SG) appears (see Fig. 1.16). The SG is considered to be characterized by
the condition such that
 
[ml ]c = 0 and m2l c = 0. (1.104)

The disordered dilute alloys such as Cu1−x Mnx (x  0.1) and Au1−x Fex (x  0.1)
are known to show a SG at low temperatures (see Sect. 7.1).
Experimentally most magnetic structures of crystalline systems are determined
by the neutron elastic scattering experiments. The readers who are interested in the
experimental determination of the magnetic structure are recommended to refer to
the books by Marshall and Lovesey [14, 15].
http://www.springer.com/978-3-642-33400-9