Вы находитесь на странице: 1из 42

July 31, 2012

11:38

World Scientific Book - 9in x 6in

Julia 8556-Quantum Theory of Solids

Chapter 1

Lattice Dynamics

When the structure and cohesion of solids are studied, we assume that the
atoms or ions in solids stay at their respective equilibrium positions. This
is sufficient for the purpose of studying their structural and binding properties. However, when we pursue to understand many other properties of
solids, such as their thermodynamic properties, the picture of static atoms
or ions in solids becomes inadequate and their dynamics must be taken into
consideration. As a matter of fact, atoms or ions in solids never stay persistently at their equilibrium positions at finite temperatures. Instead, they
move back and forth (that is, they vibrate or oscillate) constantly about
their equilibrium positions. This kind of motion is referred to as lattice vibrations and the entire subject related to lattice vibrations is called lattice
dynamics or crystal dynamics.
Lattice dynamics can be said to be the oldest branch of solid state
physics. To be convincing, we now trace some of the early important developments in lattice dynamics. In 1907, Einstein1 published his work on
the lattice specific heat, entitled Plancks theory of radiation and the theory of specific heat (the birth of the Einstein model on the lattice specific
heat). In 1912, Born and von Karman2 published their work on lattice
vibrations, entitled On vibrations in space lattices (the birth of the formal
theory of lattice dynamics), and Debye3 published his work on the lattice
specific heat, entitled On the theory of specific heat (the birth of the Debye
model on the lattice specific heat). There are many other early landmark
developments.
1 A.

Einstein, Annalen der Physik 22, 180 (1907).


Born and Th. von Karman, Physikalische Zeitschrift 13, 297 (1912); ibid. 14, 15
(1913).
3 P. Debye, Annalen der Physik (Leipzig) 39, 789 (1912).
2 M.

han

July 31, 2012

11:38

World Scientific Book - 9in x 6in

Julia 8556-Quantum Theory of Solids

A Modern Course in Quantum Theory of Solids

In 1950s, Brockhouse4 pioneered the measurement of the spectrum of


lattice vibrations (the dispersion relations of normal modes of lattice vibrations) of a solid through inelastic neutron scattering experiments. The
development of this experimental technique provided a great impetus to
the study of lattice dynamics in various types of solids.
Lattice vibrations are very important because they play vital roles in
many physical properties of solids. Lattice vibrations can scatter electrons
in a metal and thus yield resistance to the motion of electrons, which leads
to the increase in the resistivity of the metal. Lattice vibrations can take
heat from or give heat to the environment and thus affect the heat capacity
of a solid. A certain kind of lattice vibrations interact with photons and
thus have an impact on the optical properties of a solid. The interaction
of lattice vibrations with conduction electrons in a metal can even change
the ground state of the electrons in a fundamental way and render them to
be superconducting. The consequences of lattice vibrations on the physical
properties of solids are so many that one can hardly give an exhausted list
in a limited space.
Because of their paramount importance, a thorough study of lattice
vibrations is undoubtedly necessary. Surprised or not, lattice vibrations
call for both classical and quantum theories for their complete descriptions.
The necessity of a quantum theory of lattice vibrations is clearly testified
by the inability of classical theory of lattice vibrations to produce the correct temperature dependence of the specific heat of a solid as observed in
experiments. Lattice vibrations must be studied in three stages before their
effects on physical properties can be fully unveiled. In the first stage, various vibrational modes are obtained through solving the classical equations
of motion of atoms or ions. In the second stage, vibrational modes are
quantized according to the canonical quantization rules. This is the first
time that lattice vibrations are quantized. The second stage acts only as a
transition. For a better understanding of their properties and for the convenience of their applications, lattice vibrations are quantized for the second
time in the third stage. With the second quantization, the consequences of
lattice vibrations on physical properties of solids can be fully investigated.
Regardless of the type of bonding, any solid can be taken as composed
of electrons and nuclei. These two kinds of particles are intimately coupled

4 B.

N. Brockhouse and A. T. Stewart, Physical Review 100, 756 (1955).

han

July 31, 2012

11:38

World Scientific Book - 9in x 6in

Julia 8556-Quantum Theory of Solids

Lattice Dynamics

han

together. The Hamiltonian of a solid is then given by


=H
e + H
n + H
en ,
H
X ~2
1 1 X
e2
e =
H
2i +
,
2m
2 40
|ri rj |
i
i6=j
X ~2
1 1 X ZI ZJ e2
n =
H
2I +
,
2MI
2 40
|RI RJ |
I
I6=J
X ZI e2
en = 1
H
,
40
|ri RI |

(1.1)

iI

where m is the mass of an electron, MI and ZI e are the mass and charge
of nucleus I, ri and RI denote the positions of the ith electron and the Ith
e is the Hamiltonian of the subsystem of electrons,
nucleus, respectively, H
n is the Hamiltonian of the subsystem of nuclei, and H
en is the interaction
H
Hamiltonian between electrons and nuclei.
The Hamiltonian in Eq. (1.1) is referred to as the fundamental Hamiltonian of a solid in the sense that all the properties of the solid can be
could be found exactly.
computed if the eigenvalues and eigenstates of H
Unfortunately, it is not in sight at all that any one could accomplish that.
Therefore, we have no choice but make some approximations to be able
to proceed to understand any physical properties of a solid. Because
the electrons and ions are coupled together, the separation of the electronic and nuclear motions would be of great help. This is provided by
the BornOppenheimer approximation that is also known as the adiabatic
approximation.
This chapter is organized as follows. The BornOppenheimer approximation is first introduced in Sec. 1.1 so that we can concentrate only on the
motion of nuclei (or atoms or ions) thereafter. We then attempt to develop
the classical theory of lattice vibrations as gently as possible, with the full
classical theory established in the end.
In Sec. 1.2, we introduce the harmonic approximation and derive the
harmonic lattice potential energy for a three-dimensional crystal with a
multi-atom basis. We then proceed to find the normal modes of lattice
vibrations of a solid under the harmonic approximation. Attention should
be paid to the way we solve the classical equations of motion of atoms: We
expand the displacement of an atom in terms of its Fourier components (i.e.,
make a Fourier transformation of the displacement of atoms with respect
to positions of primitive cells and time) so that the differential equations
are converted into algebraic equations.

July 31, 2012

11:38

World Scientific Book - 9in x 6in

Julia 8556-Quantum Theory of Solids

A Modern Course in Quantum Theory of Solids

In finding the normal modes of lattice vibrations in a crystal, we follow


these steps:
(1) Establish the classical equations of motion of atoms using Newtons
second law with the force acting on an atom derived from the harmonic
lattice potential energy.
(2) Fourier transform the displacements of atoms and convert the differential equations into algebraic equations.
(3) Find the allowed values of wave vector.
(4) Solve the resultant algebraic equations for the frequencies of normal
modes.
(5) Introduce the normal coordinates and polarization vectors for normal
modes.
(6) Solve for the polarization vectors.
(7) Discuss the properties of polarization vectors.
(8) Derive an expression for the displacements of atoms.
(9) Derive the Hamiltonian of the crystal under study.
The results obtained in the last two steps will be used in the quantization
of lattice vibrations.
1.1

BornOppenheimer Approximation

The characteristic speed of an electron in a solid is 106 m/s, while that of a


nucleus is 105 m/s. Thus, the electrons in a solid move much faster than the
nuclei. This is because the electrons are much lighter than the nuclei, m
103 MI , while the momenta they acquire through various interactions are
comparable. Because of the much higher mobility of the electrons than that
of the nuclei, when the configuration of the nuclei changes, the electrons
can respond instantaneously and thus remain essentially in the electronic
ground state. We can thus assume that the nuclei remain at their stationary
positions when the ground state of the electronic subsystem is solved. The
full potential energy of the nuclei is then obtained by taking the electronic
contributions into account and used subsequently to solve for the motion of
the nuclei. The disentanglement of the motion of the electrons and nuclei
in such a manner is known as the BornOppenheimer approximation or the
adiabatic approximation.
We now describe the BornOppenheimer approximation in more details.
For brevity in notations, we first introduce collective notations for the co-

han

July 31, 2012

11:38

World Scientific Book - 9in x 6in

Julia 8556-Quantum Theory of Solids

Lattice Dynamics

han

ordinates of the electrons and nuclei. Let r = {r1 , r2 , , rNe } collectively


denote the electronic coordinates and R = {R1 , R2 , , RN } the nuclear
coordinates. Here Ne is the number of electrons and N the number of nuclei.
Let (r, R) = (r; R)(R) be the wave function of the solid with (r; R)
and (R) the electronic and nuclear wave functions, respectively. The semicolon in (r; R) indicates that R is taken as a parameter when the motion
of the electronic subsystem is solved. We start with the eigenequation of
= H(r,
R) (i.e., the stationary Schrodinger equation of
the Hamiltonian H
the solid)
R)(r, R) = E(r, R),
H(r,

(1.2)

R). To proceed, we rearrange the above


where E is the eigenvalue of H(r,
equation as

1 
en (r, R) (r; R)
He (r) + H
(r; R)



1
n (R) (r; R)(R) .
=
EH
(1.3)
(r; R)(R)

To emphasize the coordinate dependence, we have explicitly displayed


proper coordinate variables in the Hamiltonians. Because of the entanglement of the variables r and R, the two sides of Eq. (1.3) can not both
equal either a constant or a function of only r or R. However, since the
nuclei can be taken as remaining at their stationary positions when the
ground state of the electronic subsystem is solved, the eigenequation for
the electronic states


e (r) + H
en (r, R) (r; R) = E(R)(r; R)
H
(1.4)

can be solved with the nuclei remaining in the configuration R, where E(R)
is the electronic eigenenergy in the nuclear configuration R. Inserting the
above equation into Eq. (1.3), we obtain



n (R) + E(R) (r; R)(R) = E(r; R)(R).
H
(1.5)

The motion of the nuclei is thus separated from that of the electrons. This
is a great step forward since we now have a recipe to solve the eigenequation
of the Hamiltonian of the solid albeit it is done approximately. If the variables in Eq. (1.3) had been separated exactly, we would have had an exact
solution to the problem. In a sense, the BornOppenheimer approximation
is equivalent to solving Eq. (1.2) with the separation of variables. Hence,
the impreciseness in the BornOppenheimer approximation is caused by
the forceful application of the separation of variables to Eq. (1.2).

July 31, 2012

11:38

World Scientific Book - 9in x 6in

Julia 8556-Quantum Theory of Solids

A Modern Course in Quantum Theory of Solids

The electronic energy E(R) is referred to as the adiabatic contribution


to the potential energy of the nuclei. The other terms related to electronic
wave functions are referred to as the non-adiabatic contribution and they
can be inferred from Eq. (1.5). Multiplying both sides of Eq. (1.5) by
(r; R) from left and then integrating over r (i.e., over r1 , r2 , , rNe ),
we have
 X 2

~
2

+ (R) (R) = E(R),


(1.6)
2MI I
I

where (R) is the nuclear potential energy and is given by


1 1 X ZI ZJ e2
+ E(R) + na (R).
(R) =
2 40
|RI RJ |

(1.7)

I6=J

The term na (R) in (R) is the non-adiabatic contribution and is given by



X ~2  Z
na (R) =
dr (r; R)I (r; R) I
MI
I
X ~2 Z
dr (r; R)2I (r; R)
(1.8)

2MI
I
R
R
with dr = dr1 dr2 drNe . Due to the afore-mentioned slow motion
of the nuclei in comparison with the electrons, the non-adiabatic contribution is usually small and can be taken into account perturbatively. The
non-adiabatic contribution is very important to some physical properties
of a solid since it describes the interaction between electrons and lattice
vibrations.
When only the electronic states are of concern, Eq. (1.4) can be solved
for a set of fixed nuclear positions (i.e., a fixed nuclear configuration). In
consideration of the large masses and slow motion of the nuclei, their motion is often solved using classical mechanics. In such a case, a potential
energy surface can be mapped out by solving Eq. (1.4) for different nuclear
configurations and then used in classical computations for the motion of
nuclei.
In practice, the nuclei in Eq. (1.2) are often replaced with ions or ion
cores since the core electrons play a much less role than valence electrons
in determining the properties of a solid.

1.2

Lattice Potential Energy and Harmonic Approximation

From the above discussions, we see that the nuclear potential energy, referred to as the lattice potential energy hereafter, can be obtained only

han

July 31, 2012

11:38

World Scientific Book - 9in x 6in

Julia 8556-Quantum Theory of Solids

Lattice Dynamics

han

after the electronic motion has been solved for all the nuclear configurations. Thus, it seems that the understanding of the lattice dynamics of a
crystal is impossible without the knowledge of the electronic states. However, the lattice potential energy can also be obtained empirically with the
input from experiments. Such an approach is called the pseudopotential
method. In any event, the lattice potential energy is assumed to be known
from now on. In this sense, our treatment is of phenomenological nature.
With the kinetic energy expressed in terms of momenta of atoms, the lattice
Hamiltonian is given by
X p2
i
=
H
+ (r1 , r2 , , rN ),
(1.9)
2m
i
i
where we have used the lowercase letter i to label an atom, the lowercase
letter m to denote its mass, the bold lowercase letter p to denote its momentum, and the bold lowercase letter r to denote its position. The bold
capital letter R is now reserved for the lattice vectors. Also, for brevity we
will generally refer to atoms as the constituents of a crystal in this section
even though they may be ions. But, ions will be used when an ionic crystal is explicitly referred to. For pairwise interactions between atoms, the
lattice potential (r1 , r2 , , rN ) can be written as
(r1 , r2 , , rN ) =

N
1 X
(ri rj ),
2

(1.10)

i6=j=1

where (ri rj ) is the interaction energy between atoms i and j.


For the given lattice potential energy of a crystal, the problem we face
is what to do with it to develop a theory for the lattice dynamics of the
crystal. To accomplish this, we make good use of the fact that atoms move
only in the close vicinities of their equilibrium positions (that is, the amplitudes of their vibrations are small). In the first step, we Taylor-expand
the lattice potential energy in terms of the displacements of atoms from
their equilibrium positions and keep only up to the second-order terms in
the expansion. This practice is known as the harmonic approximation. The
lattice Hamiltonian in the harmonic approximation is referred to as a harmonic Hamiltonian. The crystal with a harmonic Hamiltonian is referred
to as a harmonic crystal . The terms of orders higher than the second
order are referred to as anharmonic terms. Ordinarily, the contributions
from the anharmonic terms are negligibly small and can be safely ignored.
However, for crystals with extraordinary properties, such as ferroelectric
crystals and crystals that can undergo structural phase transformations,

July 31, 2012

11:38

World Scientific Book - 9in x 6in

Julia 8556-Quantum Theory of Solids

A Modern Course in Quantum Theory of Solids

the anharmonic terms become important. For such crystals, the effects of
the cubic and quartic anharmonic terms are often considered.
We now consider a three-dimensional crystal with a multi-atom basis.
To be able to keep track of the algebras comfortably, we first describe clearly
how the atoms are labeled and their positions denoted.
As generally done, we label each primitive cell by the lattice site on
which the primitive cell sits. Thus, the ith primitive cell locates on the
ith lattice site and its position vector is given by Ri that is the lattice
vector of the ith lattice site. Because of the presence of bases of atoms in
a crystal, the number of atoms in each primitive cell is greater than one.
The atoms within each primitive cell are indexed by positive integers, with
Greek letters (, , ) often used for the variables of indices. For a p-atom
basis, we have = 1, 2, , p. To refer to an atom within a primitive cell,
we can say the th atom within the ith primitive cell. The position of
an atom within a primitive cell is given in the local Cartesian coordinate
system associated with the primitive cell with the origin at the tip of the
position vector of the primitive cell, denoted by d for the th atom. Thus,
the equilibrium position of the th atom within the ith primitive cell in a
crystal is given by Ri + d .
Shown in Fig. 1.1 is a simple cubic crystal with a two-atom basis. The
CsCl crystal has such a structure. Because atom 1 in a primitive cell locates
at the tip of the position vector of the primitive cell, its position vector is
zero in the local Cartesian coordinate system associated with the primitive
cell and is thus not shown in the figure.
With the displacement of an atom from its equilibrium position taken
into account, the instantaneous position ri of the th atom within the ith
primitive cell is given by
ri = Ri + d + ui .

(1.11)

In components, the above equation reads


ri, = Ri + d + ui, .

(1.12)

The lattice potential energy in Eq. (1.10) is now expressed as


=

1 X
(Ri + d Rj d + ui uj )
2

(1.13)

i6=j

for a crystal with a multi-atom basis. Taylor-expanding (Ri + d Rj


d + ui uj ) in terms of ui uj about Ri + d Rj d and keeping

han

July 31, 2012

11:38

World Scientific Book - 9in x 6in

Julia 8556-Quantum Theory of Solids

han

Lattice Dynamics

d2 ui2
ui1
i

ri1
y

ri2
Ri

O
x

x
(a)

(b)

(c)

Fig. 1.1 Lattice vibrations in a simple cubic crystal with a multi-atom basis. (a) Static
lattice. To indicate that the atoms of the second kind locate at the centers of the cubes
(the primitive cells), the body diagonals of one cube are drawn. (b) Dynamic lattice.
The ith primitive cell is shaded and marked by i close to its rear lower-left corner that
is chosen as the origin of the local Cartesian coordinate system associated with the
primitive cell. The global Cartesian coordinate system is also shown. (c) Description of
atomic positions. Shown are the position Ri of the ith primitive cell, the position d2
of the second atom within the primitive cell, the displacements ui1 and ui2 , and the
instantaneous positions ri1 and ri2 of the two atoms within the primitive cell. Note
that the position vector d1 of the first atom is zero within the local Cartesian coordinate
system associated with the primitive cell and is not shown.

only terms up to the second order, we have


(Ri + d Rj d + ui uj )
X
(Ri +d Rj d )+
(Ri +d Rj d )(ui, uj, )

1X
(ui, uj, ) (Ri +d Rj d )(ui, uj, ), (1.14)
+
2

where and denote the first- and second-order partial derivatives of


with respect to components of a lattice vector
(Ri + d Rj d )
,
Ri
2 (Ri + d Rj d )
(Ri + d Rj d ) =
.
Ri Ri
(Ri + d Rj d ) =

(1.15)

The harmonic lattice potential energy is then given by


1XX
(ui, uj, ) (Ri +d Rj d )(ui, uj, ),
harm = 0 +
4
i6=j

(1.16)

July 31, 2012

11:38

World Scientific Book - 9in x 6in

10

Julia 8556-Quantum Theory of Solids

A Modern Course in Quantum Theory of Solids

where
0 =

1 X
(Ri + d Rj d )
2

(1.17)

i6=j

is the total cohesive (or lattice) energy. The first-order term vanishes because of the equilibrium condition
X
(Ri + d Rj d ) = 0.
i (i6=j)

We can remove the constraint i 6= j on the dummy summation variables


in Eq. (1.16) upon noticing the presence of the factors (ui, uj, ) and
(ui, uj, ). However, to avoid the possible divergence in (Ri +d
Rj d ), we set it to be identically zero for i = j, which is permissible
because the term with i = j did not appear in Eq. (1.16). We can then
rearrange harm as follows
1X X
harm = 0 +
(ui, uj, ) (Ri +d Rj d )(ui uj, )
4 i, j

1X X
= 0 +
(Ri +d Rj d )(ui, ui, ui, uj, )
2 ij
,
 X

1X X
= 0 +
(Ri + d Rj 0 d 0 ) ij
2 ij
,
j0 0

(Ri + d Rj d ) ui, uj,
= 0 +

1X X
ui, D, (Ri Rj )uj, ,
2 ij

(1.18)

where we have introduced matrix D(Ri Rj ) whose (, )th element is


given by
X

0
0
D, (Ri Rj ) =
(Ri + d Rj d ) ij
j0 0

(Ri + d Rj d ).

(1.19)

Note that the row of D is indexed by the combination of and and


the column by the combination of and . Thus, for a three-dimensional
crystal with a p-atom basis, D is a 3p3p matrix. Note also that the dependence on the positions of atoms within a primitive cell has been transferred
into the subscripts. The Fourier transform of D(Ri Rj ) with respect

han

July 31, 2012

11:38

World Scientific Book - 9in x 6in

Julia 8556-Quantum Theory of Solids

han

11

Lattice Dynamics

to Ri Rj is a very important quantity from which the dynamics of the


lattice vibrations in the crystal can be inferred.
Because of the presence of the basis, not all elements of D(Ri Rj )
are even functions of Ri Rj . However, it still has several other useful
properties.
(1) As indicated by its argument, D, (Ri Rj ) depends on Ri and Rj
only in the form of Ri Rj .
(2) From its definition in Eq. (1.19), it is seen that D, (Rj Ri ) =
D, (RiRj ). It also holds that D, (Rj Ri ) = D, (RiRj ).
(3) The summation of D, (Ri Rj ) over i or j vanishes
X
D, (Ri Rj ) = 0.
(1.20)
j

1.3

Normal Modes of a Three-Dimensional Crystal with a


Multi-Atom Basis

Having discussed the harmonic lattice potential energy of a crystal, we now


turn to solving the problem of lattice vibrations for the crystal by finding
the normal modes of its lattice vibrations using classical mechanics. We
begin with setting up the classical equations of motion for atoms in the
crystal.
1.3.1

Equations of motion of atoms

Differentiating the harmonic lattice potential energy for a three-dimensional


crystal with a multi-atom basis in Eq. (1.18) with respect to ui, , we
obtain the th component of the force exerting on the th atom within the
ith primitive cell due to all other atoms in the crystal
harm
ui,
1 X X
=
ui0 0 , 0 D0 , 0 (Ri0 Rj )uj,
2 ui, 0 0 0

Fi, =

i j ,

1X X 
=
D0 , 0 (Ri0 Rj )uj, ii0 0 0
2 0 0 0
i j ,

+ ui0 0 , 0 D0 , 0 (Ri0 Rj )ij
X
D, (Ri Rj )uj, .
=
j

(1.21)

July 31, 2012

11:38

12

World Scientific Book - 9in x 6in

Julia 8556-Quantum Theory of Solids

A Modern Course in Quantum Theory of Solids

It follows from Newtons second law m u


i, = Fi, that
X
D, (Ri Rj )uj, ,
m u
i, =

(1.22)

where m is the mass of the th atom in a primitive cell. Note that there
is an equation of the above form for each atom in the crystal and for each
coordinate component of the displacement of each atom. Thus, we have
3N p equations in three dimensions. The solutions to the above equations
are to be found by expressing ui, in terms of its Fourier components
X
1

ui, (t) =
Q (k, )ei(kRi t)
(1.23)
N
m

k
with N the total number of primitive cells in the crystal. The fact that
ui, (t) takes only on real values leads to the property that Q (k, ) =
Q (k, ) for the Fourier coefficient Q (k, ) .
1.3.2

Allowed values of wave vector k

The allowed values of k are to be found from the Born-von Karman boundary condition that, for a crystal with a multi-atom basis, is stated as follows
ui1 +N1 , i2 i3 , , (t) = ui1 , i2 +N2 , i3 , , (t) = ui1 i2 , i3 +N3 , , (t)
= ui1 i2 i3 , , (t),

(1.24)

where N1 , N2 , and N3 are respectively the numbers of primitive cells along


basis vectors a1 , a2 , and a3 .
The above equation indicates that the crystals wraps itself up in all
three directions of a1 , a2 , and a3 . Inserting Eq. (1.23) into Eq. (1.24)
yields
eiN1 ka1 = eiN2 ka2 = eiN3 ka3 = 1.

(1.25)

Since the above equations do not change if k is changed by any reciprocal


lattice vector K, we restrict k to be within the first Brillouin zone with
the understanding that the wave vectors differing only by reciprocal lattice
vectors are all equivalent. To infer the allowed values of k from the above
equation, we express k as
k = x1 b1 + x2 b2 + x3 b3 ,
where 0 6 |x1 |, |x2 |, |x3 | 6 1 and b1 , b2 , and b3 are the primitive vectors
of the reciprocal lattice. Upon making use of the orthonormality relation
between bi and aj , bi aj = 2ij , we have
ei2N1 x1 = ei2N2 x2 = ei2N3 x3 = 1

han

July 31, 2012

11:38

World Scientific Book - 9in x 6in

Julia 8556-Quantum Theory of Solids

Lattice Dynamics

han

13

from which it follows that


x1 = n1 /N1 , n1 = 0, 1, 2, , (N1 /2 1), N1 /2,

x2 = n2 /N2 , n2 = 0, 1, 2, , (N2 /2 1), N2 /2,

x3 = n3 /N3 , n3 = 0, 1, 2, , (N3 /2 1), N3 /2.


The allowed values of k are then given by
k = (n1 /N1 )b1 + (n2 /N2 )b2 + (n3 /N3 )b3

(1.26)

with ni = 0, 1, 2, , (Ni /2 1), Ni /2 for i = 1, 2, 3. From the


value ranges of n1 , n2 , and n3 , we see that the total number of the allowed
values of k is N = N1 N2 N3 , the total number of primitive cells in the
monatomic crystal. This statement is applicable to any crystal regardless
of its dimensionality and no matter whether or not it has a basis.
1.3.3

Allowed values of frequency

We now find the allowed values of frequency . Substituting Eq. (1.23) into
Eq. (1.22) yields

Q (k, ) X i(kRi t)
m 2
e
N
k
XX
Q (k, ) i(kRj t)
=
D, (Ri Rj )
e
,
N m
k j

X
X
1
2 Q (k, ) =
D, (Ri Rj )eik(Ri Rj ) Q (k, ),

m m j

X

D, (k) 2 Q (k, ) = 0,
(1.27)

where D, (k) is the dynamical matrix for a three-dimensional crystal


with a multi-atom basis and is given by
X
1
D, (Ri Rj )eik(Ri Rj ) .
D, (k) =
m m j

(1.28)

To be able to infer some conclusions without solving the above equations


explicitly, we must acquire the necessary knowledge on the properties of the
dynamical matrix. We now show that it is a Hermitian matrix. Taking the
Hermitian conjugation of D(k) and making use of D, (Rj Ri ) =

July 31, 2012

11:38

World Scientific Book - 9in x 6in

14

Julia 8556-Quantum Theory of Solids

A Modern Course in Quantum Theory of Solids

D, (Ri Rj ), we have

D,
(k) =

X
1
D, (Ri Rj )eik(Ri Rj )
m m j

X
1
D, (Rj Ri )eik(Ri Rj )
m m j
X
1
=
D, (Ri Rj )eik(Ri Rj ) ,
m m j

(1.29)

that is,

D,
(k) = D, (k) or D (k) = D(k).

(1.30)

To arrive at the final result on the third line in Eq. (1.29), we have set
Rj Ri Ri Rj .
We now go back to the equations in Eq. (1.27). First of all, they imply
that the squares of the frequencies of the normal modes are the eigenvalues
of the dynamical matrix. The Hermitian property of the dynamical matrix
guarantees that all the solutions of 2 are real. They are also nonnegative
for stable crystals. Since these equations are homogeneous linear equations
for Q (k, )s, the secular equation for the determination of frequencies
follows from the sufficient and necessary condition for the existence of nontrivial solutions
det |D, (k) 2 | = 0.

(1.31)

The above equation is an algebraic equation of order 3p with p the number


of atoms in the multi-atom basis. Thus, it has 3p different solutions for
2 and 6p different solutions for at each wave vector k if no degeneracy
occurs. Hence, there are 3p branches of normal modes. The Latin letter s
is used as the branch variable. The frequency in branch s will be denoted
by ks . Since there are N different allowed values of k, there are in total
3pN normal modes of lattice vibrations in a crystal with a p-atom basis.
Note that degeneracy may occur in some regions of the first Brillouin zone.
Taking into account the fact that there are in total 6p allowed values of
, we can express the coefficient Q (k, ) in the expansion of ui, (t) in
Eq. (1.23) as follows
Q (k, ) =

3p
X

s=1


Q (k, ks )ks + Q (k, ks ),ks .

(1.32)

We now find out how many branches among the 3p branches are acoustical branches and how many are optical branches. For this purpose, we

han

July 31, 2012

11:38

World Scientific Book - 9in x 6in

Julia 8556-Quantum Theory of Solids

Lattice Dynamics

han

15

study the dynamical matrix at k = 0. Setting k to zero in Eq. (1.28) and


making use of Eq. (1.19), we obtain
X
1
D, (0) =
D, (Ri Rj ).
(1.33)
m m j
Setting k = 0 in Eq. (1.31), we obtain the equation for determining frequencies at k = 0
det |D, (0) 2 | = 0.

(1.34)

Noticing that the above equation is just the eigenequation for D(0), we
see that the zero eigenvalues of D(0) correspond to acoustical branches.
Therefore, the number of acoustical branches is given by the dimension
of D(0) less its rank. The rank of D(0) can be found by the Gaussian
elimination method in linear algebra. With the rows and columns of D(0)
indexed in the order = 1x, 1y, 1z, 2x, 2y, 2z, , px, py, pz, matrix
D(0) takes on the following form
D11, xx X D11, xy X D11, xz

m1
m1
m1
j
j
j
X
D21, xx X D21, xy X D21, xz

m2 m1 j
m2 m1 j
m2 m1
j

..
..
..

.
.
.

X Dp1, xx X Dp1, xy X Dp1, xz

mp m1 j
mp m1 j
mp m1
j
X

X D1p, xx X D1p, xy X D1p, xz

m1 mp j
m1 mp j
m1 mp

j
X D2p, xx X D2p, xy X D2p, xz

m
m
m
m
m
m
p
p
p
2
2
2

j
j
j

..
..
..

..

.
.
.
.

X Dpp, xx X Dpp, xy X Dpp, xz

mp
mp
mp
j
j
j

p
If we multiply the th column for = 1, 2, , p 1 by m /mp ,
respectively, and then add the results to the pth column, we obtain the
following result on the th row in the pth column
X
1
D, (Ri Rj ) = 0

m mp j
for = x, y, z, where we have made use of the property of D, (Ri Rj )
in Eq. (1.20). Therefore, the last three columns of D(0) have been brought
to zero through the elementary column operations to matrix D(0). After
this, no additional columns can be brought to zero because of the absence
of Dp, (Ri Rj ) for = x, y, z in matrix D(0). Therefore, the rank of
D(0) is 3p 3. This implies that three acoustical branches are present in a
three-dimensional crystal with a p-atom basis and that the remaining 3p 3
branches are optical branches. This conclusion is verified in Fig. 1.2 by
the experimental results of inelastic neutron scattering on an NaCl crystal

11:38

World Scientific Book - 9in x 6in

16

Julia 8556-Quantum Theory of Solids

A Modern Course in Quantum Theory of Solids

that has a two-ion basis. Figure 1.2 shows that NaCl has three acoustical
branches (one longitudinal and two transverse acoustical branches, that
is, one LA branch and two TA branches) and three optical branches (one
longitudinal and two transverse optical branches, that is, one LO branch
and two TO branches). Note that the two transverse acoustical branches
are degenerate along the directions (, 0, 0) and (, , ) and so are the two
transverse optical normal modes.
LO

30
h ( meV )

July 31, 2012

LO

LO
TO1

2TO

20

2TO

TO2

LA

LA
2TA

10

(, 0,0)

LA

TA2

2TA

TA1

(,, 0)

(,, )

0.5

Fig. 1.2 Dispersion relations of the normal modes in an NaCl crystal at 80 K. The
symbols denote experimental data of inelastic neutron scattering by Raunio et. al. [G.
Raunio, L. Almqvist, and R. Stedman, Physical Review 178, 1496 (1969)]. The lines
represent cubic-spline interpolations of the experimental data.

From the above results we can infer that, for a three-dimensional


monatomic crystal, there are three branches of normal modes and they are
all acoustical branches. If a monatomic crystal is taken as a crystal with a
one-atom basis for the purpose of counting branches of normal modes, we
see that the above conclusion is also applicable to a monatomic crystal.
Taking into account the facts that a one-dimensional crystal of inert
gas atoms has only one branch of acoustical normal modes and that a onedimensional ionic crystal has one branch of acoustical normal modes and
one branch of optical normal modes, we can draw a general conclusion that
there is (are) d acoustical branch(es) and d(p 1) optical branch(es) in a
d-dimensional crystal with a p-atom basis.
The above conclusion can be also stated in terms of the numbers of
acoustical and optical normal modes. In a d-dimensional crystal of size
of N primitive cells with a p-atom basis, there are dN acoustical normal
modes and d(p 1)N optical normal modes.

han

July 31, 2012

11:38

World Scientific Book - 9in x 6in

Julia 8556-Quantum Theory of Solids

17

Lattice Dynamics

1.3.4

han

Polarization vectors

Having discussed the frequencies of all normal modes, we now study


the solutions for the Fourier coefficients Q (k, ) in Eq. (1.23). Since
(s)
Q (k, ks ) for normal mode ks can only be determined within a multiplicative factor from Eq. (1.27), we set
(s)

Q(s)
(k, ks ) = Q(k, ks ) (k)
(s)

(s)

(s)

(1.35)
(s)

and demand that the vector  (k) = (1 (k), 2 (k), 3 (k)) be normalized. Inserting the above expression into Eq. (1.27) and specializing
(s)
Eq. (1.27) for branch s, we obtain the equations for  (k)s
X
 (s)
2
D, (k) ks
 (k) = 0.
(1.36)

(s)
 (k)

The vector
is referred to as the polarization vector of normal
mode ks on atom . The polarization vectors possess the following properties

(s)
(k) = (s)

(k),
X

(s0 )
(s)
(k) (k) = ss0 ,

(1.37)
(1.38)

(s)

(s)
(k) (k) = .

(1.39)

Equation (1.37) indicates that the effect of taking the complex conjugation
of a polarization vector is equivalent to taking the inversion of its wavevector variable k in k-space. Eq. (1.38) gives us the orthonormality relation
of polarization vectors. Eq. (1.39) gives us the completeness relation of
polarization vectors.
1.3.5

Displacements of atoms

We can derive an expression for the displacement of an atom in a threedimensional crystal with a multi-atom basis from Eqs. (1.23), (1.32),
and (1.35). We have
X
1
ikRj
uj, (t) =
qks (t)(s)
,
(1.40)
(k)e
N m ks
where qks (t)s are the generalized coordinates of normal modes, referred to
as the normal coordinates, and are given by
qks (t) = Q(k, ks )eiks t + Q(k, ks )eiks t .

(1.41)

July 31, 2012

11:38

World Scientific Book - 9in x 6in

18

Julia 8556-Quantum Theory of Solids

A Modern Course in Quantum Theory of Solids

Using the above expression and Q (k, ks ) = Q(k, ks ), we can verify


that qks (t) possesses the following property

qks
(t) = qks (t).

1.3.6

(1.42)

Hamiltonian of a crystal with a multi-atom basis

To derive the Hamiltonian for a monatomic crystal, we first express its kinetic and interaction potential energies in terms of the above-introduced
normal coordinates. Making use of the expression of the displacement
uj, (t) of an atom in Eq. (1.40), we have for the kinetic energy
X1
1 X X
(s0 )
0 i(k+k0 )Rj
T =
qks (t)qk0 s0 (t)(s)
m u 2j, (t) =
(k) (k )e
2
2N 0 0 j
j
kk ss

1 XX
1X
(s0 )
=
qks (t)qks (t),
qks (t)qks0 (t)(s)
(k) (k) =
2
2
0
kss

(1.43)

ks

where we have made use of the orthonormality relation of the polarization


vectors given in Eq. (1.38). With the constant term omitted and in terms of
the normal coordinates, the harmonic lattice potential energy in Eq. (1.18)
is given by
1X 2
harm =
ks qks (t)qks (t).
(1.44)
2
ks

The Lagrangian of the crystal is then given by


1X 2
1X
L = T harm =
qks (t)qks (t)
ks qks (t)qks (t).
2
2
ks

(1.45)

ks

To obtain the Hamiltonian of the crystal, we must first find the momentum
conjugate to qks (t). Differentiating L with respect to qks (t), we have
X
L
1
pks (t) =
=
qk0 s0 (t)qk0 s0 (t)
qks (t)
2 qks (t) 0 0
ks

1 X
=
qk0 s0 (t)k0 k s0 s + qk0 s0 (t)k0 k s0 s
2 0 0
ks

= qks (t) = qks


(t).

(1.46)

The Hamiltonian of the crystal then follows from the above Lagrangian in
the standard way
X
pks (t)qks (t) L
H=
ks

1X 2
1X
pks (t)pks (t) +
ks qks (t)qks (t).
=
2
2
ks

ks

(1.47)

han

July 31, 2012

11:38

World Scientific Book - 9in x 6in

Julia 8556-Quantum Theory of Solids

han

19

Lattice Dynamics

Since the Hamiltonian has been expressed as a sum of the Hamiltonians


of 3pN independent harmonic oscillators, we have hitherto solved the problem of lattice vibrations of a three-dimensional crystal with a multi-atom
basis at the level of classical mechanics. In order to see how good the results
obtained so far, we now compute the contribution of lattice vibrations to
the specific heat (the lattice specific heat) using the above results.

1.4

Classical Theory of the Lattice Specific Heat

For the computation of the lattice specific heat, we first evaluate the internal
energy u per unit volume of the crystal. The internal energy is given by
the sum of the energies of individual harmonic oscillators weighted by the
H
Boltzmann
/Z with = 1/kB T the inverse of temperature and
R Q factor e H
Z =
dq
dp
e
the canonical partition function. The internal
ks
ks
ks
energy per unit volume is then given by
Z Y
1
u=
dqks dpks HeH
ZV
ks
Z Y
1
1 ln Z
=
dqks dpks eH =
.
(1.48)
ZV
V
ks

Our problem then reduces to the evaluation of Z. Making use of Eq. (1.47),
we have
Z Y
P

Z=
dqks dpks e ks (pks pks +ks qks qks )/2
ks

YZ

dqks dpks e(pks pks +ks qks qks )/2 .

ks

The above maneuvers have reduced a 6pN -fold integral into a product

of 2-fold integrals. Note that, because of the relations qks


= qks and

pks = pks , qks


and pks are not independent variables. For our goal is the
evaluation of the lattice specific heat, we do not even need to evaluate the
2-fold integral explicitly. What we need to do is to extract the temperaturedependence of Z from the above equation. This can be easily accomplished
by making a change of variables
p
p
0
qks
= qks , p0ks = pks .
We then have
Z=

1
3pN

YZ
ks

0 0

0 0

0
dqks
dp0ks e(pks pks +ks qks qks )/2 =

 3pN
A
,

July 31, 2012

11:38

20

World Scientific Book - 9in x 6in

Julia 8556-Quantum Theory of Solids

A Modern Course in Quantum Theory of Solids

where the temperature-independent value of the 2-fold integral has been


denoted by A. The internal energy per unit volume is then given by
1
3pN kB T = 3pnkB T
V
from which the lattice specific heat per unit volume follows
u=

(1.49)

u
= 3pnkB ,
(1.50)
T
where p is the number of atoms in a primitive cell and n = N/V is the number of primitive cells per unit volume. For a three-dimensional monatomic
crystal without a multi-atom basis, p = 1. We then have cv = 3nkB . The
result in Eq. (1.50) is the well-known DulongPetit law for the lattice specific heat of solids. Expressing it in joules per kelvin per mole, we have
cv = 3R with R the gas constant, R = 8.314 JK1 mol1 . Expressing it
in calories per kelvin per mole, we have cv = 3R 6 calK1 mol1 with
R 1.986 calK1 mol1 .
Unfortunately, the result in Eq. (1.50) is in consistency with the experiment only in the high-temperature limit. While the above result indicates
that cv is a constant at all temperatures, the experiment reveals that cv
tends to zero essentially in the cubic power of T as T goes to zero. Therefore, classical theory of lattice vibrations is insufficient in explaining the
temperature dependence of the lattice specific heat. To resolve this inconsistency, we now quantize the lattice vibrations.
cv =

1.5

Quantization of Lattice Vibrations

We now quantize the normal modes of lattice vibrations derived in the


classical theory to develop a quantum theory for lattice dynamics. The
quantization process consists of two steps. In the first step, the canonical
quantization scheme is utilized to quantize the normal coordinates and
momenta of normal modes. In the second step, the combinations of the
quantum operators of the normal coordinates and momenta of the normal
modes give rise to new operators, the annihilation and creation operators
of phonons, with phonons being quanta of lattice vibrations. An important
quantity to obtain in the quantization process is the expression of the atomic
displacement in terms of the annihilation and creation operators of phonons.
This expression is referred to as the quantum field operator of the atomic
displacement since it describes the field of the atomic displacement in terms
of quantum operators.

han

July 31, 2012

11:38

World Scientific Book - 9in x 6in

Julia 8556-Quantum Theory of Solids

Lattice Dynamics

han

21

The first quantization is achieved by replacing the classical normal coordinates qks in Eq. (1.41) and the corresponding momenta pks of normal
modes by operators qks and pks that are required to satisfy the commutation relations



 

qks , pk0 s0 = kk0 ss0 , qks , qk0 s0 = pks , pk0 s0 = 0.
(1.51)
Note that qks and pks have the following properties

qks
= qks , pks = pks .
(1.52)
In the framework of the first quantization, the atomic displacements
and the Hamiltonian of the crystal corresponding to Eqs. (1.40) and (1.47),
respectively, are given by
X
1
ikRj
u
j, =
qks (s)
,
(1.53)
(k)e
N m ks
X
1X 2
=1
pks pks +
ks qks qks .
(1.54)
H
2
2
ks

ks

In the second quantization, we introduce the following annihilation and


creation operators of phonons

1/2 

ks
i
a
ks =
qks +
pks ,
2~
ks
(1.55)

1/2 


i
ks
a
ks =
qks
pks .
2~
ks
The operators a
ks and a
satisfy the following commutation relations
 ks 
a
ks , a
k0 s0 = kk0 ss0 ,
(1.56)

 

k0 s0 = 0.
a
ks , a
k0 s0 = a
ks , a
Inverting the expressions in Eq. (1.55), we can express qks and pks as
follows

1/2

~
qks =
a
ks + a
ks ,
2ks
(1.57)
1/2


~ks
a
ks a
ks .
pks = i
2
In terms of the annihilation and creation operators of phonons, the
quantum field operator of the atomic displacements and the Hamiltonian
of the crystal are given by
1/2
X

~
(1.58)
(s)
ks + a
ks eikRj ,
uj, =
(k) a
2N m ks
ks
X

=
H
~ks a
ks a
ks + 1/2 .
(1.59)
ks

July 31, 2012

11:38

22

World Scientific Book - 9in x 6in

Julia 8556-Quantum Theory of Solids

A Modern Course in Quantum Theory of Solids

are given
The eigenvalues and eigenstates of the crystal Hamiltonian H
by
En =

(nks + 1/2)~ks ,

ks

|ni =

Y
ks

|nks i =

Y
ks

nks
1

|0i,
nks ! ks

(1.60)

n = {nks | ks}, nks = 0, 1, 2, .

The time dependence of a


ks and a
ks can be derived through the Heisenberg equation of motion. It is found that
a
ks (t) = eiks t a
ks , a
ks (t) = eiks t a
ks .

(1.61)

Inserting the above expressions for the time dependence of a


ks and a
ks
into Eq. (1.58), we obtain the time-dependent quantum field operator of
the atomic displacements
1/2
X

~
iks t
(s)
a
ks + eiks t a
ks eikRj
u
j, (t) =
(k) e
2N m ks
ks

1/2
X
 (s)
~
=
 (k)ei(kRj ks t) a
ks
2N m ks
ks


i(kRj ks t)
+ (s)
a
ks .
(k)e
(1.62)
1.5.1

Statistics for phonons

In treating finite-temperature problems related to phonons, the statistics


for phonons is an indispensable piece of instrument. Since phonons are
bosons of spin zero, they obey the BoseEinstein statistics. However, we
can directly compute the thermal distribution using the eigenvalues of the
crystal Hamiltonian in Eq. (1.60). According to Boltzmann, the crystal
takes the eigenstate |ni as its state with the probability of eEn /kB T /Z
P
with Z = n eEn /kB T the canonical partition function. We now evaluate
the average number of phonons in the single-phonon state |ksi, hnks i, which
is given by
1 X
nks eEn /kB T
hnks i =
Z n
Y
1 X
=
nks
e(nk0 s0 +1/2)~k0 s0 /kB T .
Z
0 0
{nks }

ks

han

July 31, 2012

11:38

World Scientific Book - 9in x 6in

Julia 8556-Quantum Theory of Solids

Lattice Dynamics

han

23

Making use of standard algebraic manipulations in statistical mechanics,


we have
P


1 nks =0 nks e(nks +1/2)~ks /kB T X Y (nk0 s0 +1/2)~k0 s0 /kB T
P
hnks i =
e

(nks +1/2)~ks /kB T


Z
nks =0 e
{nk0 s0 } k0 s0
P
(nks +1/2)~ks /kB T
1
nks =0 nks e
= ~ /k T
.
= P

(nks +1/2)~ks /kB T


B
ks
e
e
1
nks =0
That is

1
(1.63)
e~ks /kB T 1
which is just what the BoseEinstein statistics gives for phonons. Note
that, because phonons in a crystal are constantly annihilated and created,
their number is not conserved and their chemical potential is zero.
hnks i =

1.6

Phonon Density of States

In phonon-related problems, we often need to perform the summation of


P
the form ks F (ks ) over the phonon wave vector k and branch s, where
F (ks ) depends on k only through the phonon dispersion relation ks .
Such a sum can be converted into an integral over phonon frequencies for
the benefit of reducing a three-dimensional integral (that results from converting the sum over k into an integral over k) to a one-dimensional integral.
This is especially useful in numerical computations. The conversion into
an integral over phonon frequenciesRcan be easily implemented by using the

property of the Dirac -function: d ( ks ) = 1. Inserting this


P
magic one into the summation ks F (ks ), we have
Z
1 X
1 X
F (ks ) =
d ( ks )F (ks )
V
V
ks
ks
 X

Z
1
=
d F ()
( ks )
V

ks
Z
=
d g()F (),
(1.64)

where g() is the phonon density of states, with g()d the number of
phonon states per unit volume in the frequency range from to + d,
and is given by
X Z dk
1 X
( ks ).
(1.65)
( ks ) =
g() =
V
(2)3
s
ks

July 31, 2012

11:38

24

World Scientific Book - 9in x 6in

Julia 8556-Quantum Theory of Solids

A Modern Course in Quantum Theory of Solids

The phonon density of states in branch s Zis given by


1 X
dk
gs () =
( ks ).
( ks ) =
V
(2)3

(1.66)

g() is then the summation of gs () over phonon branches. We can also express the phonon density of states in terms of an integral over the constantfrequency surface. To obtain this expression, we write dk = ddk with
d the area element on the constant-frequency surface S , ks = , and
express ( ks ) in terms of the component k of k perpendicular to the
constant-frequency surface S
(k k 0 )
,
( ks ) =
k ks
where k ks is the derivative in the direction of the normal of the constantfrequency surface S . We then have
Z
Z
Z
dk (k k 0 )
d
1

=

.
d
gs () =
(1.67)
(2)3 k ks
(2)3 S k ks
S

This alternative expression of the phonon density of states can differentiate the importance in the contributions of various normal modes to the
density of states and disclose the singularities in the dispersion relations.
If k ks = 0 at some particular wave vector k0 , this expression indicates
that the vicinity around k0 makes an important contribution to the phonon
density of states since the integrand diverges at k0 . This leads to a peak
in the phonon density of states at the corresponding frequency. Such a
frequency is known as a van Hove singularity. The wave vectors that contribute to van Hove singularities are referred to as critical points of the first
Brillouin zone.
1.7

Lattice Specific Heat of Solids

Since lattice vibrations (phonons) contribute to a variety of physical properties of solids, the results obtained in the previous section find their extensive
applications in these properties. Here we concentrate on the phonon contribution to the specific heat (the lattice specific heat) of solids since the
specific heat of solids is one of the few problems that first gave us hints on
the inaccuracy of the classical theory in its description of the microscopic
world.
We will first derive a general expression for the lattice specific heat using
the eigenvalues of the crystal Hamiltonian in Eq. (1.60). We will then study
the Debye and Einstein models for the lattice specific heat.

han

July 31, 2012

11:38

World Scientific Book - 9in x 6in

Julia 8556-Quantum Theory of Solids

Lattice Dynamics

1.7.1

han

25

General expression of the lattice specific heat

We follow the standard approach in thermodynamics for the computation


of the lattice specific heat. We first derive the internal energy of the crystal.
The lattice specific heat is then computed from the internal energy. The
thermal distribution function in Eq. (1.63) gives us the average phonon
number in the single-phonon state |ksi. Since each phonon in the singlephonon state |ksi carries an energy of ~ks , the internal energy u per unit
volume of the crystal is given by
1 X
1 X
~ks
u=
hnks i ~ks =
V
V
e~ks /kB T 1
ks
ks
X Z dk
~ks
=
,
(1.68)
(2)3 e~ks /kB T 1
s
where the k-integration is over the first Brillouin zone of the crystal. The
lattice specific heat per unit volume cv is then given by
Z
X
u
dk
~ks
.
(1.69)
=
cv =
3
~
/kB T 1
ks
T
T s
(2) e
The above equation is referred to as the general expression for the lattice specific heat . To compute the lattice specific heat using the above
expression, we must know the dispersion relations of the normal modes
(the phonon dispersion relations). From our previous experience, we know
that it is a great challenge to compute the phonon dispersion relations for
real crystals. In any event, if the phonon dispersion relations are known,
Eq. (1.69) can be then utilized to compute the lattice specific heat per unit
volume. However, even without knowing the explicit phonon dispersion relations, we can still evaluate approximately the lattice specific heat in the
high- and low-temperature limits.
1.7.2

High-temperature limit

In the high-temperature limit, ~ks /kB T  1. We can then expand the


BoseEinstein distribution function in Eq. (1.69) as follows
1
1
=
~ks /kB T + (~ks /kB T )2 /2! + (~ks /kB T )3 /3! +
e~ks /kB T 1
1
~ks
kB T
+
+ .
=
~ks
2 12kB T
The second term is a constant and does not contribute to the lattice
specific heat. The contributions from the third and other higher-order

July 31, 2012

11:38

26

World Scientific Book - 9in x 6in

Julia 8556-Quantum Theory of Solids

A Modern Course in Quantum Theory of Solids

terms are much smaller than the contribution from the first term because
~ks /kB T  1 and they are quantum corrections to the result from the
first term alone. Retaining only the first term in the above expansion, we
have
Z
X
kB T
dk
lim cv
~ks
T
T s
(2)3
~ks
X Z dk
= kB
= 3pkB /vc = 3pnkB ,
(1.70)
(2)3
s
where n = 1/vc = N/V is the number of primitive cells per unit volume
and p the number of atoms in a primitive cell. For a three-dimensional
monatomic crystal without a multi-atom basis, we have cv = 3nkB . The
result in Eq. (1.70) is just the DulongPetit law and it indicates that the
DulongPetit law is valid only at high temperatures.

1.7.3

Low-temperature limit

At low temperatures, the probabilities for the normal modes of high frequencies to be occupied by phonons are extremely small. Thus, we can take
only the normal modes of low frequencies into account in the computation
of the lattice specific heat at low temperatures. Since the optical normal
modes are of high frequencies in comparison with the acoustical phonons,
their contributions are neglected. For the acoustical normal modes, only
those of low frequencies make substantial contributions. From the computations of the dispersion relations of the normal modes in the last chapter,
we know that the dispersion relation for acoustical normal modes of low
frequencies can be well approximated by a linear dependence on the wave

is the speed of sound that depends only


number, ks cs (k)k,
where cs (k)
and does not on the magnitude of k.
on the direction of k (denoted by k)

cs (k)k/k
BT
Because e
becomes even smaller for large values of k, the error
introduced by extending the k-integration in Eq. (1.69) from over the first
Brillouin zone to over the entire reciprocal space is negligibly small at low
temperatures. We thus extend the region of the k-integration in Eq. (1.69)
to the entire reciprocal space. With the above-introduced simplifications,
the lattice specific heat of a solid at low temperatures is given by
Z
Z
3
1 X dk
~cs (k)k
dk
lim cv

T 0
BT 1
22 T s
4 0
e~cs (k)k/k

3 Z
6 kB T
x3
kB
,
= 2
dx x

~c
e 1
0

han

11:38

World Scientific Book - 9in x 6in

Julia 8556-Quantum Theory of Solids

han

27

Lattice Dynamics

where
1
1X
=
3
c
3 s

dk 1

4 c3s (k)

(1.71)

is the average of the inverse of the cubed speeds of sound of the normal
modes of the three acoustical branches. The remaining integral can be
performed by first multiplying the numerator and denominator by ex and
then expanding 1/(1 ex ) as a Taylor series
6
lim cv 2
T 0

6
2

6
= 2

kB T
~c

3

x3 ex
1 ex

kB

kB T
~c

3

kB

Z
X

kB T
~c

3

6
22
kB
=
n4
5
n=1

dx

n=1

dx x3 enx

kB T
~c

3

kB ,

(1.72)

P
4
4
where we have made use of
n=1 1/n = 90/ . This is a remarkable
result! It implies that the lattice specific heat tends to zero cubically as
the temperature goes to zero, in excellent agreement with the experiment.
The problem of the lattice specific heat at low temperatures has thus been
solved with the quantization of lattice vibrations! The experimental data
of the specific heat of diamond at low temperatures are given in Fig. 1.3
together with a fit to cv = AT 3 .

cv ( calK -1 mol -1 )

July 31, 2012

0.03
0.02
0.01
0.0
0

20

40
T [K]

60

80

Fig. 1.3 Low-temperature specific heat of diamond. The open circles represent the
experimental data [W. DeSorbo, Journal of Chemical Physics 21, 876 (1953)]. The solid
line is a fit to cv = AT 3 with A = 4.774 108 calK4 mol1 .

July 31, 2012

11:38

28

World Scientific Book - 9in x 6in

Julia 8556-Quantum Theory of Solids

A Modern Course in Quantum Theory of Solids

Since a diamond crystal is an insulator (a semiconductor with a large


band gap), its low-temperature specific heat consists of only the lattice specific heat. From Fig. 1.3, it is seen that the specific heat of diamond at low
temperatures indeed follows the T 3 -power law and that the experimental
data is well fitted to cv = AT 3 with A = 4.774 108 calK4 mol1 .
1.8

Debye Model

To evaluate the contribution of the lattice vibrations to the specific heat


of a solid, Debye put forward his model for lattice vibrations in 1912. In
the Debye model, only three acoustical branches are used to describe all
the lattice vibrations in a solid. It is assumed that all the normal modes
in the three acoustical branches have the same speed of sound c, that the
dispersion relation is linear in the wave number k, = ck, and that there
exists an upper limit (called the Debye wave vector and denoted by kD ) for
the wave number. The maximum wave number kD is determined through
demanding that the number of acoustical normal modes in the model be
equal to the actual number of acoustical normal modes in the crystal. The
quantity D = ckD is called the Debye frequency. In the Debye model,
the contribution of optical normal modes to the specific heat is taken into
account through high-frequency acoustical normal modes.
For the convenience of finding expressions for kD and D , we first assume
that they are known and derive the phonon density of states in the Debye
model. From Eq. (1.65), we have in the Debye model
Z kD
3 X
3
gD () =
( ck) =
dk k 2 ( ck)
V
22 0
k

3 2
() (D ),
=
22 c3

(1.73)

where (x) is the step function, (x) = 1 for x > 0, = 0 for x < 0. Note
that the phonon density of states in the Debye model is quadratic in for
0 < 6 D (this is a characteristic of the Debye model) and that it is zero
for < 0 or > D .
We now find expressions for kD and D . For a three-dimensional
monatomic crystal without a multi-atom basis, the total number of acoustical normal modes is 3N with N the number of primitive cells.
R The number
of acoustical normal modes in the Debye model is given by V d gD ().

han

July 31, 2012

11:38

World Scientific Book - 9in x 6in

Julia 8556-Quantum Theory of Solids

han

29

Lattice Dynamics

We thus have

Z
3V
d gD () =
3N = V
d 2 () (D )
22 c3

Z D
3
3V
D
V
2
=
d

=
(1.74)
2
3
2
2 c 0
2 c3
Z

from which it follows that


D = 62 n

1/3

c, kD = 62 n

1/3

(1.75)

where n = N/V is the number of primitive cells per unit volume.


We now compute the lattice specific heat within the Debye model. From
the general expression for the lattice specific heat in Eq. (1.69), we have
cD
v =

3
22 T

= 9nkB

kD

dk



~ck 3
e~ck/kB T

1
3 Z D /T

T
x3
dx x
,
T
D
e 1
0

(1.76)

where D = ~ckD /kB = ~D /kB is the Debye temperature. The Debye


temperature D has since been used to characterize crystals. For a plot of
cD
v versus temperature, see Fig. 1.6. In general, D depends on temperature
[see below for a more detailed discussion]. Unfortunately, the specific heat in
Eq. (1.76) can not be given in a closed form at an intermediate temperature.
However, the closed forms can be approximately obtained at high and low
temperatures.
1.8.1

High-temperature limit

In the high-temperature limit, since D /T  1, the values of the integration variable x are very small in the entire integration interval. We can then
expand the exponential function ex in the denominator of the integrand in
Eq. (1.76) as a Taylor series and retain only the first two terms. We then
have
cD
v = 9nkB



T
D

3 Z
T

D /T

dx x2

= 3nkB .

(1.77)

We have thus recovered the DulongPetit law in the high-temperature limit.

July 31, 2012

11:38

30

1.8.2

World Scientific Book - 9in x 6in

Julia 8556-Quantum Theory of Solids

A Modern Course in Quantum Theory of Solids

Low-temperature limit

Since D /T  1 in this limit, the upper limit of the integral in Eq. (1.76)
can be extended to infinity. We then have

3 Z


x3
T
cD
=
9nk
T
dx
B
v
T
D
ex 1
0


3

3
124 T
T
=
nkB 234
nkB ,
(1.78)
5
D
D
where the result for the integral in Eq. (1.72) has been used. Hence, the
lattice specific heat at low temperatures also follows the T 3 -power law in
the Debye model. The success of the Debye model at low temperatures lies
at the physical fact that only low-frequency acoustical single-phonon states
are occupied with appreciable probabilities at low temperatures.
1.8.3

Debye temperature

As mentioned in the above, the Debye temperature has been used to characterize a solid. As a matter of fact, it is one of the most important characteristics of a solid. It reflects the density, structural stability, and bonding
strength of the solid. Structure defects in a solid can be also identified
through the variation in its Debye temperature. The Debye temperature
is also the characteristic energy scale of phonons in the solid and used in
comparison of energy scales with other elementary excitations. The magnitudes of the Debye temperature vary widely among solids: It can be as
large as over 2, 000 K, such as in diamond, and as small as below 40 K,
such as in cesium. The typical value of the Debye temperature D can be
taken as several hundred Kelvins. Since the Fermi temperature F of the
electron gas in a metal is typically several ten thousand Kelvins, the ratio
D /F is typically of the order of 102 . Thus, the energy scale of phonons
in a metal is very small compared to that of electrons. This fact will be
extensively exploited in the study of the electronphonon interaction.
The Debye temperature of a solid can be inferred from several different
physical quantities of the solid, such as the entropy, the specific heat, the
speed of sound, the elastic constants, and etc.
The Debye temperature of a solid in general varies with temperature.
The variation is large in some solids and small in others. The temperature dependence of the Debye temperature of a perfect crystalline solid is
chiefly caused by the electronphonon interaction and the anharmonicity

han

11:38

World Scientific Book - 9in x 6in

Julia 8556-Quantum Theory of Solids

han

31

Lattice Dynamics

in lattice vibrations. The temperature dependence of the Debye temperature in diamond is shown in Fig. 1.4 from which it is seen that the Debye
temperature in diamond is high and that its variation is large. The Debye
temperature peaks at about 60 K with a peak value of about 2, 250 K. It is
about 1, 850 K at 25 K and 1, 870 K at 300 K. The large Debye temperature
in diamond leads to a small lattice specific heat in diamond as shown in
Fig. 1.3.
2300
2200
D [K ]

July 31, 2012

2100
2000
1900

100

200

300

T [K]
Fig. 1.4 Debye temperature as a function of temperature in diamond [W. DeSorbo,
Journal of Chemical Physics 21, 876 (1953)].

The Debye temperatures of alkali metals are small compared to the Debye temperature of diamond, with lithium having the largest Debye temperature (about 375 K) among the alkali metals and its Debye temperature not
varying appreciably with temperature. The Debye temperatures of the remaining alkali metals, sodium, potassium, rubidium, and cesium, are shown
in Fig. 1.5 as functions of temperature. It is seen the Debye temperatures
of these alkali metals do not vary much with temperature, either.
1.9

Einstein Model

Performing the derivative with respect to temperature T in the general


expression of the lattice specific heat in Eq. (1.69), we obtain
cv =

kB X
kB X (~ks /2kB T )2
=
E(~ks /2kB T ),
2
V
V
sinh (~ks /2kB T )
ks
ks

(1.79)

11:38

World Scientific Book - 9in x 6in

32

Julia 8556-Quantum Theory of Solids

A Modern Course in Quantum Theory of Solids

150
D [ K ]

July 31, 2012

100

50

10

100

200 300

T [K]
Fig. 1.5 Debye temperature as a function of temperature for alkali metals Na, K, Rb,
and Cs (from top to bottom) from T = 1 to 300 K [D. L. Martin Physical Review 139,
150 (1965)].

where we have converted the integration over k into a summation over k


and introduced an auxiliary function E(x) given by
E(x) =

x2
.
sinh2 (x)

(1.80)

The function E(x) is called the Einstein function. For the convenient evaluation of the contribution of the optical phonons to the lattice specific
heat, Einstein treated the optical normal modes as independent harmonic
oscillators and assumed that they have the identical frequency E . This is
the well-known Einstein model for the lattice specific heat. Note that the
acoustical phonons are not taken into account in the Einstein model. Because the optical phonons all have nonzero frequencies, the lattice specific
heat given by the Einstein model has an incorrect temperature dependence
at low temperatures.
The lattice specific heat per unit volume in the Einstein model is simply
given by
cE
v =

popt N kB
(E /T )2 eE /T
E(~E /2kB T ) = popt nkB
2 ,
V
eE /T 1

(1.81)

where E = ~E /kB is the Einstein temperature and popt is the number


2 E /T
of optical branches. Note that, as T 0, cE
.
v popt nkB (E /T ) e
E
Although cv 0 as T 0, the temperature dependence is incorrect as
mentioned above with the reason given there. As T , cE
v popt nkB .
If the number of optical branches is equal to three, cE
at
high
temperatures
v

han

11:38

World Scientific Book - 9in x 6in

Julia 8556-Quantum Theory of Solids

han

33

Lattice Dynamics

agrees with that given by the DulongPetit law and with that given by the
Debye model.
The temperature dependence of the lattice specific heats predicted
by the Debye and Einstein models are plotted in Fig. 1.6. For the lattice specific heat given by the Debye model, cv /3nkB is plotted, whereas
cE
v /popt nkB is plotted for the Einstein model.

cvD 3nkB , cvE popt nkB

July 31, 2012

1.0
Debye
0.5

Einstein

0.0
0.0

0.5

1.0

1.5

T D , T E
Fig. 1.6 Lattice specific heats predicted in the Debye and Einstein models as functions
of reduced temperature T /D or T /E . The solid line is for cD
v and the dashed line for
cE
v.

From Fig. 1.6, it is seen that the lattice specific heats from the Debye
and Einstein models both tends to zero as temperature goes to zero and
approach the result given by the DulongPetit law at high temperatures.
D
E
Overall, cE
v is smaller than cv . Note that cv goes to zero much faster than
E
cD
v does. This is due to the erroneous behavior of cv at low temperatures
mentioned in the above.

1.10

Effect of Thermal Expansion on Phonon Frequencies

When temperature varies, a crystal expands or shrinks, which leads to the


variation in phonon frequencies. This is the subject we investigate in this
section. We start from the general description of the thermal expansion.
The dimensionless Gr
uneisen parameter (named after Eduard Gr
uneisen),
(T ), is used to describe the thermal expansion. The Gr
uneisen parameter

July 31, 2012

11:38

World Scientific Book - 9in x 6in

34

Julia 8556-Quantum Theory of Solids

A Modern Course in Quantum Theory of Solids

is defined by
(T ) =

BT
,
cv

(1.82)

where is the volume thermal expansion coefficient, = ln V /T , BT


the isothermal bulk modulus, BT = V P/V = P/ ln V with P the
pressure, and cv the specific heat per unit volume. To derive an explicit
expression for (T
 ), we need to compute the pressure P that is given by
P = F/V T in terms of the Helmholtz free energy F , F = kB T ln Z,
where the canonical partition function Z is given by
X
YX
Y
2
Z=
eEn =
e~ks (nks +1/2) =
. (1.83)
sinh(~
/2kB T )
ks
n
n
ks

ks

ks

Here the eigenvalues of the crystal Hamiltonian in Eq. (1.60) have been
used. The Gr
uneisen parameter is then given by


 


 
1 F
kB
ln Z
(T ) =
=
T
cv T V T V
cv T
V
T V
X
ks E(~ks /2kB T )
=

ks

E(~ks /2kB T )

(1.84)

ks

where the Einstein function E(x) is given in Eq. (1.80) and ks is the mode
Gr
uneisen parameter for normal mode ks and is given by
ln ks
.
(1.85)
ln V
Note that Eq. (1.84) implies that (T ) is a weighted average of the mode
Gr
uneisen parameters with the weight for normal mode ks given by the norP
malized Einstein function: E(~ks /2kB T ) divided by ks E(~ks /2kB T ).
For a one-dimensional crystal, the mode Gr
uneisen parameter is given
by
ks =

ln ks
(1.86)
ln L
with L the length of the one-dimensional crystal. Take a one-dimensional
crystal of inert gas atoms of mass m as an example. The phonon dispersion
relation is given by k = (4K/m)1/2 | sin(ka/2)| for such a one-dimensional
crystal, where K is the force constant. We have
ks =

k =

ln k
= (ka/2) cot(ka/2).
ln a

(1.87)

han

11:38

World Scientific Book - 9in x 6in

Julia 8556-Quantum Theory of Solids

han

35

Lattice Dynamics

Note that the mode Gr


uneisen parameter is negative for all normal modes.
The Gr
uneisen parameter is then given by
X
(ka/2) cot(ka/2)E(~k /2kB T )
(T ) =

E(~k /2kB T )

(1.88)

The values of (T ) at a number of temperatures are evaluated numerically with the results plotted in Fig. 1.7 as a function of T / with
= ~(4K/m)1/2 /kB .

-0.7
-0.8

July 31, 2012

-0.9
-1.0
0.0

0.5

1.0

T
Fig. 1.7 Plot of the Gr
uneisen parameter of a one-dimensional crystal of inert gas atoms
as a function of the reduced temperature T /.

From Fig. 1.7, it is seen that the Gr


uneisen parameter for this onedimensional crystal is negative at all temperatures, which implies that the
normal mode frequencies decrease as the crystal expands.

1.11

Specific Heat of a Metal

The specific heat of a pure metallic crystal consists of the electronic and
lattice specific heats. At low temperatures, the electronic specific heat takes
on the form cev = T with the electronic specific heat coefficient and the
3
lattice specific heat takes on the form cL
v = AT [cf. Eqs. (1.72) and (1.78)].
Thus, the specific heat of a pure metal is given by
3
cv = cev + cL
v = T + AT .

(1.89)

11:38

World Scientific Book - 9in x 6in

36

Julia 8556-Quantum Theory of Solids

han

A Modern Course in Quantum Theory of Solids

cv T [ mcalK -2 mol -1 ]

cv [ m calK -1 mol -1 ]

July 31, 2012

(a)
0.8
0.4
0.0
0

0.5

1.0

1.5

0.6

(b)

0.5
0.4
0.3
0.0

0.5

T [K]

1.0
2

1.5

2.0

2.5

T [K ]

Fig. 1.8 Low-temperature specific heat of sodium. (a) Specific heat cv as a function of
temperature T . The open circles represent the experimental data [D. L. Martin, Physical
Review 124, 438 (1961)]. The solid line is a linear least-squares fit of the experimental
data to cv = T + AT 3 . (b) Specific heat divided by temperature, cv /T , as a function
of T 2 . The open circles represent the same experimental data as in (a) but now cv /T
is plotted as a function of T 2 . The solid straight line is a linear least-squares fit of the
experimental data to cv /T = + AT 2 .

To see how well the form of the low-temperature specific heat in Eq. (1.89)
is obeyed in a real simple metal, we show the low-temperature specific heat
of sodium in Fig. 1.8.
In Fig. 1.8(a), the experimental data of the low-temperature specific
heat of sodium are shown. From the linear least-squares fit of the experimental data to the expression in Eq. (1.89), it is seen that the lowtemperature specific heat of sodium does follow the law prescribed in
Eq. (1.89). Although the coefficients and A can be determined from
the above linear least-squares fit of the experimental data, it has become a custom that the coefficients are determined by plotting the experimental data in the manner of cv /T versus T 2 . Such a plot is shown
in Fig. 1.8(b) together with the linear least-squares fit of the experimental data to cv /T = + AT 2 . In such a plot, the intercept on the vertical axis yields the value for and the slope of the straight line gives
the value for A. It has been found that 0.335 mcalK2 mol1 and
A 0.107 mcalK4 mol1 .
The above example demonstrates that the low-temperature specific heat
of a simple metal follows the law given in Eq. (1.89). However, deviations
from this law are observed in metallic compounds, especially in metallic
compounds that contain d or f elements, such as heavy fermion systems5 .
5 G.

R. Stewart, Reviews of Modern Physics 56, 755 (1984).

July 31, 2012

11:38

World Scientific Book - 9in x 6in

PROBLEMS

Julia 8556-Quantum Theory of Solids

han

37

Problems
1-1 Consider a linear chain in which alternate ions have masses m1 and
m2 and only nearest neighbors interact through a spring of force
constant K. Find the dispersion relations for the normal modes.
Discuss the limiting cases for m1  m2 and m1 = m2 .
1-2 A commonly-seen example of a simple one-dimensional crystal is a
line of point masses, each of which has two nearest neighbors: One
is distance d away and the other distance (a d) away (d  a)
in equilibrium. The two neighboring point masses are connected by
springs, with the force constants of the springs between the neardistanced point masses and between the far-distanced point masses
given by K and G (K > G), respectively.
(1) Write down the harmonic crystal potential energy in terms of the
displacements of point masses from their equilibrium positions.
(2) Set up the classical equations of motion for the point masses.
(3) Solve for the frequencies and polarization vectors of the normal
modes of the lattice vibrations from the classical equations of
motion.
1-3 Consider a one-dimensional crystal of atoms of mass m. Only the
interactions up to the next nearest neighbors are taken into account
and are modeled by springs with the force constant for the nearestneighbor interaction given by K and that for the next-nearestneighbor interaction given by G.
(1) Compute the dispersion relation of the normal modes.
(2) Find the condition on G so that the dispersion curve peaks inside
the first Brillouin zone.
(3) Find the expressions for the group and phase velocities and evaluate them at the peak position of the dispersion curve under the
condition found in (2).
1-4 Consider a linear chain of atoms of mass m with the nearest neighboring atoms connected by springs of force constant K. In addition,
the motion of each atom is damped, with the damping force u j
exerted on the jth atom, where uj is the displacement of the jth
atom from its equilibrium position. Assume that  (mK)1/2 .
(1) Write down the equations of motion of atoms with the damping
taken into account.

July 31, 2012

11:38

38

World Scientific Book - 9in x 6in

Julia 8556-Quantum Theory of Solids

A Modern Course in Quantum Theory of Solids

(2) Find the dispersion relation k


(3) Find the relaxation time of the normal modes.
1-5 Consider a linear chain of polarizable molecules with the nearestneighbor equilibrium distance a. All the molecules are fixed to their
positions. However, each molecule has an internal degree of freedom
that obeys the equation of motion 2 p/t2 = 02 p + E02 , where p
is the electric dipole moment of the molecule (assumed to be parallel
to the chain), E the local electric field due to all other molecules,
and the polarizability. Find the dispersion relation (k) for small
amplitude polarization waves (optical normal modes). Discuss the
dependence of (0) on .
1-6 A triatomic linear chain consists of three different types of atoms
of masses m1 , m2 , and m3 , respectively. As usual, it is assumed
that only nearest-neighboring atoms interact and the interactions are
modeled as being mediated through springs of force constants between atoms of types 1 and 2, between atoms of types 2 and 3, and
between atoms of types 3 and 1. Derive an equation that determines the frequencies of normal modes and describe the properties of
solutions.

ky
/a

/a

Z
/a
kx
X

/a
Fig. 1.9 First Brillouin zone of the two-dimensional square lattice. Three highsymmetry points, , X, and M , and three high-symmetry lines, , , and Z, are
shown.

1-7 Consider a two-dimensional crystal with a square Bravais lattice


whose first Brillouin zone is given in Fig. 1.9. With only interactions
between the nearest and next nearest neighbors taken into account,

han

July 31, 2012

11:38

World Scientific Book - 9in x 6in

Julia 8556-Quantum Theory of Solids

han

39

PROBLEMS

the harmonic lattice potential energy of the crystal is given by


X
X


1
~ i R
~ j )(~ui ~uj ) 2 + 1 G
~ i R
~ j )(~ui ~uj ) 2 ,
= 2K
(R
(R
2a
4a2
hiji

(ij)

where hiji indicates summation over the nearest neighbors and (ij)
summation over the next nearest neighbors. Here a is the lattice
~ i s are Bravais lattice vectors, and ~ui s are deviations of
constant, R
atoms from their equilibrium positions.
(1) Construct the dynamical matrix.
(2) Find the frequencies and polarization vectors of normal modes
along the lines , , and Z, respectively.
(3) Plot the dispersion relations along these three high-symmetry
lines.
1-8 The lattice dynamics of a simple cubic crystal of lattice constant a
and atom mass M is studied here with only interactions between
nearest-neighboring atoms taken into account. The interactions are
modeled as being mediated through springs of force constant .
(1) Write down the potential energy of the crystal and construct the
dynamical matrix.
(2) Solve for the dispersion relations.
(3) Plot the dispersion relations along [100] and label branches properly. Give the physical reason for the zero-frequency normalmode branches.
(4) Indicate the pattern of displacements of eight atoms in the conventional cell for the mode at k = (/a, 0, /a) with displacements along ex .

1-9 Consider a three-dimensional monatomic Bravais lattice in which each


ion of mass M interacts only with its nearest neighbors with the
interaction potential energy given by (ri rj ) = K(|ri rj | d)2 /2,
where d is the equilibrium spacing between the atoms and K the force
constant of the spring connecting the atoms.
(1) Show that the frequencies of the three normal modes at each
wave vector k are given by s (k) = [s (k)/M ]1/2 , where s (k)s
P
are the eigenvalues of the 3 3 matrix D = 2K R6=0 sin2 (k
R
with , = x, y, z.
R/2)R
(2) Now apply the above result to a monatomic FCC crystal. Show
that, if k is in the [100] direction, k = (k, 0, 0), then the frequency of the longitudinal acoustical normal mode is given by

July 31, 2012

11:38

40

World Scientific Book - 9in x 6in

Julia 8556-Quantum Theory of Solids

A Modern Course in Quantum Theory of Solids

L = 2(2K/M )1/2 sin(ka/4) and the frequency of the two degenerate transverse acoustical normal modes is given by T =
2(K/M )1/2 sin(ka/4). Also consider the cases for k in the [110]
and [111] directions
1-10 A three-dimensional crystal has a two-atom basis. The masses of the
two atoms in the basis are m1 and m2 , respectively. Let v1 and v2 be
their velocities. Show that, for an optical normal mode at the center
of the first Brillouin zone (k = 0), m1 v1 + m2 v2 = 0.
1-11 The quantum field operator of atomic displacements for a threedimensional crystal with a multi-atom basis is given in Eq. (1.62).
We now derive the quantum field operator of atomic momenta. In
analogy with the definition of momentum in classical mechanics, let
Pj, (t) = m u
j, (t)/t.
(1) Write down the explicit expression of Pj, (t) in terms of operators a
ks and a
ks .
(2) Show that [
uj, (t), P`, (t)] = i~j` . Therefore, Pj, (t)
is the momentum field operator conjugate to the displacement
field operator u
j, (t).
(3) Show that u
j, (t) and Pj, (t) are Hermitian operators.
1-12 In this problem, the Hamiltonian for a three-dimensional crystal with
a multi-atom basis will be derived.
(1) Specializing the above-obtained expression for Pj, (t) to the
time-independent case and making use of the resultant expression, express the kinetic energy of the crystal, T =
P

ks and a
ks .
j, Pj, Pj, /2m in terms of operators a
(2) Using the expression of uj, , express the harmonic lattice po harm = (1/2) P P
tential energy of the crystal,
j,
j`
, u
D, (Rj R` )
u`, in terms of operators a
ks and a
ks .

(3) Derive the Hamiltonian for a three-dimensional crystal with a


multi-atom basis.
1-13 Take phonons in a crystal as if they move in a box of volume V and
study the thermodynamics of this gas of phonons.
(1) Evaluate the canonical partition function of phonons, Z =
P En /kB T
with En an eigenvalue of the crystal Hamiltonian,
ne
and the Helmholtz free energy F = kB T ln Z.

han

July 31, 2012

11:38

World Scientific Book - 9in x 6in

Julia 8556-Quantum Theory of Solids

han

41

PROBLEMS


(2) From the thermodynamic relation S = F/T V , compute
the entropy S.
(3) Express S in terms of the thermal average of the occupation
number nB (~ks ) hnks i of the single-phonon state |ksi.
1-14 Consider a one-dimensional crystal of inert gas atoms. Let L be the
length of the crystal and N the number of atoms. Let a be the
lattice constant. (i) Evaluate the phonon density of states for this
crystal. (ii) Derive an integral expression for the lattice specific heat
of the crystal. (iii) Evaluate the lattice specific heat in the high- and
low-temperature limits.
1-15 Reconsider the above problem within the Debye model. (i) Find the
phonon density of states within the Debye model. (ii) Determine the
Debye frequency. iii. Find a general expression for the lattice specific
heat. (iv ) Evaluate the lattice specific heat in the high- and lowtemperature limits. (v ) Compare the exact and Debye results for the
lattice specific heat by plotting them together from zero temperature
to the Debye temperature.
1-16 Consider the relative size of the electronic and lattice contributions
to the specific heat of a metal using the Sommerfeld theory for the
electrons and the Debye model for the phonons. (i) Find an expres
sion for the ratio cev /cL
v . (ii) Determine the temperature T at which
e
L
cv = cv . (iii) Give an estimate on the order of magnitude for T in
alkali metals.
1-17 A number of values of the specific heat of potassium at low temperatures are given in Table 1.1. (i) Plot Cv versus T and Cv /T versus
T 2 . (ii) Perform a linear least-squares fit of the experimental data
for Cv /T to Cv /T = + AT 2 and determine and A. (iii) Estimate
the Debye temperature of potassium at low temperatures.
Table 1.1 Low-temperature specific heat of potassium in mJ K1 mol1 [W. H. Lien
and N. E. Phillips, Physical Review 133, A1370 (1964)]. The temperature T is in K.
T
0.260
0.278
0.295
0.250
0.265
0.269

4
1
3
1
0
8

Cv
0.585
0.630
0.678
0.559
0.596
0.606

2
6
6
2
9
6

T
0.288
0.289
0.306
0.327
0.337
0.347

5
4
7
0
9
8

Cv
0.657
0.665
0.710
0.768
0.796
0.836

8
7
4
7
2
2

T
0.364
0.373
0.393
0.399
0.423
0.427

4
4
5
4
1
4

Cv
0.885
0.918
0.973
1.003
1.021
1.102

8
0
3
0
0
0

T
0.451
0.457
0.483
0.496
0.543
0.594

5
8
5
9
5
4

Cv
1.177
1.208
1.302
1.353
1.551
1.786

0
0
0
0
0
0

July 31, 2012

11:38

42

World Scientific Book - 9in x 6in

Julia 8556-Quantum Theory of Solids

A Modern Course in Quantum Theory of Solids

1-18 Assume that the dispersion relation of an optical phonon branch in a


solid takes on the form (k) = 0 Ak 2 near k = 0, where 0 and A
are positive constants. Find the phonon densities of states for < 0
and > 0 , respectively.
1-19 The dispersion relation of the acoustical branch in a one-dimensional
crystal with a two-atom basis of identical atoms of mass m is given
by

1/2 o1/2
1 n
a (k) = 1/2 K + G K 2 + G2 + 2KG cos(ka)
,
m
where K and G are the force constants of the springs connecting the
two nearest neighbors of an atom. Evaluate the phonon density of
states for this acoustical branch.
1-20 At finite temperatures, phonons in a crystal are constantly created
P
and annihilated so that their number, nph = ks nks with nks the
number of phonons in single-phonon state |ksi, fluctuates greatly.
(i) Evaluate the thermal average number of phonons per unit volP
P P
ume hnph i = Z 1 n nph eEn /kB T = Z 1 ks n nks eEn /kB T with
P En /kB T
Z =
and En the eigenvalues of the harmonic latne
tice Hamiltonian in a three-dimensional crystal with a multi-atom
basis. (ii) Find the expressions for hnph i in the high- and lowtemperature limits. (iii) Evaluate the variance of the number of
2
phonons, var(nph ) = hn2ph i hnph i .
1-21 Reexamine the Gr
uneisen parameter (T ) of a one-dimensional crystal of inert gas atoms with the phonon dispersion relation given by
k = (4K/m)1/2 | sin(ka/2)|. (i) Show algebraically that (0) = 1
at T = 0. (ii) Show analytically that (T ) = ln 2 as T .
(iii) Evaluate numerically (T ) for T / from 0 to 1.5 and plot the
results. Here = ~(4K/m)1/2 /kB .

han

Вам также может понравиться