Sunteți pe pagina 1din 32

Chapter 5



so far: we considered mainly a static crystal with the atom s at fixed position
now: consider elastic vibrations of a crystal,
with atom s at position ~xs (t) = ~xs0 + ~us (t),
i.e. we consider (small) displacements ~us (t) that depend on time
and which are correlated to displacements of other atoms nearby
e.g. deformations or elongations of whole lattice planes.

Deformation of solids can be treated by the classical
theory of linear elasticity2 .
gives both the energies and dynamics associated with small deformations,
provided the wavelength of the deformations is large
in comparison with the lattice constants.
However, this (continuum) theory fails when the wavelength of deformations
becomes comparable to3 interatomic distances in the crystal lattice.
collective excitations/vibrations of the atoms (at frequency )
~ propagating through the crystal
formation of waves (with wave vector K)
important questions:
what are the vibrating modes in a crystal lattice ?
~ ?
what is the relation (K)
how do the lattice vibrations interact with mechanical, electromagnetic
or other forces ?

in contrast to uncorrelated vibrations, e.g. due to thermal fluctuations, which was considered for
the derivation of the Debye-Waller factor that enters into the expression for the intensity of the Bragg
peak for x-ray diffraction
see e.g. Kittel Introduction to solid state physics (1966) at the end of chapter 3
or smaller than



Chapter 5 Phonons
lattice vibrations have particle-like character
quantum mechanical treatment (particle-wave-dualism):
propagating waves phonons with energy ~
(analogy to photons)

importance of phonons:

determine many thermodynamic properties of a crystal

(e.g. specific heat)
contribution to thermal conductivity
interaction with electrons
large contribution to resistivity in metals
formation of (Cooper pairs) in (conventional) superconductors
absorption/emission of light in semiconductors
How many modes of oscillations can exist in a crystal with N atoms?
1 atom: 3 degrees of freedom (translation in x, y, z direction4 )
2 uncoupled atoms: 23 degrees of freedom
2 coupled atoms: also 6 degrees of freedom:
3 translation
2 rotation (perpendicular to the line connecting the
1 oscillation
N uncoupled atoms: 3N degrees of freedom (translation)
N coupled atoms: also have to have 3N degrees of freedom:
3 translation (crystal as a whole)
3 rotation (crystal as a whole)
3N 6 oscillation
One finds: the 3N 6 degrees of freedom for oscillations can be classified into relatively simple modes of oscillations charac~ (dispersion relation)
terized by their dependence (K)

neglecting rotational motion of the single atoms




Most simple situation is obtained in the [100], [110] and [111] propagation directions
of cubic crystals (correspond to directions along the cube edge, the face diagonal and
the body diagonal, respectively).
If a wave is propagating along one of these directions entire planes of atoms move in
phase with displacements either parallel (longitudinal) or perpendicular (transverse)
to the direction of the wavevector K
allows simple description of the displacement of a whole plane s from its equilibrium
1-dimensional problem
Each wavevector has three polarization modes one longitudinal (Fig.5.1) and two
transverse (Fig.5.2).

Figure 5.1: (Dashed lines) Planes of atoms when in equilibrium. (Solid lines) Planes
of atoms when displaced as for a longitudinal wave. The coordinate u measures the
displacement of the planes [from Kittel, Introduction to solid state physics (1996);

Figure 5.2: Planes of atoms as displaced during passage of a transverse wave [from
Kittel, Introduction to solid state physics (1999); Fig.4.3].


Chapter 5 Phonons


1-dimensional chain of identical atoms

Chain consists of N atoms with masses M .

allow only motion of atoms along x-direction
one degree of freedom corresponds to translation of the whole chain along x
N 1 degrees of freedom for oscillations
arrangement of atoms:

Figure 5.3: Displacement of atoms in 1dimensional chain

Atoms located at positions xs = sa + us (t) with lattice constant a
choose for simplicity periodic boundary conditions:

Atom 1 has always same position and momentum as atom N + 1

(in open chain: e.g. atom 1 and atom N is fixed in space)

Force acting on atom s due to atom s + 1:

Fs = C(us+1 us ) ,

with force constant C


Fs is linear in the displacements Hookes law is valid if deflection us is not

too large
For simplicity we consider only interactions with nearest neighbors s + 1 and s 1;
then the total force acting on atom s is
Fs = C(us+1 us ) + C(us1 us )


equations of motion of atoms (s = 1 . . . N ) with mass M :


d 2 us
M us = C(us+1 + us1 2us )



Ansatz for time dependence of displacements: us = us eit

us = 2 us in (5.3)

M 2 u0s = C(u0s+1 + u0s1 2u0s )


which is a difference equation in the displacements u,

with travelling wave solutions
u0s = ueiKsa


(= longitudinal wave along x with wavevector K)

with (5.5) in (5.4)
(divide by u eiKsa )

M 2 ueiksa = Cu(eiK(s+1)a + eiK(s1)a 2eiKsa )



1-dimensional chain of identical atoms


M 2 = C(eiKa + eiKa 2) = 2C(1 cos Ka)


(with 2 cos x = ei x + eix )

This finally yields the dispersion relation (K)

(1 cos Ka);


4C Ka


1 cos x = | sin(x/2)|)

for the travelling wave us = uei(sKat) .

Figure 5.4: Dispersion relation

(K) of monatomic linear chain.

The regime of physically significant values of K is limited due to eiKa ei(Ka+2n) ,

(with n = 0, 1, 2, . . .)

independent K vectors are only specified within the interval


1. Brillouin zone
connection between K values within and outside the 1. Brillouin zone:

A vector K outside the 1. Brillouin zone can be transformed into a vector K 0

inside the 1. Brillouin zone by subtracting or adding a multiple of 2/a,
i.e. via the transformation K 0 = K + G where G is a reciprocal lattice vector.
The displacement ration of neighboring atoms is
u e[i(s+1)Ka]
= eiKa
u eisKa

ei(Ka2n) eiK a
= eiKa = e|{z}

i.e. any displacement can always be described by a wavevector within the

1. Brillouin zone.




Chapter 5 Phonons
In other words:
The range of Ka covers all independent values of the exponential
eiKa = us+1 /us . I.e., saying that two adjacent atoms are out of phase by 0.3 or
2.3 has the same physical meaning, as shown in the figure below.

Figure 5.5: The wave represented by the solid curve conveys no information not given
by the dashed curve. Only wavelengths longer than 2a are needed to represent the
motion [from Kittel, Introduction to solid state physics (1996); Fig.4.5].
limit of long wavelength:
if K a wavelength 2/K a
one can approximate in (5.8) |sin(x)| x, i.e.

Ka K


this limit corresponds to the continuum approximation

there exists a maximum value for the frequency max = (4C/M )1/2 ,
reached when the K vector lies on the boundary of the 1. Brillouin zone
Kmax = /a min = 2a.
Typical values in solids are max 1013 Hz.

Solution at Kmax = a (i.e. at the zone boundary): Kmax sa = s, hence

us = u eis eit = u (1)s eit
= standing wave (due to Bragg reflection at the zone boundary)

compare with Bragg condition 2d sin = n :

d = a,
= /2 sin = 1

= 2/Kmax = 2/(/a) = 2a

2d sin = 2a
n = 2a



1-dimensional chain of identical atoms


group velocity:

The group velocity of a wave packet is given as the slope of the dispersion curve
~vg = gradK (K)
vg =
(corresponds to the velocity of energy propagation in a crystal)

With the dispersion relation (5.8) the group velocity is

vg =


Figure 5.6: Group velocity vs. K, for

monatomic 1-dim. chain. vg = 0 at the
zone boundary, i.e. there is no transport
of energy

limit q
of long wavelength:

vg =


a = const.

corresponds to sound waves with long wavelength;

Distribution of the N 1 oscillatory degrees of freedom

from periodic boundary conditions us = us+N

eiKsa = ei[K(s+N )a+2n]

N Ka = 2n
n, n = N/2 n N/2)
N discrete values for K; with K = 0 corresponding to translation
oscillatory degrees of freedom are being used up
just within the 1. Brillouin zone.


Chapter 5 Phonons
generalization to 3D:

we used Hookes law F = C u which is valid in that form only in one dimension;
in 3D this relation becomes more complex:

replace force by stress5 field (Spannungsfeld) (tensor; 33 matrix)

replace displacement by strain field (Dehnungsfeld) (tensor; 33 matrix)

then: =C e ;

C is a tensor (4. Stufe) with 81 componentsCiklm ;

hereby: 21 components are independent of each other
(s. Kittel, chapter 3)
derivation of force constants from experiment

So far, we used the approximation of only nearest neighbor interactions, i.e. for
the force on atom s from the displacement of atom s + p we only considered
p = 1.
Particularly in metals the effective forces can be of quite long range, i.e. interactions of lattice planes up to 20 lattice spacings apart from each other (i.e. p = 20)
can be significant.

For that case one can find a generalization of the dispersion relation (5.7) to p
nearest neighbors as
2 X
2 =
Cp (1 cos pKa)
M p>0
Solving for the interplanar force constants Cp at range pa yields
Cp =


dK K
cos pKa


for a structure with a monatomic basis.

The measurement of the dispersion relation hence allows the determination of
the range of the interplanar forces.

stress is defined as force per unit area


1-dimensional chain of two different atoms



1-dimensional chain of two different atoms

Atoms shall have masses M1 , M2

Atoms with masses M1 are displaced by us1 , us , us+1 ,. . .
and atoms with masses M2 by vs1 , vs , vs+1 ,. . .

Figure 5.7: 1-dimensional chain of two

different atoms with displacements vs , . . .
and us , . . .
The force constant shall be again denoted as C, i.e. force between two (different)
neighboring atoms is F = C(vs us ) (Hooks law).
equations of motion:
M1 us = C (vs + vs1 2us )
M2 vs = C (us+1 + us 2vs )


travelling waves, now with different amplitudes
us = u ei(sKat)
vs = v ei(sKat)


where the lattice constant a is defined as the distance between nearest neighbors of
atoms with same mass (for zero displacement).
Substitution of (5.17) in (5.16) yields
2 M1 u = Cv[1 + eiKa ] 2Cu
2 M2 v = Cu[1 + eiKa ] 2Cv


This system of linear equations has a solution only if the determinant of the coefficients
of the unknowns u, v vanishes

2C M1 2

C [1 + eiKa ]
2C M2 2

M1 M2 4 2C(M1 + M2 ) 2 + 2C 2 (1 cos Ka) = 0



Chapter 5 Phonons

solution for ka 1 :
with cos Ka 1 12 K 2 a2 + . . . one finds the two solutions

optical branch
M1 M2

K 2 a2
2(M1 + M2 )

acoustical branch



solution at the zone boundary Kmax = a :

cos Ka = 1, hence
2 =



2 =



The full dispersion relatione, including the discussed limiting cases is shown below for
the case M1 > M2 .

Figure 5.8: Optical and acoustical branches of the dispersion relation for a diatomic
linear lattice, showing the limiting frequencies at K = 0 and K = Kmax = a . The
lattice constant is a [from Kittel, Introduction to solid state physics (1996); Fig.4.7].


1-dimensional chain of two different atoms


The figure below shows the atom displacements transverse to the K vector.

Figure 5.9: Transverse optical

and transverse acoustical waves
in a diatomic linear lattice, illustrated by the particle displacements for the two modes at the
same wavelength. [from Kittel, Introduction to solid state
physics (1996); Fig.4.10].

For the transverse optical (TO) branch at K = 0 one finds on substitution of

(5.21) in (5.18)
The minus sign means that the atoms vibrate against each other.
If the two atoms have opposite charges one can excite this type of vibration by
the electric field of an incident light wave optical mode.
In the K = 0 limit of (5.21) one finds as a solution for the transverse acoustical
branch v = u.
This means that the atoms move together, as in long wavelength acoustical vibrations acoustical mode.
frequency gap:

For certain frequencies here for (2C/M1 )1/2 < < (2C/M2 )1/2 there are no
wavelike solutions.
characteristic for polyatomic lattices

Solutions for real yield a complex wavevector

us = u es|K|a eit damping in space!

Note: the 2N degrees of freedom fit again into the 1. Brillouin zone.


Chapter 5 Phonons


Generalization and examples

extension to p atoms in the basis: p branches in 1D

extension to 3D:
one longitudinal (L) and two transverse (T) polarization modes
E.g. for NaCl or diamond with p = 2 atoms in the primitive cell:
for each of the three polarization modes (1L and 2T) one gets two branches
(acoustical and optical)
2 3 = 6 branches
The figures below show examples for germanium and KBr also with p = 2
atoms per primitive cell.

3 acoustical branches (1LA, 2TA)

3p 3 optical branches (p 1)LO, 2(p 1)TO
(valid for each direction K)



Figure 5.10: Phonon dispersion relations in the [111] direction: Left in germanium
at 80 K. The twoTA phonon branches are horizontal at the zone boundary position
12 21 12 . The LO and TO branches coincide at K = 0 ; this also is a conseKmax = 2
quence of the crystal symmetry of Ge. The results were obtained with neutron inelastic
scattering by G. Nielsson and G. Nelin. Right in KBr at 90 K, after A.D.B. Woods
et al. The extrapolation to K = 0 of the TO and LO branches are called T and L .
[from Kittel, Introduction to solid state physics (1996); Fig.4.8].


Generalization and examples


Example: Silicon
Si crystallizes in the diamond lattice with p = 2 atoms per cell. The figure below shows
the Brillouin zone of such a cubic crystal.

Figure 5.11: Brillouin zone of cubic crystals. Some points and lines of high symmetry are drawn. denotes the center
(0, 0, 0), the point X is given by 2
(0, 1, 0),
the point L by a (1, 1, 1) and the point
K by 3
(0, 1, 1) [from Bergmann-Schaefer,
Lehrbuch der Experimentalphysik, Bd. 6
Festkorperphysik (1992); Abb.1.18].
Due to p = 2, silicon has 6 phonon branches which are degenerate along the lines
and .

Figure 5.12: Dispersion of phonons in Si. Shown is the phonon energy ~, with phonon
frequency . , X, K and L denote different points and , and different lines in
outside the Brillouin
the Brillouin zone, as defined in fig. 5.4. Extending the line K
zone one reaches a point which is equivalent to X. L and T means longitudinally
and transversely polarized wave, respectively, A, acoustical wave, and O optical wave.
[from Bergmann-Schaefer, Lehrbuch der Experimentalphysik, Bd. 6 Festkorperphysik
(1992); Abb.1.19].


Chapter 5 Phonons

Example: Tl2 Ba2 CaCu2 O8 (high-Tc superconductor)

The figures below show the Brillouin zones of a body-centered tetragonal (bct) and a
simple tetragonal (st) lattice of a complex cuprate superconductor, together with the
calculated phonon dispersion relations of the bct structure.

Figure 5.13: Left: Brillouin zones of the body-centered tetragonal (top) and simple
tetragonal (bottom) lattices. The symmetry directions along which the dispersion
curves are displayed are indicated with
lines. For the body-centered tetragonal

lattice, X = (1, 1, 0) a , G1 = (1, 0, 0) 1 + c2 a , and Z = (0, 0, 2) c = (2, 0, 0) a . For

the simple tetragonale lattice, X = (1, 0, 0) a , M = (1, 1, 0) a and Z = (0, 0, 1) c . a, a, c
are the sides of the conventional unit cells of both lattices. Right: Phonon dispersion
curves of bct Tl2 CaBa2 Cu2 O3 (2:1:2:2:8) along symmetry directions of the bct Brillouin
zone. [from Kulkarni et al., Physical Review B 43, 5451 (1991); Fig. 1 and 3].


Quantized lattice vibrations: Phonons



Quantized lattice vibrations: Phonons

Quantum mechanics: Energy of a wave with frequency :

E = n+
with n: number of quantum particles; n = 0, 1, 2, . . .
electromagnetic (light) wave: Quantum particle = Photon
lattice vibrations: Quantum particle = Phonon
E = (n+ 12 )~ is the energy of an elastic mode of frequency when the mode is excited
to the quantum number n,
i.e. the mode is occupied by n phonons.
For n = 0 E = ~/2 = zero point energy of the mode.
Occurs for both phonon and photon due to their equivalence to a quantum mechanical harmonic oscillator of frequency
Phonon momentum:
~ can interact with particles as if it had a momentum p~ = ~K
A phonon of wavevector K
However, a phonon does not carry a physical momentum
~ = 0, corresponding to the translation of a crystal as a whole)
(except for K
~ is only defined modulo a reciprocal lattice vector G.
Note: K
Explanation (in the particle picture):
~ during Bragg
the crystal as a whole can change its momentum by ~G
reflection (recoil);
~ 2 /2Mcrystal very small, since Mcrystal very large
Recoil energy Erecoil = (~K)
possible interactions of phonons with:
electrons, neutron, photons,. . .
other phonons (due to anharmonic terms in the elastic energy)
selection rules in a crystal:
govern allowed transitions between quantum states,
e.g. for elastic scattering of an x-ray photon by a crystal we found
~k 0 = ~k + G


~ is a reciprocal lattice vector, ~k is the wavevector of the incident photon

where G
and ~k 0 is the wavevector of the scattered photon.


Chapter 5 Phonons
inelastic scattering by phonons:
scattering of a photon with a phonon involved can be described by either
~k 0 + K
~ = ~k + G


~ or by
i.e. the creation (emission) of a phonon of wavevector K
~k 0 = ~k + K
~ +G


~ during the scattering process.

i.e. the absorption of a phonon K
interaction with neutrons:

Inelastic scattering of neutrons by absorption or emission of a phonon is a widely

used method for the determination of the phonon dispersion relation (K)

Kinematics are described by

conservation of momentum:
~kneutron + G
~ = ~k 0
neutron Kph


(: creation/anihilation of a phonon)
~ is chosen such that K
~ lies in the 1. Brillouin zone.
Here, G
conservation of energy:
~2 kneutron
~2 kneutron


(: creation/anihilation of a phonon)
~ is based on the two conservation laws (5.28),
Experimental determination of (K)
Then in the experiment, one needs to find the energy gain or energy loss of the
scattered neutrons as a function of the scattering direction ~k ~k 0 .

Advantage of using neutrons:

Eneutron , neutron are comparable to Eph , ph


Quantized lattice vibrations: Phonons


Figure 5.14: Phonon dispersion in GaAs as obtained from neutron scattering (after
Strauch and Dorner). The different points and lines on/along the wave vector axis
are defined in Fig.5.11; is the frequency [from Bergmann-Schaefer, Lehrbuch der
Experimentalphysik, Bd. 6 Festkorperphysik (1992); Abb.1.30].


Chapter 5 Phonons


Thermal properties of phonons

contribution to the heat capacity

Heat capacity is very sensitive to elementary excitations and ordering phenomena in

crystals (electrons, phonons, magnetic order, spin waves, superconductivity etc.).
Each phenomenon has very specific properties (e.g. T dependence)
heat capacity is a well suited probe;
particular advantage: detects volume properties
In general one defines the heat capacity either as
CV (U/T )V at constant volume or
Cp (U/T )p at constant pressure;
here T is the temperature and U is the energy.
The phonon contribution to the heat capacity of a crystal is called the
lattice heat capacity Clat .
some results from quantum statistics (without proof):
A system of quantum particles (e.g. vibrations) shall be characterized by energies E n
(n: index number)
question: What is the probability f (En ) of finding a quantum particle
at the energy En ?
answer: depends on the nature of the particle and on temperature:
particles with semi-integer spin (e.g.
e.g. electrons:
f (En ) =

e(En )/kB T



= fermions,

Fermi-Dirac distribution function

Here, is the(chemical potential), which has to be determined in such
a way that the number of particles in the system (e.g. all electrons in the
crystal) is conserved.
particles with integer spin (e.g. 0~, 1~) = bosons,
e.g. photons, phonons:
f (En ) =

e(En )/kB T

Bose-Einstein distribution function


particular property of photons, phonons:

there is no conservation of the number of particles = 0
classical statistics (for distinguishable particles) yields probability
f e(E)/kB T

Boltzmann distribution function



Thermal properties of phonons


Back to the phonon heat capacity:

internal energy of a vibrating crystal:






~ multiplied by the probability to have a state occupied at this

= energy ~(K)
~ and polarization states p.
energy and with summation over all wave vectors K
there do exist a lot of states K
K~ can be transformed into an integral d D(),

with D(): number of states within the frequency interval d

(= density of states or Zustandsdichte)

determination of D():
consider cube with length L, plus periodic boundary conditions
possible vibrating states
(compare to linear chain with L = N a where we found K = 2n/N a):
Kx = 0, 2/L, 4/L, . . . ; Ky , Kz analog

1 state per


8 3


~ space within a sphere with radius K:

number of states in K
NK =

4 3 V
K 3


(for each polarization p)

The density of states for each polarization type is hence given by
D() =
next task: determine dK/d

dNK dK
V K 2 dK

dK d
2 2 d



Chapter 5 Phonons

Debye model of density of states:

The velocity of sound is taken as constant for each polarization type
(like in classical elastic continuum).
Gives simple dispersion relation
= vs K


with sound velocity vs .

Hence, the density of states is given as
D() =

V 2
2 2 vs3


Remark :
Debye model is o.k. for acoustical phonons;
at sufficiently low temperature 6 mainly those phonons are excited
(energy of the optical phonons is large for all K!)
With N primitive cells in the crystal volume V :
The total number of acoustic phonon modes is N
From (5.35) we know then that all N vibrating states are located within the volume
(in K space)
V 4
V 4
N = 3 K 3 = 3 (/vs )3
8 3
8 3
This relation defines a maximum frequency, or cutoff frequency
D =

6 2 vs3 N


(Debye frequency)


With the dispersion relation one gets the corresponding cutoff wavevector
KD =

6 2 N


(Debye wave vector)


Note: D and KD depend only on the density N/V , which is physically reasonable.

compare distribution function


Thermal properties of phonons


Thermal energy for each polarization type:

Now we can use the obtained density of state to make the transformation from the
summation over all K vectors in (5.33) to the integration over all possible frequencies
up to the maximum frequency D
V 2
U = dD() ~/k T
d 2 3 ~/k T
2 vs e
If vs is identical for all three polarizations
4 4 Z xD
3V kB
d ~/k T
U =3 2 3
2 vs 0
2 2 vs3 ~3 0
ex 1
with x

kB T




kB T


Here we defined = ~D /kB as the Debye temperature


6 2 N



With this definition we can express the thermal energy as

3 Z xD
U = 9N kB T
dx x

e 1


with N atoms in the specimen. Thus, the heat capacity (plotted in Fig.5.15)) is
= 9N kB
Cv =

3 Z xD
x4 e x
dx x

(e 1)2


Figure 5.15: Heat capacity CV of a solid, according to the Debye approximation. [from
Kittel, Introduction to solid state physics (1996); Fig.5.7].


Chapter 5 Phonons

For low temperature: xD

(Debye approximation is anyway only a good approximation at low temperature)


3 4 N kB T 4


dx x
e 1


ex 1

12 4 N kB T 3
and cv =
234N kB

cV T 3


(Debyes T 3 law)



simple explanation:
We have 3N possible modes (N : number of primitive cells).
Only those modes having ~ < kB T will be excited at low temperature to an energy
close to kB T .
~ space only a fraction (T /D )3 = (T /)3 will
Hence, from the available volume in K
be occupied by excited modes.
T 3
number of excited phonons 3N
each phonon has energy kB T
energy 3N kB T (T /)3 , and CV 12N kB (T /)3
The range of validity of the T 3 is below 0.1.
For T the heat capacity approaches the classical value 3N kB .
Typical values for :

: 334 K
: 2230 K
: 165 K
: 91 K

Figure 5.16: Heat capacity of silicon and germanium. Not the decrease at low temperatures. To convert a value in cal/molK to J/MolK, multiply by 4.186. [from Kittel,
Introduction to solid state physics (1996); Fig.5.8].


Thermal properties of phonons


Einstein model of the density of states:

consider N oscillators at the same frequency 0
(o.k. for many optical phonons)
Density of states:
D() = N ( 0 )


Thermal energy:
U = 3N


heat capacity:
Cv = 3N kB
as plotted in the figure below

kB T

3 due to polarization

e~0 /kB T
(e~0 /kB T 1)2



Figure 5.17: Comparison of experimental values of the heat capacity of diamond with
values calculated on the earliest quantum (Einstein) model, using the characteristic
temperature E = ~/kB = 1320 K. To convert to J/MolK, multiply by 4.186. [from
Kittel, Introduction to solid state physics (1996); Fig.5.11].
For low temperature: Cv e~0 /kB T
(not as good as Debye model which is in better agreement with experimental data)
For high temperature: Cv 3N kB
(reproducing the Dulong-Petit value, as the Debye model does)


Chapter 5 Phonons

General expression for the density of states:

A general expression for D() the number of states per unit frequency range requires
the knowledge of the phonon dispersion relation (K).
The number of allowed K values for which the phonon frequency is between and
+ d is
D()d = 3
d3 K
8 shell
(We found earlier that we have one state per V /(8 3 );
~ space bounded by the two surfaces
The integral goes over the volume of the shell in K
on which the phonon frequency is constant and either or + d)
evaluate the volume of this shell !
~ space as shown in the figure below.
We consider a surface of constant in K
Here, we denote dS an element area on this surface

Figure 5.18: Element of area dS on a

constant frequency surface in K space.
The volume between two surfaces of constant frequency
at and + d is
equal to dS d/|K | [from Kittel, Introduction to solid state physics (1996);

With the perpendicular distance dK between the surfaces of constant and constant
+ d (see next figure) one gets:
d k = dS dk


Thermal properties of phonons


Figure 5.19: The quantity dK is the perpendicular

distance between two constant frequency surfaces in
K space, one at frequency , the other at frequency
+ d [from Kittel, Introduction to solid state physics
(1996); Fig.5.13].
~ ~ |dK = d; here
~ ~ is the gradient in K space
furthermore |
= dS
~ ~ |

dS dK = dS
D() =

8 3


~ space
Integral goes over area of constant in K
D() can contain singularity points = Van-Hove singularities,
if the group velocity vg = 0
(this is e.g. the case at Brillouin zone boundary)

Figure 5.20: Density of states as a function of frequency for (a) a solid according to
the Debye model and (b) an actual crystal structure. The spectrum for the crystal
starts as 2 for small , but discontinuities develop at singular points [from Kittel,
Introduction to solid state physics (1996); Fig.5.14].


Chapter 5 Phonons


Thermal expansion of solids

So far, we only considered a linear force-displacement relation to describe lattice vibrations (Hooks law).
This leads to a potential energy which is quadratic in the interatomic displacements
(harmonic theory).
As one of its consequences there should be no thermal expansion which we will consider
Naive view: solid expands, while atoms oscillate
about their equilibrium position.
but: if one considers only harmonic oscillations, the average position of
the atom remains unchanged, no matter how large the amplitude
of oscillation is.
thermal expansion of solids results from anharmonic potential
Taylor expansion of the potential energy of an atom in the crystal potential
(equilibrium position shall be at x = 0)
(consider 1-dim. solid)
U (x) = cx2 gx3 f x4 . . .


only 1. term: harmonic approximation

additional terms: anharmonic
term gx3 : Asymmetry in mutual repulsion of neighboring atoms
term f x4 : Softening of the vibrations (attenuation of the spring constant at large amplitudes
average displacement of the atom:
use Boltzmann distribution which weights all possible values of displacement
x with the factor eU (x)/kB T (classical statistics o.k. for high temperatures)

hxi =

dx x eU (x)

dx e


U (x)

with = kB1T . The denominator is a normalization factor; normalizes the total probability to 1.


Thermal properties of phonons


expand the integrands for small displacements

i.e. such that the anharmonic terms in the energy are small as compared to kB T

3 g 3/2
U (x)
dx xe

dx e
(x + gx + f x ) =

4 c5/2

U (x)
dx e

kB T
thermal expansion proportional to T due to 3. order term in the potential


Measurements of the lattice constant a vs. temperature of solid argon are shown below.
The slope da/dT is proportional to the thermal expansion coefficient (vanishes at low
T ), which is significant at high T .

Figure 5.21: Lattice constant of solid argon as a function of temperature [from Kittel,
Introduction to solid state physics (1996); Fig.5.15].



Chapter 5 Phonons

Phonon contribution to thermal conductivity

consider long slab with temperature gradient dT /dx

there is a steady-state flow of heat
jU = KU



here KU is the thermal conductivity coefficient (Warmeleitzahl), and jU is the flux of

thermal energy (or the energy transmitted across unit area per time)
transfer of thermal energy is a diffusive process, subject to random scattering processes
inside the slab;
otherwise jU would only be determined by the temperature difference T at the ends
of the slab.
the random nature of the heat transfer is responsible for
(i) the development of a thermal gradient, and
(ii) for the appearance of a mean free path ` in the expression of the thermal flux.
Idea for calculation: consider phonons as a gas of particles
general result of the kinetic gas theory: KU = 13 Cv` with:
C: heat capacity per volume
v: mean particle velocity (here: velocity of sound, few km/s)
`: mean free path (typically a few nm at 300 K; a few 10 nm at 77 K)
consider flux of particles nhvx i along x direction (n: density of particles)
The associated thermal flux is then given by
ju = nhvx iU
this is due to the fact that a particle will give up energy U = cT at temperature T :
(if particle comes from a region with higher temperature T + T ;
c: heat capacity of the particle)
between end points of a mean free path `:
T =

hvx i
ju = nhvx2 ic

with : mean collision time

= nhv 2 ic



with ` = v and C = nc
jU = Cvl
|3 {z } dx




Thermal properties of phonons


Discussion of heat transport mediated by phonon gas

Above, we introduced the mean free path for phonon scattering to derive the heat
transport mediated by phonons.
required: inelastic scattering processes, which restore thermal equilibrium.
collisions of phonons with lattice defects, or with surface:
does not work, because elastic the phonon frequency remains unchanged
collisions between phonons:
(note: in this case anharmonic terms in the potential energy U (x) of the
atoms are important!)
~1 + K
~2 = K
~3 + G
in general: K
~ = 0: Normal processes (N processes)
for G
~ 6= 0: Umklapp-Processes (U processes)
for G
N processes:
The total phonon momentum is J~ = J~ nK~ ~K;
here nK~ is the number of phonons with wavevector K.
The change in total phonon momentum is
~ 3 ~K
~ 1 ~K
~ 2 = ~(K
~3 K
~1 K
~ 2) = 0
J~ = J~0 J~ = ~K


i.e. for N processes the total phonon momentum is conserved

A distribution of hot phonons with J~ 6= 0 propagating down a rod will leave J~
unchanged, i.e. N processes will not be able to establish thermal equilibrium.
no thermal resistance, as illustrated in Fig.5.23(top) for the flow of hot gas
molecules in a tube with frictionless walls and in Fig.5.24(top).
U processes:
U processes are responsible for the thermal conductivity due to phonons
as they allow change of the total phonon momentum (Fig.5.23(bottom),

Figure 5.22: (a) Normal (N )

and (b) Umklapp (U ) phonon
collision processes in the reciprocal lattice of a 2-dim. square
lattice. The square represents
the 1. Brillouin zone [from Kittel, Introduction to solid state
physics (1999); Fig.5.17c].


Chapter 5 Phonons

Figure 5.23: (top) Flow of gas molecules in a state of drifting equilibrium down a long
open tube with frictionless walls. Elastic collision processes among the gas molecules do
not change the momentum or energy flux of the gas because in each collision the velocity
of the center of mass of the colliding particles and their energy remain unchanged.
Thus energy is transported from left to right without being driven by a temperature
gradient. Therefore the thermal resistivity is zero and the thermal conductivity is
infinite. (bottom) The usual definition of thermal conductivity in a gas refers to
a situation where no mass flow is permitted. Here the tube is closed at both ends,
preventing the escape or entrance of molecules. With a temperature gradient the
colliding pairs with above-average center of mass velocities will tend to be directed to
the right, those with below-average velocities will tend to be directed to the left. A
slight concentration gradient, high on the right, will be set up to enable the net mass
transport to be zero while allowing a net energy transport from the hot to the cold
end. [from Kittel, Introduction to solid state physics (1999); Fig.5.16a].


Thermal properties of phonons


Figure 5.24: top: In a crystal we may arrange to create phonons chiefly at one end,
as by illuminating the left end with a lamp. From that end there will be a net flux
of phonons toward the right end of the crystal. If only N processes (K1 + K2 = K3 )
occur, the phonon flux is unchanged in momentum on collision and some phonon flux
will persist down the length of the crystal. On arrival of phonons at the right end we
can arrange in principle to convert most of their energy to radiation, thereby creating a
sink for the phonons. Just as in Fig.5.23(top) the thermal resistivity is zero. bottom:
In U processes there is a large net change in phonon momentum in each collision event.
An initial net phonon flux will rapidly decay as we move to the right. The ends may act
as sources and sinks. Net energy transport under a temperature gradient occurs as in
Fig.5.23(bottom) [after Kittel, Introduction to solid state physics (1999); Fig.5.16cd].
very high temperatures T > :
all phonon modes are excited and the phonon density n T (compare with
Bose-Einstein statistics)
U processes will provide relaxation to thermal equilibrium
mean free path ` 1/n 1/T KU 1/T
low temperatures:
number of phonons which are suited to U processes decreases e/T

K e/T

very low temperatures:

mean free paths reaches dimension of the crystal D
KU 13 CvD T 3 , since C T 3


Chapter 5 Phonons

example (NaF):

Figure 5.25: Thermal conductivity of a highly purified crystal of sodium fluoride (NaF),
after H.E. Jackson, C.T. Walker and T.F. McNelly [from Kittel, Introduction to solid
state physics (1999); Fig.5.18].
Final remarks:
in metals:
additional (strong) heat transport due to electrons
some insulators (e.g. sapphire, Al2 O3 ):
have nevertheless a thermal conductivity which is comparable to metals,
e.g. maximum 200 W/(cmK) at 30 K for sapphire
for comparison:
maximum thermal conductivity of copper is 100 W/(cmK)