PHONONS PHOTONS

• Quanta of lattice vibrations • Quanta of electromagnetic

• Energies of phonons are radiation
quantized • Energies of photons are
quantized as well

h s hc
E phonon  E photon 
 
~a0=10-10m
~10-6m

h h
p phonon  p photon 
 
Energy of harmonic oscillator
Obtained by in a classical way of considering the normal modes
that we have found are independent and harmonic.

 1
 n   n   
 2 • Make a transition to Q.M.

Energy, E
• Represents equally spaced
energy levels

 Energy levels of atoms
vibrating at a single
 frequency ω


It is possible to consider  n as constructed by adding n excitation
quanta each of energy  to the ground state.
1
 0  
2

A transition from a lower energy level to a higher energy level.

 1  1
   n2     n1  
 2  2

   n2  n1      
  
unity

absorption of phonon
The converse transition results an emission of phonon
with an energy  .
Phonons are quanta of lattice vibrations with an
angular frequency of  .
Phonons are not localized particles.
Its momentum is exact, but position can not be
determined because of the uncertainity princible.
However, a slightly localized wavepacket can be
considered by combining modes of slightly different 
and .
 
Assume waves with a spread of k of ; so this wavepacket will
be localized within 10 unit cells.
10 a

This wavepacket will represent a fairly localized phonon moving with

group velocity
d .
dk

Phonons can be treated as localized particles within some limits.

 Energy of
phonons
1D crystals
 

k Multiply by 
k

Crystal momentum

•Phonons are not conserved

•They can be created and destroyed during collisions .
 Thermal energy and lattice vibrations
•Atoms vibrate about their equilibrium position.
•They produce vibrational waves.
•This motion is increased as the temperature is
raised.

In a solid, the energy associated with this vibration and perhaps also
with the rotation of atoms and molecules is called as thermal energy.

Note: In a gas, the translational motion of atoms and molecules

contribute to this energy.
Therefore, the concept of thermal energy is fundamental to an
understanding many of the basic properties of solids. We would like
to know:
•What is the value of this thermal energy?
•How much is available to scatter a conduction electron in a metal;
since this scattering gives rise to electrical resistance.
•The energy can be used to activate a crystallographic or a magnetic
transition.
•How the vibrational energy changes with temperature since this gives
a measure of the heat energy which is necessary to raise the
temperature of the material.
•Recall that the specific heat or heat capacity is the thermal energy
which is required to raise the temperature of unit mass or 1gmole by
one Kelvin.
Heat capacity from Lattice vibrations
The energy given to lattice vibrations is the dominant
contribution to the heat capacity in most solids. In non-magnetic
insulators, it is the only contribution.
Other contributions;
•In metals from the conduction electrons.
•n magnetic materials from magneting ordering.
Atomic vibrations leads to band of normal mode frequencies from zero
up to some maximum value. Calculation of the lattice energy and heat
capacity of a solid therefore falls into two parts:
i) the evaluation of the contribution of a single mode, and
ii) the summation over the frequency distribution of the modes.
 Energy and heat capacity of a harmonic
oscillator, Einstein Model
_
   Pn n
n
Avarage energy of a harmonic
oscillator and hence of a lattice
mode of angular frequency at
temperature T Energy of oscillator

 1
The probability of the oscillator being in this  n   n   
level as given by the Boltzman factor  2
exp( n / k BT )
 
 1   1 
_   n     exp 
  n     / k T
B 
   Pn n  2    2  (*)
_
  n0 
n 
  1 

n0
exp 
 
 
n 
2


  / k T
B 



1 
z   exp[(n  ) ]
n0 2 k BT
z  e   / 2 kBT  e 3 / 2 kBT  e 5  / 2 kBT  .....
z  e   / 2 kBT (1  e   / kBT  e 2  / kBT  .....
z  e   / 2 kBT (1  e   / kBT ) 1

According to the Binomial expansion for x«1 where x   / k BT

Eqn (*) can be written
_
1 z 2 
  k BT 2
 k BT (ln z )
z T T
_
2   e   / 2 k B T 
  k BT ln  
T  1  e   / kBT 
_
    / 2 k B T
  k BT 2
T  ln e  ln  1  e   / k B T
 

_       
  k BT   
2
 ln  1  e   / k B T
  x'
T 2 k T  T (ln x) 
  B   x x
   k B   / k B T 
_  2k  k 2T 2 e  1   e   / k B T
  k BT 2  B2 2  B   / k T    
 4 k BT  1 e B   2  1  e   / k B T

 

_
1 
      / k B T
2 e 1
 _
1 
      / k B T
2 e 1
This is the mean energy of phonons.The first term in the above
equation is the zero-point energy. As we have mentioned before even
at 0ºK atoms vibrate in the crystal and have zero-point energy. This is
the minimum energy of the system.
The avarage number of phonons is given by Bose-Einstein
distribution as
_
(number of phonons) x (energy of phonon)=(second term in  )
1
n( )  

e k BT
1
The second term in the mean energy is the contribution of
phonons to the energy.
  k BT Mean energy of a
harmonic oscillator
as a function of T
1

2
T

low temperature limit

 k BT
1_
 Since exponential term
      gets bigger
2 e kBT  1
_
1
   Zero point energy
2
  k BT Mean energy of a
harmonic oscillator as
a function of T
2
1 x
 e x  1  x   ..........
2 2!
T 

e k BT
 1
high temperature limit  k BT k BT
_
1 
   
2 
• is independent of frequency of 1 1
oscillation. k BT
_
1
•This is the classical limit because the      k BT
energy steps are now small compared with 2
the energy of the harmonic oscillator.
•So that is the thermal energy of the _
classical 1D harmonic oscillator.   kBT
 Heat Capacity C
Heat capacity C can be found by differentiating the average energy of
phonons of
_
1 
    
2 e kBT  1

  k B 

  e kBT


  
2
 k BT 
2
d e k BT

Cv   Cv  k B
e   B 
 
2 2 2
dT 
k T 
kBT
1 e k BT
1

2 
   e T

Let  Cv  k B  
 e 1
2
k T  
T
 Plot of Cv as a function of T
2 

  e T

Cv  k B   where 
 e 1
2
T  
T k

Cv Specific heat vanishes

exponentially at low T’s and tends to
kB classical value at high temperatures.

Area= The features are common to all
2 quantum systems; the energy tends
T to the zero-point-energy at low T’s
 and to the classical value of
kB Boltzmann constant at high T’s.
 Plot of Cv as a function of T
Specific heat at constant volume depends on temperature as shown in
figure below. At high temperatures the value of Cv is close to 3R,
where R is the universal gas constant. Since R is approximately 2
cal/K-mole, at high temperatures Cv is app. 6 cal/K-mole.

This range usually includes RT.

Cv From the figure it is seen that Cv is
equal to 3R at high temperatures
3R regardless of the substance. This fact
is known as Dulong-Petit law. This law
states that specific heat of a given
number of atoms of any solid is
T, K independent of temperature and is the
same for all materials!
 Classical theory of
heat capacity of solids
The solid is one in which each atom is bound to its side by
a harmonic force. When the solid is heated, the atoms vibrate
around their sites like a set of harmonic oscillators. The
average energy for a 1D oscillator is kT. Therefore, the
averaga energy per atom, regarded as a 3D oscillator, is 3kT,
and consequently the energy per mole is
 = 3NkBT  3RT
where N is Avagadro’s number, kB is Boltzmann constant and
R is the gas constant. The differentiation wrt temperature
gives;
d
Cv  Cv  3R  3  6.02  1023 (atoms / mole)  1.38  1023 ( J / K )
dT
J Cal
Cv  24.9 ;1J  0.2388Cal  Cv 6
( K  mole) ( K  mole)
 Einstein heat capacity of solids
 The theory explained by Einstein is the first quantum theory of solids.
He made the simplifying assumption that all 3N vibrational modes of
a 3D solid of N atoms had the same frequency, so that the whole
solid had a heat capacity 3N times 
2
  e T
Cv  k B  
 
2
 T  e T  1

 In this model, the atoms are treated as independent oscillators, but

the energy of the oscillators are taken quantum mechanically as

This refers to an isolated oscillator, but the atomic oscillators in a solid
are not isolated.They are continually exchanging their energy with
their surrounding atoms.
 Even this crude model gave the correct limit at high temperatures, a
heat capacity of
3 Nk B  3R
Dulong-Petit law where R is universal gas constant.
 • At high temperatures, all crystalline solids have a specific heat of
6 cal/K per mole; they require 6 calories per mole to raise their
temperature 1 K.
•This arrangement between observation and classical theory break
down if the temperature is not high.
•Observations show that at room temperatures and below the
specific heat of crystalline solids is not a universal constant.
Cv
In all of these materials
cal (Pb,Al, Si,and Diamond)
6
Kmol specific heat approaches
constant value asymptotically
at high T’s. But at low T’s, the
specific heat decreases
T towards zero which is in a
 complete contradiction with
Cv  3 R the above classical result.
kB
 The Discrepancy of Einstein model
Einstein model also gave correctly a specific heat tending to
zero at absolute zero, but the temperature dependence near T=0
did not agree with experiment.

Taking into account the actual distribution of vibration

frequencies in a solid this discrepancy can be accounted using
one dimensional model of monoatomic lattice
Density of States
According to Quantum Mechanics if a particle is constrained;
the energy of particle can only have special discrete energy
values.
it cannot increase infinitely from one value to another.
it has to go up in steps.
 These steps can be so small depending on the system that the
energy can be considered as continuous.
This is the case of classical mechanics.
But on atomic scale the energy can only jump by a discrete
amount from one value to another.

Definite energy levels Steps get small Energy is continuous

In some cases, each particular energy level can be
associated with more than one different state (or
wavefunction )
This energy level is said to be degenerate.

The density of states  ( ) is the number of discrete states

per unit energy interval, and so that the number of states
between  and   d  will be  ( )d  .
 There are two sets of waves for solution;
Running waves
Standing waves
Running waves:
k
4 2 0 2 4 6
 
L L L L L
These allowed k wavenumbers corresponds to the running
waves; all positive and negative values of k are allowed. By
means of periodic boundary condition an integer
Na 2 2 2
L  Na  p     k  pk  p Length of
p k Na L the 1D
chain

These allowed wavenumbers are uniformly distibuted in k at a

density of  R  k  between k and k+dk.
L
running waves  R  k  dk  dk
2
 Standing waves: 5 4 3
L L L
6
k L 2
0  2 3 L
L L L 7 0 
L L

In some cases it is more suitable to use standing waves,i.e. chain

with fixed ends. Therefore we will have an integral number of half
wavelengths in the chain;
n 2 2 n n
L  ;k  k  k 
2  2L L

These are the allowed wavenumbers for standing waves; only

positive values are allowed.
2 for  for
k p k p
L running waves L standing waves
These allowed k’s are uniformly distributed between k and k+dk
at a density of  S (k )

L
 S (k )dk  dk DOS of standing wave

L
 R  k  dk  dk DOS of running wave
2

•The density of standing wave states is twice that of the running waves.
•However in the case of standing waves only positive values are
allowed
•Then the total number of states for both running and standing waves
will be the same in a range dk of the magnitude k
•The standing waves have the same dispersion relation as running
waves, and for a chain containing N atoms there are exactly N distinct
states with k values in the range 0 to  / a .
The density of states per unit frequency range g():

The number of modes with frequencies  and +d will be

g()d.
g() can be written in terms of S(k) and R(k).

dR modes with frequency from  to +d corresponds

dn modes with wavenumber from k to k+dk

 dn   R (k )dk  g ( )d ; dn   S (k )dk  g ( )d

Choose standing waves to obtain g ( )

g ( )   S (k ) dk
d
Let’s remember dispertion relation for 1D monoatomic lattice

4K 2 ka K ka
 
2
sin 2 sin
m 2 m 2

d 2a K K kaka 1 m1 1
a cos
cos g ( )   S (k )
dk 2 mm 2 2 a KK  ka 
ka
a cos
cos  
m 2 2 
 1 m 1
g ( )   S ( k )
a K cos  ka / 2 

sin 2 x  cos 2 x  1  cos x  1  sin 2 x  ka  2  ka 

cos    1  sin  
 2   2 
1 m 1 4 Multibly and divide
g ( )   S (k )
a K  ka  4
1  sin 2  
 2 
Let’s remember:
1 2 L
g ( )   S (k )  S (k )dk  dk
a 4K 4K 2  ka  
 sin  
m m  2  L  Na
4K 2  ka 
2Na
LN 2 2 1 2 1/ 2  
2
sin  
g ( )   max   
 a max
m  2 
2
2 4K
max 
2

True density of states m

 g ( )
2N 2
g ( ) 

 max   
2 1/ 2

N m
True density of states by
 K means of above equation

K K
max 2 
m m
constant density of states

K
True DOS(density of states) tends to infinity at max  2 ,
m
since the group velocity d / dk goes to zero at this value of  .
Constant density of states can be obtained by ignoring the
dispersion of sound at wavelengths comparable to atomic spacing.
 The energy of lattice vibrations will then be found by
integrating the energy of single oscillator over the distribution
of vibration frequencies. Thus

1  
       / kT   g    d 
0
2 e 1 

2N 2
Mean energy of a harmonic 
 max    for 1D
2 1/ 2

oscillator

One can obtain same expression of g ( ) by means of using

running waves.

It should be better to find 3D DOS in order to compare the

results with experiment.
3D DOS
Let’s do it first for 2D
Then for 3D.
Consider a crystal in the shape of 2D box with crystal lengths of L.

y ky

+ -  
L L
L
- +

+ -
0 x kx
L

Standing wave pattern for a Configuration in k-space

2D box
•Let’s calculate the number of modes within a range of
wavevector k.
•Standing waves are choosen but running waves will lead
same expressions.
•Standing waves will be of the form

U  U 0 sin  k x x  sin  k y y 

• Assuming the boundary conditions of

•Vibration amplitude should vanish at edges of
x  0; y  0; x  L; y  L
positive integer
Choosing
p q
kx  ; ky 
L L
 y ky
+ -  
L L
L - +
+ -
0 x kx
L

Standing wave pattern for Configuration in k-space

a 2D box
•The allowed k values lie on a square lattice of side  / L in
the positive quadrant of k-space.
•These values will so be distributed uniformly with a density
of   per unit area.
2
L / 
• This result can be extended to 3D.
 L
Octant of the crystal:
kx,ky,kz(all have positive values)

The number of standing waves;

L 3
L 3 V 3
s  k  d k    d k  3 d k
3

L   
1
L /  4 k 2 dk
8
kz V 1
 s  k  d k  3  4 k 2 dk
3

 8
2
Vk
 s  k  d 3k  2 dk
dk

2
k ky Vk 2
S  k   2
2
kx
 Vk 2
•   k  2 is a new density of states defined as the
2
number of states per unit magnitude of in 3D.This eqn

can be obtained by using running waves as well.

• (frequency) space can be related to k-space:
dk
g    d     k  dk g      k
d

Let’s find C at low and high temperature by means of

using the expression of g    .

High and Low Temperature Limits

Each of the 3N lattice d

  3Nk BT modes of a crystal C C  3 Nk B
containing N atoms dT

 This result is true only if T 
kB
At low T’s only lattice modes having low frequencies
can be excited from their ground states;

Low frequency long 

sound waves

  vs k vs 
0  k k
a
  k 1 dk 1 Vk 2 dk
vs      and g   2
k  vs d vs 2 d 

 2 
V 2 
 vs  1
g    at low T’s
2 2 vs

vs depends on the direction and there are two transverse,

one longitudinal acoustic branch:
V2 1 V2  1 2 
g     g    2  3
 3
2 vs
2 3
2  vL vT 

Velocities of sound in
longitudinal and
transverse direction
 
1  
       / kT   g    d Zero point energy=  z
0
2 e 1 

1   V  2  1 2 
       / kT   2  3  3  d 
0
2 e  1  2  vL vT 
 V  1 2    

 3 x
   z  2  3  3      / kT d 
 2  vL vT    0  e  1  k BT
 
3 k BT
 k BT  3  x
  
  x 
 3    k T
0 e / kT  1 d   0 e x  1 4 4 dx B
k BT
V  1 2   k BT   d  dx
   z 3 2  3  3  4  3 3 
 2  vL  kvBTT   x 15
0 e / kT  1 d 2 3 0 e x  1dx 3
d  V2  2 1  12  2 4 k3BT 
Cv   V3 k3B  33  kB34 154T  at low temperatures
dT 15 30  vL  vvLT  vT    
How good is the Debye approximation
at low T?
3
d 2  1 2   k BT 
Cv   V  kB  3  3  
2

dT 15 v
 L vT    

The lattice heat capacity of solids thus

varies as T 3 at low temperatures; this is
referred to as the Debye T 3 law.
Figure illustrates the excellent aggrement
of this prediction with experiment for a
non-magnetic insulator. The heat
capacity vanishes more slowly than the
exponential behaviour of a single
harmonic oscillator because the vibration
spectrum extends down to zero
frequency.
The Debye interpolation
scheme
The calculation of g ( ) is a very heavy calculation for 3D, so it
must be calculated numerically.
Debye obtained a good approximation to the resulting heat
capacity by neglecting the dispersion of the acoustic waves, i.e.
assuming
  s k
for arbitrary wavenumber. In a one dimensional crystal this is
equivalent to taking g ( ) as given by the broken line of density of
states figure rather than full curve. Debye’s approximation gives the
correct answer in either the high and low temperature limits, and the
language associated with it is still widely used today.
 The Debye approximation has two main steps:

1. Approximate the dispersion relation of any branch by a linear

extrapolation of the small k behaviour:

Einstein Debye
approximation approximation
to the to the
dispersion dispersion
  vk
 Debye cut-off frequency D
2. Ensure the correct number of modes by imposing a cut-off
frequency D, above which there are no modes. The cut-off
freqency is chosen to make the total number of lattice modes
correct. Since there are 3N lattice vibration modes in a crystal
having N atoms, we choose D so that
D

 g ( )d   3 N
0 
V 1 2
2 V 1 2 D 2
2 2 vL3 vT3 0
g ( )  (  ) (  )  d  3N
2 2 vL3 vT3

V 1 2 V 1 2 3N 9N
(  ) D3  3 N (  )  3 
6 2 3
vL vT 3 2 2 vL3 vT3 D3 D3

9N 2
g ( )  3  g ( ) /  2
D
The lattice vibration energy of

1 
E   (    / kBT )g ( )d
0
2 e 1
becomes
9N
D
1  9 N  D
  3 D
 3 
E 3 0 ( 2   e / kBT  1) d  D3  0 2 d  0 e / kBT  1d 
2

D  
and,
D
9 9N  3 d 
E  N  D  3
8 D 
0
e  / k B T  1
First term is the estimate of the zero point energy, and all T dependence is
in the second term. The heat capacity is obtained by differentiating above eqn
wrt temperature.
 dE
The heat capacity is C
dT
D
9 9N
D
 d 3
dE 9 N  2 4 e / kBT
E  N  D  3  CD   3  d
dT D  
2 2
8 D 0
e  / k B T  1 0
k BT e / kBT  1

Let’s convert this complicated integral into an expression for

the specific heat changing variables to x

 d kT 
kT
x  x
k BT dx  

and define the Debye temperature D

 D
D 
kB
The Debye prediction for lattice specific heat

4 D / T
dE 9 N k BT  k BT     2
x 4e x
CD   3    2   dx
dT D      k BT   e  1
x 2
0

3  /T
 T  D
x 4e x
CD  9 Nk B    dx
   1
2
 D 0 e x

 D
where D 
kB
How does Climit
D at high and low temperatures?

High temperature T  D
x2 x3
X is always small e  1 x 
x
 
2! 3!

x 4e x x 4 (1  x) x 4 (1  x)
   x 2

   
2 2 2
e x
 1 1  x  1 x

3  /T
 T  D

T  D  CD  9 Nk B 

 D
 
0
x 2 dx  3Nk B
How does CD limit at high and low temperatures?

Low temperature T  D

For low temperature the upper limit of the integral is infinite; the
integral is then a known integral of 4. 4 /15

3  /T
 T  D
x 4e x
T  D  CD  9 Nk B    dx
 e  1
2
 D 0
x

3
We obtain the Debye T law in the form
3
12 Nk B 4  T 
CD   
5 
 D
 Lattice heat capacity due to Debye interpolation
scheme
Figure shows the heat capacity C
between the two limits of high and 3 Nk B T
low T as predicted by the Debye 1
interpolation formula.
3  /T
 T  D x 4e x
CD  9 Nk B    dx
  D  0  e  1
x 2

Because it is exact in both high and low T

limits the Debye formula gives quite a good
representation of the heat capacity of most solids, 1
T / D
even though the actual phonon-density of states Lattice heat capacity of a solid as
curve may differ appreciably from the Debye predicted by the Debye interpolation
scheme
assumption.
Debye frequency and Debye temperature scale with the velocity of sound in
the solid. So solids with low densities and large elastic moduli have high  D . Values of
for D various solids is given in table. Debye energy  D can be used to
estimate
the maximum
Solid Ar phononNa
energy inCs
a solid. Fe Cu Pb C KCl
D (K ) 93 158 38 457 343 105 2230 235
Anharmonic Effects
 Any real crystal resists compression to a smaller volume than its equilibrium value
more strongly than expansion due to a larger volume.
 This is due to the shape of the interatomic potential curve.
 This is a departure from Hooke’s law, since harmonic application does not produce
this property.
 This is an anharmonic effect due to the higher order terms in potential which are
ignored in harmonic approximation.

 r  a
2
 d 2V 
V (r )  V (a )   2   ....................
2  dr r  a

 Thermal expansion is an example to the anharmonic effect.

 In harmonic approximation phonons do not interact with each other, in the absence
of boundaries, lattice defects and impurities (which also scatter the phonons), the
thermal conductivity is infinite.
 In anharmonic effect phonons collide with each other and these collisions limit
thermal conductivity which is due to the flow of phonons.
Phonon-phonon collisions
The coupling of normal modes by the unharmonic terms in the
interatomic forces can be pictured as collisions between the phonons
associated with the modes. A typical collision process of

phonon1 3 , k3
1 , k1
After collision another phonon is
produced

2 , k 2
phonon2 3  1  
and
2 k3  k1  k2
3  1  2 conservation of energy

k3  k1  k2 conservation of momentum

Phonons are represented by wavenumbers with
 
 k
a a
If k3 lies outside this range add a suitable multible of 2 to
a
bring  
 k k3  k1  k2
a a
it back within the range of . Then,
This phonon is indistinguishable n 2
becomes
k3 k3   k1  k2
from a phonon with wavevector a
where k1 , k2 , and k3 are all in the above range.
Longitudinal  3 3'  3
1 1 2
Transverse 2
k k
 0   0 
 
a a a a
n Normal
0  process n Umklapp
0  process
(due to anharmonic effects)
 
Phonon3 has k  ; Phonon3 has k  and Phonon3=Phonon3’
a a
 Thermal conduction by phonons
 A flow of heat takes place from a hotter region to a cooler region
when there is a temperature gradient in a solid.
The most important contribution to thermal conduction comes from
the flow of phonons in an electrically insulating solid.
Transport property is an example of thermal conduction.
Transport property is the process in which the flow of some
quantity occurs.
Thermal conductivity is a transport coefficient and it describes the
flow.
The thermal conductivity of a phonon gas in a solid will be
calculated by means of the elementary kinetic theory of the transport
coefficients of gases.
Kinetic theory
In the elementary kinetic theory of gases, the steady state flux of a property P
in the z direction is
1 _ dP
flux  l 
3 dz
Angular average
Constant average speed for
Mean free path
molecules
In the simplest case where P is the number density of particles the transport
coefficient obtained from above eqn. is the diffusion coefficient D  1 l  .
_

3
If P is the energy densityE then the flux W is the heat flow per unit area so
that 1 _ dE 1 _ dE dT
W  l  l
3 dz 3 dT dz
Now dE / dT is the specific heat C per unit volume, so that the thermal
conductivity;
1 _
K  l C Works well for a phonon gas
3
 Heat conduction in a phonon and real gas
The essential differences between the processes of heat
conduction in a phonon and real gas;

Phonon gas Real gas

•Speed is approximately constant. •No flow of particles

•Both the number density and energy •Average velocity and kinetic energy per
density is greater at the hot end. particle are greater at the hot end, but the
number density is greater at the cold end,
•Heat flow is primarily due to phonon
and the energy density is uniform due to the
flow with phonons being created at the
uniform pressure.
hot end and destroyed at the cold end
•Heat flow is solely by transfer of kinetic
energy from one particle to another in
collisions which is a minor effect in phonon
hot cold case.

hot cold
Temperature dependence of thermal conductivity K
1 _
K  l C Approximately equal to
Vanishes exponentially at
3 velocity of sound and so
temperature independent.
low T’s and tends to
classical value k B at high T’s

•Temperature dependence of phonon mean free length is determined by

phonon-phonon collisions at low temperatures
•Since the heat flow is associated with a flow of phonons, the most effective
collisions for limiting the flow are those in which the phonon group velocity
is reversed. It is the Umklapp processes that have this property, and these are
important in limiting the thermal conductivity
Conduction at high temperatures
 At temperatures much greater then the Debye temperature  the heat capacity is
D
given by temperature-independent classical result of

C  3 Nk B
 The rate of collisions of two phonons phonon density.
 If collisions involving larger number of phonons are important, however, then the
scattering rate will increase more rapidly than this with phonon density.

 At high temperatures the average phonon density is constant and

the total lattice energy  T; phonon number  T , so
Scattering rate  T and mean free length  T 1

Then the thermal conductivity of

1 _
K  l C  T 1.
3
 Experimental results do tend towards this behaviour at high temperatures
as shown in figure (a).

10

10 0

0
1
10-1 T3
T
10-1

5 10 20 50 100 2 5 10 20 50 100
T (K ) T (K )

(a)Thermal conductivity of a quartz (b)Thermal conductivity of artificial

crystal sapphire rods of different diameters
Conduction at intermediate temperatures

Referring figure a
At T<  D ; the conductivity rises more steeply with falling temperature,
although the heat capacity is falling in this region. Why?

This is due to the fact that Umklapp processes which will only occur if there are
phonons of sufficient energy to create a phonon with k3   / a. So

Energy of phonon must be  the Debye energy ( k D )

The energy of relevant phonons is thus not sharply defined but their number is
expected to vary roughly as
e  D / bT when T  D ,
where b is a number of order unity 2 or 3. Then
l  e D / bT
This exponential factor dominates any low power of T in thermal conductivity,
3
such as a factor of T from the heat capacity.
Conduction at low temperatures

l for phonon-phonon collisions becomes very long at low T’s and eventually
exceeds the size of the solid, because

number of high energy phonons necessary for Umklapp processes decay

exponentially as

 D / bT
l e
is then limited by collisions with the specimen surface, i.e.

l  Specimen diameter

T dependence of K comes from Cv which obeys T3 law in this region

3
12 Nk B  T 
4
Temperature dependence of Cv dominates.
CD   
5 
 D
Size effect
 When the mean free path becomes comparable to the dimensions of the sample,
transport coefficient depends on the shape and size of the crystal. This is known as a
size effect.

 If the specimen is not a perfect crystal and contains imperfections such as dislocations,
grain boundaries and impurities, then these will also scatter phonons. At the very
lowest T’s the dominant phonon wavelength becomes so long that these imperfections
are not effective scatterers, so;
the thermal conductivity has a T 3 dependence at these temperatures.
 D / bT
 The maximum conductivity between T 3 and e region is controlled by
imperfections.

 For an impure or polycrystalline specimen the maximum can be broad and low
[figure (a) on pg 59], whereas for a carefully prepared single crystal, as illustrated in
figure(b) on pg 59, the maximum is quite sharp and conductivity reaches a very high
value, of the order that of the metallic copper in which the conductivity is
predominantly due to conduction electrons.