1:0 Phonons

In this chapter we will tum our attention to the interaction between light and the
phonons in a solid. Phonons are vibrations of the atoms in a crystal lattice, and
have resonant frequencies in the infrared spectral region. This contrasts with
10.1 Infrared active phonons 204 the optical properties of bound electrons, which occur at visible and ultraviolet
10.2 Infrared reflectivity and frequencies.
absorption in polar The main optical properties of phonons can be explained to a large extent by
solids 206 classical models. We will therefore make extensive use of the classical dipole
10.3 Polaritons 214 oscillator model developed in Chapter 2. This will allow us to understand
10.4 Polarons 215 why polar solids reflect and absorb light strongly within a band of infrared
10.5 Inelastic light scattering 218 frequencies. We will then introduce the concepts of polaritons and polarons,
10.6 Phonon lifetimes 222 before moving on to discuss the physics of inelastic light scattering. We will
see how Raman and Brillouin scattering techniques give us complementary
information to infrared reflectivity data, which is why they are so extensively
used in phonon physics. Finally we will briefly discuss why phonons have
a finite lifetime, and how this affects the reflectivity and inelastic scattering
spectra.
We will assume that the reader is familiar with the basic physics of phonons,
which is covered in all introductory solid state physics texts. A partial list of
suitable preparatory reading is given under Further Reading at the end of the
chapter.

10.1 Infrared active phonons

The atoms in a solid are bound to their equilibrium positions by the forces that
hold the crystal together. When the atoms are displaced from their equilibrium
positions, they experience restoring forces, and vibrate at characteristic fre­
quencies. These vibrational frequencies are determined by the phonon modes
of the crystal.
The resonant frequencies of the phonons occur in the infrared spectral re­
gion, and the modes that interact directly with light are called infrared active
The group theory approach is beyond the (IR active). Detailed selection rules for deciding which phonon modes are IR
scope of this book, although we will give active can be 9erived by using group theory. At this level we just discuss the
some simple arguments based on symmetry
when we consider inelastic light scattering in
general rules based on the dispersion of the modes, their polarization, and the
Section 10.5. nature of the bonding in the crystal.
The phonon modes of a crystal are subdivided into two general categories:
• acoustic or optical;
• transverse or longitudinal.
It will come as no surprise to realize that it is the 'optical' rather than the
acoustic modes that are directly IR active. These optically active phonons are
 10.1 Infrared active phonons 205

able to absorb light at their resonant frequency. The basic process by which
a photon is absorbed by the lattice and a phonon is created is represented in ----1

co, k
•�
►
Fig. 10.1. Conservation laws require that the photon and the phonon must Q, q
have the same energy and momentum. We will see below that this condition
Fig. 10.1 Lattice absorption process by an
can only be satisfied for the optical modes. infrared active phonon. The straight arrow
Figure 10.2 shows the generic dispersion curves for the acoustic and optical represents the photon that is absorbed, while
phonons in a simple crystal. The angular frequency r.l of the acoustic and the wiggly arrow represent the phonon that is
optical phonons is plotted against the wave vector q in the positive half of created.
the first Brillouin zone. At small wave vectors the slope of the acoustic branch
is equal to Vs , the velocity of sound in the medium, while the optical modes are
essentially dispersionless near q = 0. The phonon dispersion curves for real crys­
The figure also shows the dispersion of the light waves in the crystal, which tals are more complicated than those shown
have a constant slope of v = c / n, where n is the refractive index. The refractive in Fig. 10.2 because the longitudinal and
transverse polarizations tend to have different
index has been highly exaggerated here in order to make the dispersion of the frequencies.
photon noticeable on the same scale as the phonon dispersion. The requirement
that the photon and phonon should have the same frequency and wave vector is
satisfied when the dispersion curves intersect. Since c / n » V s , the only inter­
section point for the acoustic branch occurs at the origin, which corresponds to
the response of the crystal to a static electric field. The situation is different for
the optical branch: there is an intersection at finite w, which is identified with
the circle in Fig. 10.2. Since the optical branch is essentially flat for small q,
the frequency of this resonance is equal to the frequency of the optical mode
at q = 0.
Electromagnetic waves are transverse, and can only apply driving forces
to the transverse vibrations of the crystal. Therefore they can only couple to
the transverse optic (TO) phonon modes. This does not mean that we can now
completely forget about the longitudinal optic (LO) phonons. As we will see in
Section 10.2.2, the LO modes do in fact play an important role in the infrared
properties of crystals.
Photons couple to phonons through the driving force exerted on the atoms
by the AC electric field of the light wave. This can only happen if the atoms
are charged. Therefore, if the atoms are neutral, there will be no coupling to

..
t"t;,-----�optical branch

acoustic branch
Fig. 10.2 Dispersion curves for the acoustic
and optical phonon branches in a typical crys­
7t tal with a lattice constant of a. The dispersion
0 Wave vector (q) of the photon modes in the crystal is shown
a by the dotted line.
206 Phonons

the light. This means that the crystal must have some ionic character in order
for its TO phonons to be optically active.
The ionicity of a solid arises from the way the crystal binding occurs. An
ionic crystal consists of an alternating sequence of positive and negative ions
held together by their mutual Coulomb attraction. Covalent crystals, by con­
trast, consist of neutral atoms with the electrons shared equally between the
Table 10.1 Infrared activity of the neighbouring nuclei. This means that none of the optical phonons of purely
phonon modes in polar and non­
polar crystals. LA: longitudinal covalent solids like silicon are IR active. Most other materials fall somewhere
acoustic, TA: transverse acoustic, between these two limits. For example, the bond in a III-V semiconductor is
LO: longitudinal optic, TA: trans­ only partly covalent, and the shared electrons lie slightly closer to the group V
verse optic. atoms than to the group III atoms, which gives the bond a partly ionic character.
The bonds with an ionic character are called polar bonds to stress the point
Mode Polar Non-polar
crystal crystal
that the asymmetric electron cloud between the atoms creates a dipole that can
interact with electric fields. Provided the bond has some polar character, its
LA no no phonons can be IR active.
TA no no The conclusions of this section are summarized in Table 10.1.
LO no no
TO yes no

10.2 Infrared reflectivity and absorption in polar

solids
Experimental data show that polar solids absorb and reflect light very strongly
in the infrared spectral region when the frequency is close to resonance with the
TO phonon modes. We have come across several examples of this already. For
example, the transmission spectra of sapphire and CdSe given in Fig. 1.4 show
that there are spectral regions in the infrared where no light is transmitted. This
is a consequence of lattice absorption.
The aim of this section is to account for this result by modelling the interac­
tion of photons with TO phonons. To do this we will make extensive use of the
classical oscillator model developed in Chapter 2, especially Section 2.2. This
will allow us to calculate the frequency dependence of the complex dielectric
constant Er(w), from which we will be able to determine the important optical
properties such as the reflectivity and absorption.

10.2.1 The classical oscillator model

The interaction between electromagnetic waves and a TO phonon in an ionic
crystal is most easily treated by considering a linear chain, as illustrated in
Fig. 10.3. The chain consists of a series of unit cells, each containing a positive
ion (black circle) and a negative ion (grey circle). The waves are taken to be
propagating along the chain in the z direction. We are dealing with a trans­
verse mode, and so the displacement of the atoms is in the x or y directions.
Furthermore, in an optic mode the different atoms within each unit cell move
in opposite directions, with a fixed ratio between their displacements which is
not necessarily equal to unity.
We are interested in the interaction between a TO phonon mode with q �
LO modes do not interact with light because
the displacement of the atoms generates a 0 and an infrared light wave of the same frequency and wave vector. This
longitudinal electric field, which is perpen­ means that we are considering phonons with a very long wavelength of ~
dicular to that of the light wave. 10 µ,m matched to that of an infrared photon. This phonon wavelength is huge
 10.2 Infrared reflectivity and absorption in polar solids 207

Fig. 10.3 Interaction of a TO phonon mode

propagating in the z direction with an electro­
z magnetic wave of the same wave vector. The
black circles represent positive ions, while
the grey circles represent the negative ions.
The solid line represents the spatial depen­
dence of the electric field of the electromag­
netic wave.

compared to the size of a unit cell in a crystal, which is usually less than
10- 9 m. The size of the atoms has been highly exaggerated in Fig. 10.3 to
make the physics of the interaction clearer. In fact, the real size of the atoms
is tiny compared to the wavelength, and there will be thousands of unit cells
within one period of the wave.
The solid line in the figure represents the spatial dependence of the AC
electric field of the infrared light wave. At resonance, the wave vector of
the photon and the phonon are the same. This means that the driving force
exerted by the light on the positive and negative ions is in phase with the lattice
vibration. At the same time, the antiparallel displacements of the oppositely
charged atoms generate an AC electric field in phase with the external light.
This implies that there is a strong interaction between the TO phonon mode
and the light wave when the wave vectors and frequencies match.
For long wavelength TO modes with q � 0, the motion of the atoms in
different unit cells is almost identical, and we therefore need to concentrate on
what is happening within the unit cell itself. This enables us to see that there is
a close connection between the TO phonons at q = 0 and the vibrational modes
of the molecules from which the crystal is formed. We can therefore make use The data for Si02 glass shown in Fig. 2.7
of some of the principles developed in molecular physics, for example: the illustrates the connection between the in­
frared absorption in solids and that of the
selection rules for deciding whether a particular phonon mode is IR or Raman constituent molecules quite well. The glass is
active. (cf. Section 10.5.2.) amorphous, and therefore does not have long
The interaction between the TO phonon and the light wave can be modelled range order with delocalized phonon modes.
by writing down the equations of motion for the displaced ions. The displace­ The absorption in the range 10 1 3-10 14 Hz
is basically caused by the vibrational ab­
ments of the positive and negative ions in a TO mode are in opposite directions sorption of the Si02 molecules themselves,
and are given the symbols x + and x_ respectively, as indicated in Fig. 10.3. although the frequencies are not necessarily
The appropriate equations of motion are: exactly the same in the solid as in the free
molecule.

(10.1)

(10.2)

where m+ and m_ are the masses of the two ions, K is the restoring constant
of the medium, and 8(t) is the external electric field due to the light wave. The For a strongly ionic crystal such as NaCl,
effective charge per ion is taken to be ±q. q would just be equal to ±e. However, for
By dividing eqn 10.1 by m+ and eqn 10.2 by m_, and then subtracting, we crystals with polar covalent bonds such as
the III-V compounds, q will represent an
obtain: effective charge which is determined by the
d2 K q
- (x + - x_) = --(x+ - x_) + - 8(t), (10.3) asymmetry of the electron cloud within the
dt 2
µ µ bond.
208 Phonons

whereµ., is the reduced mass given by

1 1 1
-=-+-.
m m_
(10.4)
µ, +

By putting x = x+ - x_ for the relative displacement of the positive and

negative ions within their unit cell, we can recast eqn 10.3 in the simpler form:

d2 x
-
2
+ QTO = -q S(t), (10.5)
dt 2
x

µ,
where we have written Qiu for K / µ,. S1To represents the natural vibrational
frequency of the TO mode at q = 0 in the absence of the external light field.
Equation 10.5 represents the equation of motion for undamped oscillations
of the lattice driven by the forces exerted by the AC electric field of the light
wave. In reality, we should have incorporated a damping term to account for the
finite lifetime of the phonon modes. The physical significance of the phonon
lifetime will be discussed further in Section 10.6. At this stage, we simply
introduce a phenomenological damping rate y, and rewrite eqn 10.5 as
d2 x d q
+ y dt + Q2TO = µ S(t). (10.6)
x

dt 2 x

This now represents the response of a damped TO phonon mode to a resonant

light wave.
Equation 10.6 is identical in form to eqn 2.5 in Chapter 2, with mo replaced
byµ.,, wo by S1rn and -e by q. Therefore, we can use all the results derived in
Section 2.2 to model the response of the medium to a light field of angular fre­
quency w with c(t) = Soeiwt . In particular, we can go directly to the formula
for the frequency dependence of the dielectric constant without repeating all
the steps in the derivation. By adapting the symbols appropriately in eqn 2.14,
we immediately write down:

q N 2
1
Er(w) = 1 + X+ - 2 - W 2 - 1yw (10.7)
fQ/J., ( QTO . ) ,

where Er(w) is the complex dielectric constant at angular frequency w. x rep­

resents the non-resonant susceptibility of the medium, and N is the number of
unit cells per unit volume.
Equation 10.7 can be tidied up by introducing the static and high frequency
dielectric constants Est and E oo respectively. In the limits of low and high
frequency, we obtain from eqn 10.7:

(10.8)

and
In principle, we should consider the local E 00 = Er(oo) = 1 + X. (10.9)
field corrections discussed in Section 2.2.4
here. This is an unnecessary complication at Thus we can write:
this level which does not add much to the
main conclusions. We will therefore neglect
local field effects, and base our discussion on (10.10)
eqn IO.IO.
 10.2 Infrared reflectivity and absorption in polar solids 209

This is our main result, which will be used in the next subsections to derive the
infrared optical coefficients. As discussed in Section 2.2.2, and in particular
in connection with Fig. 2.6, we should understand 'w = oo' in a relative
sense here. Eoo represents the dielectric constant at frequencies well above the
phonon resonance, but below the next natural frequency of the crystal due, for
example, to the bound electronic transitions in the visible/ultraviolet spectral
region.

10.2.2 T he Lyddane-Sachs-Teller relationship

Before working out the frequency dependence of the infrared reflectivity, it is
useful to investigate one rather striking implication of eqn 10.10. Suppose we
have a lightly damped system so that we can set y = 0. Then at a certain
frequency which we label w', eqn 10.10 tells us that the dielectric constant can
fall to zero. The condition for this to happen is:

(10.11)

This can be solved to obtain:

W
1
= (-
Est ½,--,
) ��TO- (10.12)
Eoo

What does Er = 0 mean physically ? In a medium with no free charges, the

total charge density will be zero. Hence Gauss's law (eqn A.10) tells us that We came across another situation in which
Er is zero when we discussed plasmons in
v' · D = v' · (ErEo€) = 0, (10.13) Section 7.5. We saw there that the plasma
oscillations that cause Er to be zero also cor­
respond to longitudinal electric field waves.
where we have made use of eqn A.3 to relate the electric displacement D to
the electric field € in a dielectric medium. When we consider the propagation
of electromagnetic waves through the dielectric, we look for wave solutions of
the form:
€(r, t)
= €a ei (k-r-wt). (10.14)
On substituting eqn 10.14 into eqn 10.13, we usually assume that Er i- 0 and
therefore conclude that k · € = 0. This tells us that the electric field must
be perpendicular to the direction of the wave and therefore that the waves
are transverse. However, if Er = 0, we can satisfy eqn 10.13 with waves in
which k • € i- 0, that is, with longitudinal waves. Thus we conclude that the
dielectric can support longitudinal electric field waves at frequencies which
satisfy Er(w) = 0.
In the same way that TO phonon modes generate a transverse electric field
wave, the LO phonon modes generate a longitudinal electric field wave. Thus
the waves at w = w' correspond to LO phonon waves, and we identify w' with
the frequency of the LO mode at q = 0, namely QLO· This allows us to rewrite
eqn 10.12 in the following form:

Est
(10.15)
Eoo
210 Phonons

Table 10.2 Comparison of the measured This result is known as the Lyddane-Sachs-Teller (LST) relationship. The va­
ratio nLo /nTo for several materials
to the value predicted by the Lyddane­ lidity of the relationship can be checked by comparing the values of QLO / QTO
Sachs-Teller relationship. After [I]. deduced from neutron or Raman scattering experiments with those calculated
from eqn 10.15 using known values of the dielectric constants. Some results
are given in Table 10.2. It is apparent that the agreement is generally very good.
Crystal nw/nrn (Est/Eoo ) 2
An interesting corollary of the LST relationship is that it implies that the
Si LO phonon and TO phonon modes of non-polar crystals are degenerate. This
GaAs 1.07 1.08 follows because there is no infrared resonance, and therefore E st = E 00 . This
A!As 1.12 I.II
BN 1.24 1.26 is indeed the case for the purely covalent crystals of the group IV elements,
ZnSe 1.19 1.19 namely diamond (C), silicon and germanium.
MgO 1.81 1.83
AgF 1.88 1.88

10.2.3 Restrahlen
Having discussed the properties of the system at the special frequency of w =
nLo, we can now calculate the infrared optical constants. It is easier to under­
stand the general behaviour if we assume that the damping term is small. We
thus set y = 0 in eqn 10.10, and discuss the properties of a material with a
dielectric constant that has the following frequency dependence:

Er(v) = E 00 + (E st - Eoo) 2 2) · (10.16)

( VTO - V

We have divided all the angular frequencies by 2:rr here, so that we can com­
pare the predictions to experimental data, which are usually presented against
frequency (v) rather than angular frequency (w). We will discuss the effect of
including the damping term when we compare our model to the experimental
data in connection with Fig. 10.5.
Figure 10.4(a) plots the frequency dependence of the dielectric constant
Er(v) calculated from eqn 10.16 for a polar crystal with the following parame­
1 THz = 10 12 Hz ters: VTo = lOTHz, VLo = 11 THz,E st = 12.1 and E 00 = 10. These figures are
quite close to those that would be found in a typical III-V semiconductor. Note
that the phonon frequencies have been chosen to satisfy the LST relationship
given in eqn 10.15.

1.0
40
0.8
(a) (b)
20 0
:� 0.6
w....
0 � 0.4
0.2
-20
0.0
6 8 10 12 14 6 12 14
Frequency (THz) Frequency (THz)

Fig. 10.4 Frequency dependence of the dielectric constant and reflectivity for a crystal with VTO = 10 THz, VLO = 11 THz, Est = 12.1 and
E oo = 10. The curves have been calculated from eqns 10.16 and 10.17. Phonon damping is ignored in this calculation.
 l 0.2 Infrared reflectivity and absorption in polar solids 211

0.8

:� 0.6
(.)
(!.)
i:;:::::
0.4
�
0.2 Fig. 10.5 Infrared reflectivity of InAs and
GaAs at 4.2 K. A wave number of I cm-1 is
0.0 ......._........._......._____,___�..__-....___, equivalent to a frequency of 2.998 x 10 10 Hz.
200 220 240 260 280 300 After [2], copyright © 1967 by Academic
Press, reproduced by permission of the pub­
Wave number (cm- 1) lisher.

At low frequencies the dielectric constant is just equal to Est• As v increases

from 0, Er(v) gradually increases until it diverges when the resonance at VTO
is reached. Between VTO and VLO, Er is negative. Precisely at v = VLO, Er = 0.
Thereafter, Er is positive, and gradually increases asymptotically towards the
value of E 00 • We can see from eqn 1.17 that .jEoo corre­
The most important optical property of a polar solid in the infrared spectral sponds to the refractive index of the medium
at frequencies well above the optical phonon
region is the reflectivity. This can be calculated from the dielectric constant

I l I JEr _ 1
resonances. This will be the refractive index
using eqn 1.26: measured at near-infrared and visible fre­
ii _ 1 2 1 2 quencies the below band gap of the material.
(10.17)
R= ii+l = JEr+l
Figure 10.4(b) plots the reflectivity calculated using eqn 10.17 for the di­
electric constant shown in Fig. 10.4(a). At low frequencies the reflectivity is
(.JEst - 1) 2 /(.JEst + 1)2 . As v approaches VTO, R increases towards unity.
In the frequency region between VTO and VLO, JEr
is imaginary, so that R
remains equal to unity. R drops rapidly to zero as v increases above VLO
(see Exercise 10.2), and then increases gradually towards the high frequency
asymptote of (Foo - 1) 2 /(Foo+ 1) 2 .
We see from this analysis that the reflectivity is equal to 100 % in the
frequency region between VTo and vw. This frequency region is called the
restrahlen band. Restrahlen is the German word for 'residual rays'. Light
cannot propagate into the medium in the restrahlen band.
Figure 10.5 shows experimental data for the reflectivity of InAs and GaAs
in the infrared spectral region. InAs has TO and LO phonon frequencies at Experimental infrared spectra are frequently
218.9 cm- 1 and 243.3 cm- 1 respectively, while for GaAs we have VTO plotted against the wave number v = 1 / A.
273.3cm-1 and VLo = 297.3 cm-1. We see that the reflectivity is very high = The wave number is effectively a frequency
unit, with I cm-1 equivalent to 2.998 x
for frequencies between the TO and LO phonon frequencies in both materials, 10 10 Hz.
and there is a sharp dip in the reflectivity just above the LO phonon resonance.
On comparing these results with the prediction shown in Fig. 10.4(b), we see
that the general agreement between the model and the experimental data is very
good. The main difference is that in both materials the maximum reflectivity
in the restrahlen band is less than 100 %. This reduction in the reflectivity is
caused by ignoring the damping term. (See Example 10.1 and Exercise 10.4.)
The damping also broadens the edge so that there is only a minimum in R just
above VLO rather than a zero.
212 Phonons

The magnitude of y can be found by fitting the experimental data to the full
dependence given in eqn 10.10. The values of y obtained in this way are around
1011 -1012 s-1 , which implies that the optical phonons have a lifetime of about
lps=I0-12s. 1-10 ps. The physical significance of this short lifetime will be discussed in
Section 10.6.

10.2.4 Lattice absorption

When we introduced the classical oscillator model in Section 2.2 of Chapter 2,
we made the point that we expect high absorption coefficients whenever the
frequency matches the natural resonances of the medium. The reader might
therefore be wondering why we have been concentrating on calculating the
reflectivity rather than the absorption due to the TO phonon resonances.
This question is further prompted by recalling the analogy between the
infrared absorption of polar solids and that of isolated molecules. In both
cases we are basically treating the interaction of photons with quantized vi­
brational modes. In molecular physics we usually discuss this in terms of
the infrared absorption spectrum. The absorption spectra show strong peaks
whenever the frequency coincides with the infrared active vibrational modes
and the molecule can absorb a photon by creating one vibrational quantum.
This is directly analogous to the process for solids shown in Fig. 10.1 in which
a photon is absorbed and a phonon is created.
The answer to these questions is that the lattice does indeed absorb very
strongly whenever the photon is close to resonance with the TO phonon. As
stressed in Chapter 2, the fundamental optical properties of a dielectric - the
absorption, refraction and reflectivity - are all related to each other because
they are all determined by the complex dielectric constant. The distinction
between absorption and reflection is merely a practical one. Polar solids have
such high absorption coefficients in the infrared that unless the crystal is less
than ~ 1 µ,m thick, no light at all will be transmitted. This is clearly seen
in the transmission spectra of A}zO3 and CdSe shown in Fig. 1.4. For this
reason, it is only sensible to consider lattice absorption in thin film samples. In
thick crystals, we must use reflectivity measurements to determine the vibra­
tional frequencies. This contrasts with molecular physics, where we are usually
dealing with low density gases, which give rise to much smaller absorption
coefficients.
The absorption coefficients expected at the resonance with the TO phonon
can be calculated from the imaginary part of the dielectric constant. At w =
QTO we have from eqn 10.10:
. QTO
Er(QTo) = Eoo + !(Est - Eoo)--. (10.18)
y
The extinction coefficient K can be worked out from Er using eqn 1.23, and
then the absorption coefficient a can be determined from K using eqn 1.16.
Typical values for a are in the range 106 -107 m- 1. (See Example 10.1 and
Exercise 10.6.) This is why the sample must be thinner than~ 1 µ,m in order to
perform practical absorption measurements. Infrared absorption measurements
on thin film samples do indeed confirm that the absorption is very high at the
TO phonon resonance frequency.
 10.2 Infrared reflectivity and absorption in polar solids 213

Example 10.1
The static and high frequency dielectric constants of NaCl are Est = 5.9 and
E oo = 2.25 respectively, and the TO phonon frequency VTO is 4.9THz.
(i) Calculate the upper and lower wavelengths of the restrahlen band.
(ii) Estimate the reflectivity at 50 µ,m, if the damping constant y of the
phonons is 10 12 s-1.
(iii) Calculate the absorption coefficient at 50 µ,m.

Solution
(i) The restrahlen band runs from VTO to VLO· We are given VTO, and we can
calculate VLo from the LST relationship (eqn 10.15). This gives

VLO = (.:i!_) i
E 00
x VTO =( · ) i X 4.9THz
59
2.25
= 7.9THz .
Therefore the restrahlen band runs from 4.9 THz to 7.9 THz, or 38 µ,m
to 61 µ,m.
(ii) At 50 µ,m we are in middle of the restrahlen band. We therefore expect
the reflectivity to be high. We insert the values for Est, E oo , y and QTO =
2n VTo into eqn 10.10 with w = 2n v ( v = 6THz) to find:
(4.9)2
Er= 2.25 + 3.65 = -5.0 + 0.57i .
- 62 - i(1)(6)/27l'
(4.9)2
We then obtain the real and imaginary parts of the refractive index from
eqns 1.22 and 1.23:
1 1 1
n= ( - 5.0 + [(-5.0)2 + (0.57)2 ]2)2 = 0.13,
Ji.
and
1
= + (0.57)2 ] 2) 2 = 2.2.
1 1
K ( + 5.0 + [(-5.0)2
Jj_
We finally substitute these values of n and K into eqn 1.26 to find the
reflectivity:
(n - 1)2 + K 2 (-0.87)2 + (2.2)2
R=----­ = 0.91.
(n + 1)2 + K 2 (1.13)2 + (2.2)2
This value is close to the measured reflectivity of NaCl in the restrahlen
band at room temperature.
(iii) We can calculate the absorption coefficient ct from the extinction coeffi­
cient using eqn 1.16. We have already worked out that K = 2.2 in part
(ii). Hence we find:
4 x 2.2
ct= 41l'K = n = 5.5 x 105 m-1
A 50 x I0-6
This shows that the light would be absorbed in a thickness of about
2µ,m.
214 Phonons

20.-----,---,---....----r----,r---�
/co= Vst q

co= v= q
-------------- QL0/2n
I=! 10 -
QTO/2n
---N
8 5
Fig. 10.6 Polariton dispersion predicted from
eqn 10.19 with Er given by eqn 10.16. The
curves are calculated for a crystal with lJTO = 0 --�--�--�-�--�
10 THz, Est = 12.1 and E oo = 10. The 0.0 0.5 1.0 1.5 2.0 2.5
asymptotic velocities Vst and v 00 are equal to
c/ JEst and c/ � respectively. Wave vector q (106 m- 1)

10.3 Polaritons
The dispersion curves of the photons and TO phonons were discussed in broad
terms in connection with Fig. 10.2. We now wish to consider the circled in­
tersection point in Fig. 10.2 in more detail. As we will see, the two dispersion
curves do not actually cross each other. This is a consequence of the strong
coupling between the TO phonons and the photons when their frequencies and
wave vectors match. This leads to the characteristic anticrossing behaviour
which is observed in many coupled systems.
The coupled phonon-photon waves are called polaritons. As the name sug­
gests, these classical waves are mixed modes which have characteristics of
both polarization waves (the TO phonons) and the photons. The dispersion of
the polaritons can be deduced from the relationship:
C
W= Vq = --q, (10.19)
Fr
where the second part of the equation comes from eqn A.29, with /.Lr = 1. The
resonant response of the polar solid is contained implicity in the frequency
dependence of Er.
Figure 10.6 shows the polariton dispersion calculated for a lightly damped
medium. The dielectric constant is given by eqn 10.16, and is plotted for the
same parameters as in Fig. 10.4(a). At low frequencies the dielectric constant
is equal to Est, and the dispersion of the modes is given by w = cq/ �­
As w approaches S"ho, the dielectric constant increases, and the velocity of
the waves decreases, approaching zero at S"ho itself. For frequencies in the
restrahlen band between S'ho and QLO, the dielectric constant is negative. No
modes can propagate, and all the photons that are incident on the medium
are reflected. For frequencies above QLO, Er is positive again and propagat­
ing modes are possible once more. The velocity of the waves gradually in­
creases with increasing frequency, approaching a value of c / Foo at high
frequencies.
The dispersion of the polariton modes has been measured for a number of
materials. Figure 10.7 shows the measured dispersion of the TO phonons and
LO phonons in GaP at small wave vectors. The results were obtained by Raman
scattering techniques. (See Section 10.5.2.) The experimental data reproduce
very well the polariton dispersion model indicated in Fig. 10.6. The solid line
 10.4 Polarons 215

52
--\- -0-0-- -0- -------- -0- -0

- 48 upper polariton
branch
LO phonon

'-' 44
\
lower polariton branch
40 Fig. 10.7 Dispersion of the TO and LO
GaP phonons in GaP measured by Raman scat­
tering. The solid lines are the predictions of
36 the polariton model with hl!'fo = 45.5 meV,
Eoo = 9.1 and Est = I 1.0. After [3], copy­
0 1 2 3
right 1965 American Institute of Physics,
1
Wave vector (10 m- )
6
reprinted with permission.

is the calculated polariton dispersion, which gives a very accurate fit to the
experimental points. Note that the LO phonons do not show any dispersion
here because they do not couple to the light waves.

10.4 Polarons
So far in this chapter we have been considering the direct interaction between
a light wave and the phonons in a crystal. As we have seen, this gives rise
to strong absorption and reflection in the infrared spectral region. The optical
phonons can, however, contribute indirectly to a whole host of other optical
properties that depend primarily on the electrons through the electron-phonon
coupling. In this section we will consider the polaron effect, which is one of
the most important examples of this.
Consider the motion of a free electron through a polar solid, as shown in
Fig. 10.4. The electron will attract the positive ions that are close to it, and
repel the negative ones. This produces a local displacement of the lattice in
the immediate vicinity of the electron. The lattice distortion accompanies the
electron as it moves through the crystal. The electron with its local lattice

• ---.., .,,,_ • ------ ... •

,.
·:1 ·· a..

, ·�
, •
'lliS'
--
·:.t _I, �-iii.
,.1,.,

• ' '-·• \

,.. �-
I -� ,,,1.i"
••
�
., �
I � e-
r �·-....
•
I
-.r \
6 lie
� �,.,.,. I

• '.'I ' ... ----- -- , �;

Fig. 10.8 Schematic representation of a po-
I laron. A free electron moving through an
·111111• ·�, '
Ii ·1
� ionic lattice attracts the positive (black) ions,

e� • • .. •
and repels the negative (grey) ones. This pro-
duces a local distortion of the lattice within
!, ;;··�
·� the polaron radius shown by the dashed circle
216 Phonons

distortion is equivalent to a new elementary excitation of the crystal, and is

It can be shown that the average number called a polaron.
of virtual LO phonons that move with the The polaron effect can be conceived in terms of an electron surrounded by
electron is equal to aep/2. We do not consider
the longitudinal acoustic modes here because
a cloud of virtual phonons. We think of the electron absorbing and emitting
they do not produce a polarization in the phonons as it moves through the crystal. These phonons produce the local
medium: the positive and negative ions move lattice distortion. The displacement of the ions is in the same direction as the
in the same direction, and this produces no electric field of the electron, we are therefore dealing with longitudinal optic
electric dipole moment.
phonons.
The strength of the electron-phonon interaction in a polar solid can be
The polaron theory can be applied equally to quantified by the dimensionless coupling constant aep, which is given by:
electrons or holes by taking the appropriate
effective masses in the formula:. In a non­
polar crystal such as silicon, Eoo = Est, and 1 m*c2 ½ 1 1
aep = 0. There is therefore no polaron effect.
aep = ( ) [ ] (10.20)
137 2/iQLO E oo - Est '

where 1/137 is the fine structure constant from atomic physics. The mass m*
that appears here is the usual effective mass deduced from the curvature of the
band structure (c.f. eqn C.6):

= li2 (dE -l
2
m* ) (10.21)
dk2

Values for aep for three binary compound semiconductors, namely GaAs, ZnSe
and AgCl, are given in Table 10.3. We see that the coupling constant increases
from GaAs (0.06) through ZnSe (0.40) to AgCl (2.2). This is because the ion­
icity increases as we go from the III-V semiconductor, in which the bonding
Table 10.3 Electron-phonon coupling con­
stant aep calculated from eqn 10.20 for
is predominantly covalent, to the I-VII compound, which is highly ionic.
GaAs, ZnSe, and AgCI. The figures for This effective mass given by eqn 10.21 is calculated by assuming that the
ZnSe are for the cubic crystal structure. lattice is rigid. However, the concept of a rigid lattice is only a theoretical
After [1]. one, and any experiment we perform to measure m* will actually measure the
polaron mass m** instead. This is because it is not possible to hold the lattice
GaAs ZnSe AgCI
rigid as the electron moves. The polaron mass is larger than the rigid lattice
m':/mo 0.067 0.13 0.30 mass because the electron has to drag the local lattice distortion with it as it
E oo 10.9 5.4 3.9 moves.
Est 12.4 7.6 11.1 An example of an experiment to measure the effective mass is cyclotron res­
r.!w (THz) 53.7 47.7 36.9
onance. In this technique, we measure the infrared absorption in the presence
Ciep 0.06 0.40 2.2
of a magnetic field B. As discussed in Section 3.3.6, the electron energy is
quantized in terms of the cyclotron energy:

(10.22)

where n is an integer, and

eB
We=-. (10.23)
m*
Optical transitions with �n = ±1 can take place between the ladder of levels
defined by eqn 10.22. We therefore observe absorption at a wavelength A given
by:
he
= eliB (10.24)
m*
This absorption usually occurs in the far-infrared spectral region, and the ef­
fective mass can be deduced from the values of A and B at resonance. In a
 10.4 Polarons 217

typical experiment, we use a fixed wavelength source from an infrared laser A particularly clear manifestation of the
and find the value of B which gives the maximum absorption. For example, electron-phonon coupling can be observed in
the cyclotron resonance occurs at about 6.1 T in GaAs (m* = 0.067mo) for cyclotron resonance experiments when B =
m*!:2Lo/e, so that We = QLO· The degen­
the 118 µ,m line from a methanol laser. The effective mass we find this way is erate electron and phonon modes anticross
the polaron mass m**, not the value determined by the curvature of the bands with each other as the field is swept through
given by eqn 10.21. this condition, and the cyclotron resonance
line splits into a doublet. The magnitude of
If the electron-phonon coupling constant aep is small, we can give an ex­
the splitting is directly proportional to the
plicit relationship between the rigid lattice effective mass m* and the polaron electron-phonon coupling constant aep• This
mass m**: effect was first observed in n-type InSb.
m** 1 1
-----�I+-aep (10.25)
m* 1 - l¥ep/6 6 ·

·Values of m* are actually worked from the measured values of m** by applying
eqn 10.25. For III-V semiconductors like GaAs with aep < 0.1, m** only dif­
fers from m* by about 1 %. The polaron effect is thus only a small correction.
This correction becomes more significant for II-VI compounds (e.g. ~ 7 %
for ZnSe). With highly ionic crystals like AgCl, the small aep approximation
is not valid. The actual polaron mass of AgCl is 0.43, which is about 50 %
larger than the rigid lattice value.
It can be shown that, in addition to the change of the mass, the polaron effect
causes a reduction in the band gap by an amount:

(10.26)

With a III-V material like GaAs, this again produces only a relatively small
effect: l::!..E g ~ -0.1 %. In practice, when we measure Eg by optical spec­
troscopy we always measure the polaron value.
Another important parameter of the polaron is its radius, rp, which spec­
ifies how far the lattice distortion extends. This is depicted schematically in
Fig. 10.4 by the dashed circle drawn around the electron that causes the lattice
distortion. If aep is small, we can give an explicit formula for rp:

(10.27)

This gives rp = 4.0 nm for GaAs and 3.1 nm for ZnSe. Both values are signifi­
cantly larger than the unit cell size ( ~ 0.5 nm), which is important because the
theory used to derive eqns 10.25-10.27 assumes that we can treat the medium
as a polarizable continuum. This approximation is only valid if the radius of the
polaron is very much greater than the unit cell size. A polaron which satisfies
this criterion is called a large polaron. In highly ionic solids such as AgCl and
the alkali halides, aep is not small and the polaron radius is comparable to the
unit cell size. In this case we have a small polaron. The mass and radius have
to be calculated from first principles.
The small polaron effect in highly ionic crystals leads to self-trapping of the
charge carriers. The local lattice distortion is very strong, and the charge carrier
can get completely trapped in its own lattice distortion. The carrier effectively
digs itself into a pit and cannot get out of it. This is particularly the case for the
holes in alkali halide crystals. The only way they can move is by hopping to Polaronic hopping effects are also important
a new site. The electrical conductivity of most alkali halide crystals is limited in the conduction processes in organic semi­
by this thermally activated hopping process at room temperature. conductors like polydiacetylene.
218 Phonons

Self-trapping effects are important in determining the energies of Frenkel

excitons. As discussed in Section 4.5, these are bound electron-hole pairs
localized at individual atom or molecule sites within the lattice. The self­
trapping of either the electron or hole can exacerbate the tendency for the
exciton to localize, thereby instigating the transition from Wannier (free) to
Frenkel exciton behaviour. The ground state excitons observed in many alkali
halide, rare gas and organic crystals are of the self-trapped Frenkel type.

10.5 Inelastic light scattering

Inelastic light scattering describes the phenomenon by which a light beam is
scattered by an optical medium and changes its frequency in the process. It
contrasts with elastic light scattering, in which the frequency of the light is
unchanged. The interaction process is illustrated in Fig. 10.9. Light incident
with angular frequency w1 and wave vector k1 is scattered by an excitation
of the medium of frequency Q and wave vector q. The scattered photon has
frequency w2 and wave vector k2. Inelastic light scattering can be mediated by
many different types of elementary excitations in a crystal, such as phonons,
We presented a Raman scattering spectrum magnons or plasmons. In this chapter we will be concerned exclusively with
from plasmons in n-type GaAs in Fig. 7.13. phonon processes.
Inelastic light scattering from phonons is generally subdivided as to whether
it is the optical or acoustic phonons that are involved:

• Raman scattering. This is inelastic light scattering from optical

phonons.
• Brillouin scattering. This is inelastic light scattering from acoustic
phonons.

The physics of the two processes is essentially the same, but the experimental
techniques differ. We will thus consider the general principles first, and then
consider the details of each technique separately.

10.5.1 General principles of inelastic light scattering

Inelastic light scattering can be subdivided into two generic types:

• Stokes scattering;
• Anti-Stokes scattering.

Stokes scattering corresponds to the emission of a phonon (or some other type
of material excitation), while anti-Stokes scattering corresponds to phonon
absorption. The interaction shown in Fig. 10.9 is thus a Stokes process. Con­
servation of energy during the interaction requires that:
Fig. 10.9 An inelastic light scattering pro- WJ = w2 ± Q , (10.28)
cess. The straight arrows represent pho-
tons, while the wiggly arrow represents the
phonon. The process shown corresponds to while conservation of momentum gives:
Stokes scattering in which the photon is
shifted to lower frequency. k1 = k2 ± q. (10.29)
 10.5 Inelastic light scattering 219

The + signs in eqns 10.28 and 10.29 correspond to phonon emission (Stokes
scattering), while the - signs correspond to phonon absorption (anti-Stokes
scattering). Thus the light is shifted down in frequency during a Stokes process,
and up in frequency in an anti-Stokes event.
Anti-Stokes scattering will only be possible if there are phonons present
in the material before the light is incident. The probability for anti-Stokes
scattering therefore decreases on lowering the temperature as the phonon pop­
ulations decrease. This means that the probability for anti-Stokes scattering
from optical phonons is very low at cryogenic temperatures. On the other hand,
Stokes scattering does not require a phonon to be present and can therefore
occur at any temperature. The full quantum mechanical treatment shows that
.the ratio of anti-Stokes to Stokes scattering events is given by:
!anti-Stokes
= exp ( -rt�G
i'a ,......
1 kB T) (10.30)
Istokes

This will be the ratio of the intensities of the anti-Stokes and Stokes lines
observed in the Raman or Brillouin spectra.
The frequencies of the phonons involved can be deduced from the frequency
shift of the scattered light using eqn 10.28. Thus the main use of inelastic
light scattering is to measure phonon frequencies. This means that inelastic
light scattering can give complementary information to that obtained from
the infrared spectra. For example, infrared reflectivity measurements tell us
nothing about the acoustic phonons, but we can measure the frequencies of
some of the acoustic modes using Brillouin scattering experiments. We will
consider this complementarity in more detail when we discuss the selection
rules for Raman scattering in subsection 10.5.2 below.
The maximum phonon frequency in a typical crystal is about 10 12 -10 13 Hz.
This is almost two orders of magnitude smaller than the frequency of a photon
in the visible spectral region. Equation 10.28 therefore tells us that the maxi­
mum frequency shift for the photon will be around 1 %. The wave vector of
the photon is directly proportional to its frequency, and we can therefore make
the approximation:
nw
lk2 I �lk 1 I = -, (10.31)
C
where n is the refractive index of the crystal and w is the angular frequency of
the incoming light.
We know from eqn 10.29 that lql = lk 1 - k2 I. The maximum possible value
of lql thus occurs for the back-scattering geometry in which the outgoing
photon is emitted in the direction back towards the source. In this case, we
have: nw
q�lk -(-k)l�2-. (10.32)
By inserting typical values into eqn 10.32, we conclude that the maximum
value of q that can be accessed in an inelastic light scattering experiment is of
order 107 m-1. This is very small compared to the size of the Brillouin zone
in a typical crystal ( ~ 10 10 m- 1 ). Inelastic light scattering is thus only able to
probe small wave vector phonons.
Raman and Brillouin scattering are generally weak processes, and we there­
fore expect that the scattering rate will be small. This is because we are dealing
with a higher order interaction than for linear interactions such as absorption.
220 Phonons

Figure 10.9 shows us that three particles are present in the Feynman diagram
for inelastic light scattering rather than the two for absorption (see Fig. 10.1).
Therefore, a higher order perturbation term must be involved. This means that
we usually have to employ very sensitive detectors to observe the signals even
when using a powerful laser beam as the excitation source.

10.5.2 Raman scattering

C. V. Raman was awarded the Nobel prize in 1930 for his discovery of inelastic
light scattering from molecules. The process which now carries his name refers
to scattering from high frequency excitations such as the vibrational modes of
molecules. In the present context of phonon physics, it refers specifically to
inelastic light scattering from optical phonons.
Optical phonons are essentially dispersionless near q = 0. We argued above
that inelastic light scattering can only probe the phonon modes with q � 0.
Therefore, Raman scattering gives little information about the dispersion of
optical phonons, and its main use is to determine the frequencies of the LO
and TO modes near the Brillouin zone centre. For example, when Raman
techniques are used to measure polariton dispersion curves (see Section 10.3,
and especially Fig. 10.7), we are only probing a very small portion of the
Brillouin zone near q = 0.
The complementarity of infrared reflectivity and inelastic light scattering
measurements become more apparent when we consider the selection rules for
deciding whether a particular optical phonon is Raman active or not. These
rules are not the same as those for determining whether the mode is IR active.
The full treatment requires the use of group theory. However, a simple rule can
be given for crystals that possess inversion symmetry. In these centrosymmet­
ric crystals, the vibrational modes must either have even or odd parity under
inversion. The odd parity modes are IR active, while the even parity modes
are Raman active. Thus the Raman active modes are not IR active, and vice
versa. This is called the rule of mutual exclusion, and is a well-known result
in molecular physics. In non-centrosymmetric crystals, some modes may be
simultaneously IR and Raman active.
As an example of these rules, we can compare silicon and GaAs. Silicon
has the diamond structure with inversion symmetry, while GaAs has the non­
centrosymmetric zinc blende structure. The TO modes of silicon are not IR
active, but they are Raman active, while the TO modes of GaAs are both Raman
and IR active.
The observation of a Raman spectrum requires specialized apparatus to
overcome the difficulties that are inherent to the technique. We pointed out
above that the signal is relatively weak, which means that we have to use an
intense source such as a laser to produce a sizeable scattering rate. However,
the frequency shift of the scattered photons is quite small. We thus need to re­
solve a weak Raman signal which is very close in wavelength to the elastically
scattered light from the laser.
Figure 10.10 shows a basic experimental arrangement that can be used to
measure Raman spectra. The sample is excited with a suitable laser, and the
scattered light is collected and focussed onto the entrance slit of a scanning
spectrometer. The number of photons emitted at a particular wavelength is
 10.5 Inelastic light scattering 221

laser beam
scanning
sample reflected
entrance slit double
in cryostat laser
spectrometer computer
Fig. 10.10 Experimental apparatus used to
record Raman spectra. The sample is excited
CCD array
with a laser, and the scattered photons are
detector collected and focussed into a spectrometer.
The signals are recorded using a sensitive
photon-counting detector such as a photo­
multiplier tube or a charge coupled device
lenses (CCD).

--1.A._ GaAs
_1_1_:nP

Ji A!Sb

_A-1
Fig. 10.11 Raman spectra for the TO and
GaP LO phonons of GaAs, InP, AlSb and GaP at
300K 300 K using a Nd: YAG laser at 1.06 µm. The
spectra are plotted against the wave number
shift: I cm-1 is equivalent to an energy shift
of 0.124 meV. The LO mode is the one at
200 300 400 higher frequency. After [4], copyright 1972
Excerpta Medica Inc., reprinted with permis­
Energy shift (cm- 1) sion.

registered using a photon-counting detector and then the results are stored on a
computer for analysis. Photomultiplier tubes have traditionally been employed
as the detector in this application, but modem arrangements now tend to use
array detectors made with charge coupled devices (CCD arrays). By orientat­
ing the sample appropriately, the reflected laser light can be arranged to miss
the collection optics. However, this still does not prevent a large number of One way to achieve good stray light rejection
elastically scattered laser photons entering the spectrometer, and this could is to use a double spectrometer, which is
essentially two spectrometers in tandem. This
potentially saturate the detector. To get around this problem, a high resolution both increases the spectral resolution and
spectrometer with good stray light rejection characteristics is used. enhances the rejection of unwanted photons.
Figure 10.11 shows the Raman spectrum obtained from four III-V crystals
at 300 K. The laser source was a Nd:YAG laser operating at 1.06 µ,m, and a
double monochrom.ator with a photomultiplier tube were used to detect the
signal. Two strong lines are observed for each crystal. These correspond to
the Stokes-shifted signals from the TO phonons and LO phonons, with the
LO phonons at the higher frequency. The values obtained from this data agree
very well with those deduced from infrared reflectivity measurements. (See
Exercise 10.13.)
222 Phonons

10.5.3 Brillouin scattering

L. Brillouin gave a theoretical discussion of the scattering of light by acoustic
waves in 1922. The technique named after him now refers to inelastic light
scattering from acoustic phonons. Its main purpose is to determine the disper­
sion of these acoustic modes.
The frequency shift of the photons in a Brillouin scattering experiment is
given by (see Exercise 10.14):
2nw 0
8w = Vs-- sin , (10.33)
c 2
where w is the angular frequency of the incident light, n is the refractive index
of the crystal, Vs is the velocity of the acoustic waves, and 0 is the angle through
which the light is scattered. Measurements of 8w therefore allow the velocity
of the sound waves to be determined if the refractive index is known.
The experimental techniques used for Brillouin scattering are more sophisti­
cated than those for Raman scattering due to the need to be able to detect much
smaller frequency shifts. Single-mode lasers must be used to ensure that the
laser linewidth is sufficiently small, and a scanning Fabry-Perot interferometer
is used instead of a grating spectrometer to obtain the required frequency
resolution.

Example 10.2
When light from an argon ion laser operating at 514.5 nm is scattered by
optical phonons in a sample of AlAs, two peaks are observed at 524.2 nm and
525.4 nm. What are the values of the TO phonon and LO phonon energies?

Solution
We can work out the energies of the phonons by using eqn 10.28. The photons
have been red-shifted, and thus we are dealing with a Stokes process. For the
524.2 nm line we therefore have:

For the 525.4 nm line we find Q = 7.6x 10 13 Hz. The higher frequency phonon
is the LO mode. Hence we find Ii r.lm = 45 meV and Ii r.lw = 50 meV.

10.6 Phonon lifetimes

The discussion of the phonon modes as classical oscillators in Section 10.2 led
us to introduce a phenomenological damping constant y. This damping term is
needed to explain why the reflectivity in the restrahlen band is less than unity.
Analysis of the experimental data led us to conclude that y is typically in the
range 10 1 1 -10 12 s- 1 . This very rapid damping is a consequence of the finite
 10.6 Phonon lifetimes 223

lifetime r of the optical phonons. Since y is equal to r-1, the data implies that
r is in the range 1-10 ps.
The very short lifetime of the optical phonons is caused by anharmonicity (a)
in the crystal. Phonon modes are solutions of the equations of motion with the
assumption that the vibrating atoms are bound in a harmonic potential well. In
reality, this is only an approximation that is valid for small displacements. In
general, the atoms sit in a potential well of the form:

(10.34)

An example of how interatomic interactions lead to a potential of this form is

considered in Exercise 10.15.
The term in x 2 in eqn 10.34 is the harmonic term. This leads to simple
harmonic oscillator equations of motion with a restoring force -dU /dx pro­
(b)
portional to -x. The terms in x 3 and higher are the enharmonic terms. These
anharmonic terms allow phonon-phonon scattering processes. For example,
the term in x 3 allows interactions involving three phonons. Figure 10.12 illus­
trates two possible permutations for a three-phonon process. Fig. 10.12 Three phonon interaction pro­
Figure 10.12(a) shows a three-phonon interaction in which one phonon is cesses. Each wiggly arrow represents a
annihilated and two new phonons are created. This type of anharmonic in­ phonon. These processes are caused by an­
harmonicity in the crystal.
teraction is responsible for the fast decay of the optical phonons. We can
see why this is so by referring to the generic phonon dispersion curve for
the first Brillouin zone shown in Fig. 10.13. Lattice absorption or Raman
scattering creates optical phonons with q � 0. Three-phonon processes allow
these phonons to decay into two acoustic phonons as indicated in Fig. 10.13.
Momentum and energy can be conserved if the two acoustic phonons have
opposite wave vectors, and their frequency is half that of the optical phonon.
With more complex dispersion relationships, and also the possibility for higher
order processes, many other types of decay can contribute to the short lifetime
of the optical phonons.
The lifetime of the optical phonons can be deduced from Raman data in
two different ways. Firstly, the spectral width of the Raman line is affected by
lifetime broadening. Provided that other sources of broadening are smaller, the
linewidth in frequency units is expected to be (2Jrr)-1. Thus measurements of

QTO
2

0 Fig. 10.13 Decay of an optical phonon into

7t 0 7t two acoustic phonons by a three-phonon in­
a Wave vector (q) a teraction of the type shown in Fig. !0.12(a).
224 Phonons

the linewidth give a value for r independently of the reflectivity data. Secondly,
r can be measured directly by time-resolved Raman spectroscopy using short
pulse lasers. The lifetime of the LO phonons in GaAs has been determined in
this way to be 7 ps at 77 K. This value agrees with the linewidth measured in
the conventional Raman spectrum. It is also similar to the lifetime of the TO
phonons deduced from reflectivity measurements.

Chapter summary

• The TO phonon modes of polar solids couple strongly to photons

when their frequencies and wave vectors match. Acoustic phonons
and LO phonons do not couple directly to light waves.
• The interaction between the light and the TO phonon can be
modelled by using the classical oscillator model. This model
explains why the reflectivity of a polar solid is very high for
frequencies in the restrahlen band between vro and VLO•
• The reflectivity in the restrahlen band is 100 % for an undamped
system, but damping due to the finite phonon lifetime reduces the
reflectivity in real crystals.
• The frequencies of the TO and LO phonon modes are rela
to each other by the Lyddane-Sachs-Teller relationship given in
eqn 10.15.
• The lattice absorbs stron at the TO phonon frequency. The
absorption can in thin film samples.
• The strongly co n waves at frequencies near
the restrablen lariton modes.
lar crystals leads to polaron
surrounded by a local lattice
the electron or hole increases
are strong in ionic crystals like the alkali

e inelastic light scattering pro-­

honons respectively. Energy and
g process.
ring processes cor­
respectively. Anti-
very improbable at low

• e possibility of decay
interactions.

Further reading
Introductory reading on phonons may be found in practically any solid state
physics text, for example: Ashcroft and Mermin (1976), Bums (1985), Ibach
and Luth (1995) or Kittel (1996).
 Exercises 225

The theory of polaritons and polarons is described in more detail in

Madelung (1978). Pidgeon (1980) and Seeger (1997) discuss cyclotron
resonance experiments in detail. The properties of self-trapped excitons are
covered by Song and Williams (1993), while Pope and Swenberg (1999)
discuss polaronic hopping transport, especially in organic semiconductors.
A classic text on the infrared physics of molecules and solids is Houghton
and Smith (1966). The techniques of inelastic light scattering are described in
detail by Mooradian (1972) or Yu and Cardona (1996). The study of phonon
dynamics by ultra-fast laser techniques is described by Shah (1999).

References
[1] Madelung, 0. (1996). Semiconductors, basic data, (2nd edn). Springer­
Verlag, Berlin.
[2] Hass, M. (1967). In Semiconductors and Semimetals, Vol. 3: Optical prop­
erties of 111-V compounds (ed. R.K. Willardson and A.C. Beer). Aca­
demic. Press, New York, p. 3.
[3] Henry, C.H. and Hopfield, J.J. (1965). Phys. Rev. Lett., 15, 964.
[4] Mooradian, A. (1972). In Laser Handbook, Vol. II ( ed. F.T. Arecchi and
E.O. Schulz-duBois). North Holland, Amsterdam, p. 1409.
[SJ Turner, W.J. and Reese, W.E. (1962). Phys. Rev., 127, 126.

Exercises
(10.1) State, with reasons, which of the following solids would (i) the frequencies of the TO and LO phonons of
be expected to show strong infrared absorption: (a) ice, AlSb near the Brillouin zone centre;
(b) germanium, (c) solid argon at 4 K, (d) ZnSe, (e) SiC. (ii) the static and high frequency dielectric constants,
(10.2) Show that the reflectivity of an undamped polar solid Est and E oo ;
falls to zero at a frequency given by (iii) the lifetime of the TO phonons.

v
= ( Est - l ) ½ \J'J'O, Are the experimental values found in parts (i) and (ii)
consistent with the Lyddane-Sachs-Teller relationship?
E oo - l
1.0 .--..--�...----r-"'T'"""--r---,--r--r-"""T""--r----.
where Est and E oo are the low and high frequency
dielectric constants, and \J'J'O is the frequency of the TO
0.8
phonon mode at the Brillouin zone centre. AlSb
(10.3) The static and high frequency dielectric constants of
·;::
·.:: 0.6
LiF are Est = 8.9 and E oo = 1.9 respectively, and <l)

the TO phonon frequency \J'J'O is 9.2 THz. Calculate the <l) 0.4
upper and lower wavelengths of the restrahlen band.
0.2
(10.4) Estimate the reflectivity in the middle of the restrahlen
band for a crystal with \J'J'O = 10 THz, Est = 12.1, and 0.0
Eoo = 10, when the damping constant y is (a) 1011 s- 1 16 20 24 28 32 36 40
and (b) 1012 s-1 . Wavelength (µm)
(10.5) Figure 10.14 shows the measured infrared reflectivity of Fig. 10.14 Infrared reflectivity of A!Sb. After [5], copyright
AlSb crystals. Use this data to estimate: 1 962 American Institute of Physics, reprin ted with permission.
226 Phonons

(10.6) Estimate the absorption coefficient at the TO phonon angular frequency Q. By considering the conservation
frequency in a typical polar solid with a damping of momentum in the process, show that Q is given by:
constant y of (a) 1011 s-1 and (b) 1012 s-1 . Take
2nw 0
V'fO = 10 THz, Est = 12.1, and E oo = 10. Q = Vs-- sin -
C 2,
(10.7) Explain qualitatively why the reflectivity of NaCl in the
middle of the restrahlen band is observed to decrease where Vs and n are the velocity of sound and the
from 98 % at 100 K to 90 % at 300 K. refractive index in the medium respectively. (You may
assume that w » Q.) Hence justify eqn 10.33.
(10.8) The static and high frequency dielectric constants of InP
are Est = 12.5 and Eoo = 9.6 respectively, and the TO A Brillouin scattering experiment is carried out on a
phonon frequency VTO is 9.2THz. Calculate the wave crystal with a refractive index of 3 using light from
vector of a polariton mode with a frequency of 8 THz. a laser with a wavelength of 488 nm. The scattered
(Ignore phonon damping.) photons are found to be down-shifted in frequency by
10 GHz when observed in the back-scattering geometry
(10.9) In an infrared absorption experiment on n-type CdTe, with 0 = 180 ° . Calculate the speed of sound in the
the cyclotron resonance condition is satisfied at 3.4 T crystal.
for the 306 µm line from a deuterated methanol laser.
Calculate (i) the polaron mass, and (ii) the rigid lattice (10-15)* The potential energy per molecule of an ionic crystal
electron effective mass, given that Eoo = 7.1, Est = with a nearest neighbour separation of r may be approx­
10.2, and QLO = 31.9THz. imated by the following form:
(10.10) Discuss the qualitative differences you would expect fJ ae 2
between the Raman spectrum observed from diamond U(r)= - - -­
r l2 4Juor'
to that shown for the III-V crystals in Fig. 10.11.
where a is the Madelung constant of the crystal, and fJ
(10.11) In an inelastic light scattering experiment on silicon
is a fitting parameter.
using an argon ion laser at 514.5 nm, Raman peaks
are observed at 501.2 nm and 528.6 nm. Account for (i) Account for the functional form of U (r ).
the origin of the two peaks, and estimate their intensity (ii) Show that U(r) has a minimum value when r =
ratios if the sample temperature is 300 K. ro, where rJ1 = 48/37rEo/ae2 .
(10.12) NaCl is a centrosymmetric crystal. Would you expect (iii) Expand U(r) as a Taylor series about ro, and
the TO phonon modes to be IR active, or Raman active, hence show that the potential takes the form given
or both? by eqn 10.34 for small displacements about ro,
(10.13) Use the data in Fig. 10.11 to deduce the energies in meV stating the value of the constant C3 in terms of a
of the TO and LO phonons of GaAs, InP, A!Sb, and GaP and r0.
at 300 K. How do the values for GaAs obtained from (10.16) High resolution Raman experiments on a GaAs crystal
this data relate to the infrared reflectivity data given in indicate that the LO phonon line has a spectral width
Fig. 10.5 ? of 0.85 cm-1. Use this value to estimate the lifetime of
(10.14) A photon of angular frequency w is scattered inelas­ the LO phonons, on the assumption that the spectrum is
tically through an angle 0 by an acoustic phonon of lifetime-broadened.

* Exercises marked with an asterisk are more challenging.