Chapter 6

Classical Models for the

Dielectric Functions
The dielectric function () is determined by the properties of the matter and
its interaction with a light wave propagating through that matter. A physically
correct treatment of this problem requires a quantum mechanical treatment of
the properties of solids and their interaction with light which is quite complicated. The basic features, however, can be seen already on the basis of a classical
treatment as it has been developed within the frame of the electromagnetic theory of light propagation and electron theory by Lorentz, Drude and others. This
classical theory is frequently used instead of more sophisticated quantum mechanical theories to interpret the results of laboratory measurements of optical
constants of materials of practical interest. The basic concepts of the classical
approach are described in this chapter.

6.1

Lorentz model for insulators

The optical properties of insulators first have been modelled by Lorentz (1906)
by assuming that the electrons and ions in the matter form a system of harmonic
oscillators that are excited to forced oscillations by a light wave propagating
through the matter.
The basic assumption is that the ions and electrons forming a solid are located at fixed equilibrium positions in the solid determined by internal forces,
if no external forces are applied. The electric field of a light wave propagating
through the solid exerts forces on the charged particles which displaces them
from their equilibrium positions, while the interaction with the surrounding
atoms results in forces which tend to drive a displaced particle back into its
equilibrium position. It is assumed that this force is proportional to the displacement x. The particles then can perform harmonic oscillations around their
equilibrium positions with some frequency 0 . On the other hand, the interaction of an oscillator with other oscillators in its environment may transfer part
129

130

Chapter 6. Classical Models for the Dielectric Functions

of its vibrational energy to other oscillators. This results in a damping of the

oscillations.
In a first step the oscillators are assumed to be isotropic such that the properties of the oscillations do not depend on the direction of the displacement.
This assumption is a strong simplification since most solids are crystals with a
non-isotropic structure. This simplification is later relaxed, but first we consider
the most simple case because its shows already most of the basic features of the
problem.

6.1.1

Equation of motion

The basic assumption of the Lorentz model then is that a light wave propagating
through the medium excites the system of identical damped harmonic oscillators
in the matter. The displacements of the charged particles from their equilibrium
positions result in a dielectric polarisation of the matter, which in turn modifies
the external electric field applied to the matter and this modifies the mode of
propagation of the light wave in the material. The basic problem is to calculate
the mutual interaction between the system of oscillators and the fields of the
light wave.
The differential equations for the displacement x of the charged particles
from their equilibrium positions due to the action of a local electric field E loc
are
m
x + m x + m02 x = eE loc .
(6.1)
m and e are the mass and electric charge, respectively, of the particles considered, 0 is the frequency of the harmonic oscillations and is the damping
constant of these oscillation. E loc is the local electric field at the location of the
oscillator, which is different from the macroscopic field E, which is an average
over a region which contains a big number of oscillators.
The solution of equation (6.1) is obtained by the standard method of introducing the Fourier decomposition of the time dependent quantities E loc (t) and
x(t) into equation (6.1). The result for the spectral components x() of the
displacements is
x() =

02

e
m
2

E loc () .

(6.2)

If the damping constant is different from zero, then the factor of proportionality between x() and E loc () is complex.

6.1.2

Polarization

The polarisation of the matter induced by the oscillations of the particles in the
solid is
X
(xi ) ,
(6.3)
P =
all dipoles i
per unit volume

where i are the individual dipole moments associated with the oscillators.
These are expanded in terms of the small displacements from the equilibrium

6.1. Lorentz model for insulators

131

position
P =

X
i

(0) +

(xi ) (x) +

all dipoles i
per unit volume

(6.4)

The first term describes a static polarisation in the matter resulting from permanent dipoles. This term is neglected because such materials are presently
not considered in astromineralogy. The second term describes the polarisation
associated with the displacements of the oscillators from their equilibrium positions.
For the dielectric displacement D = E + 4P one obtains by Fourier transformation
X
(xi ) x .
(6.5)
D() = E() + 4
all dipoles i
per unit volume

Since it is assumed that the oscillators are isotropic the induced dipole moments
of the individual oscillators are i = ei xi , where ei are the charges associated
with the individual oscillators. Hence
i
= ei
x
and
X

D() = E() + 4

xi ()ei .

(6.6)

all dipoles i
per unit volume

In a solid one has many different vibrational modes. For each such a vibrational mode there exists a huge number of oscillators per unit volume corresponding to the mode. Denote by an index j the various types of oscillators,
and by an index i the individual oscillators of a given type. Then one writes
X
X
xj,i ()ej,i .
(6.7)
D() = E() + 4
j

all dipoles i of type j

per unit volume

If all oscillators of type j have the same properties (i.e., resonant frequencies,
damping constants, associated charges), the summation over i simply yields the
number density nosc,j of oscillators of type j. Then
D() = E() + 4

nosc,j xj ()ej .

(6.8)

Since according to the causality principle there exists a dielectric function ()

with the property D() = ()E() (see section 5.4.1), comparison with the
preceding equation shows the dielectric function to be given by
() = 1 + 4

X
j

nosc,j ej

xj ()
,
E()

(6.9)

132

Chapter 6. Classical Models for the Dielectric Functions

where E is the field strength of the macroscopic electric field. The displacement
x() is given by equation (6.2). The dielectric function of matter, the properties
of which are modelled by damped oscillators interacting with the electric field
of a light wave, is therefore given by
() = 1 + 4

X
j

6.1.3

nosc,j ej

ej Ej,loc ()
1
2 2 i .
mj E() 0j
j

(6.10)

Local field

The calculation of the local field E loc at the position of a particular oscillator in
terms of the macroscopic field E is quite complex for most crystal symmetries of
the solid. In the following only the case of isotropic matter or cubic symmetry
is considered.
In order to calculate the local field E loc at some point in the matter one
considers within the matter some sphere with radius R around that point. The
contribution of the dipole moments associated with the oscillators outside the
sphere to the local field at the centre is treated in a continuum approximation,
the contributions of the dipoles associated with the oscillators inside the sphere
to the local field are individually summed up.
The polarisation of the matter due to the oscillators outside the sphere corresponds to some electrical surface charge density at the surface of the sphere
,
= P n

(6.11)

is the unit surface normal vector and P the polarisation in the matter.
where n
At the centre the electric charge of a particular surface element of area f
generates an electric field
E f =

P n
f .
n
2
R

Summation over all surface elements yields

I
Z

P n
4
4

E loc,outside = df n
= P e d n
P e ,
n
= P e =
2
R
3
3
where e is a unit vector in the direction of the -axis of a Cartesian coordinate
system.
The local electric field from the matter outside the sphere now is
E loc,outside =

4
P.
3

(6.12)

This does not depend on the choice of R.

The contribution of the dipoles inside the sphere to the local field at the
centre vanishes for certain lattice symmetries, for instance cubic symmetry. It

6.1. Lorentz model for insulators

133

is here assumed that this contribution can be neglected. The local field at some
point within the matter then is
E loc = E +

4
P.
3

(6.13)

On the other hand one has

P =

DE
1
=
E
4
4

(6.14)

and then
E loc = E +

6.1.4

1
+2
E=
E.
3
3

(6.15)

Dielectric function of the Lorentz oscillator model

Inserting the expression for the local field strength into equation (6.10) yields
1
4 X nosc,j e2j
() 1
=
2 2 i .
() + 2
3 j
mj
0j
j

(6.16)

This is the well known Clausius-Mosotti-equation for the dielectric function. If

one writes this as
() 1
=Z
() + 2
and solves for Z one finds
=1+

1
Z

3
1

Thus one finally has

() = 1 + 4

X nosc,j e2j
j

mj

with
2
2

0j
= 0j

0j

1
2 ij

4 nosc,j e2j
.
3
mj

(6.17)

(6.18)

This is the same result as equation (6.10) if one would have Eloc = E. The
influence of the environment on an individual oscillator thus merely results in
some shift of the resonance frequency of the free oscillator. Since the oscillators
of the present model do not exist as free oscillators isolated from the environment
of the solid, the frequency 0 has no particular meaning within the frame of the
present model and in the following the frequency
0 is renamed as 0 since only

134

Chapter 6. Classical Models for the Dielectric Functions

this quantity can be determined from laboratory measurements. Further one

introduces the following abbreviation
2
pj
=

4 nosc,j e2j
3
mj

and splits equation (6.17) into real and imaginary part. Then


2
2
X pj
0j
2
r () = 1 +


2 2 2 + 22
0j
j
j
i () =

X
j

2
0j

2
pj
j
.

2
2
+ j2 2

(6.19)

(6.20)
(6.21)

These are the final equations for the dielectric function of the Lorentz model of
damped oscillators. These equations are frequently used to describe the optical
properties of insulators.
Given these equations, one can calculate from equations (??) and (??) the
real and imaginary part of the complex index of refraction. For transparent
media with i r one has for instance approximately
nr =

r ,

ni =

1 i
2 nr

(6.22)

and from equation (??) one obtains for the absorption coefficient of the transparent medium
2
X pj
j2 2
.
(6.23)
=


j c n r 2 2 2 + 2 2
j
j
0j
The second factor in this expression can be written in terms of a normalised
frequency x and a normalised damping constant as
l(x) =

2 x2
,
[1 x2 )2 + 2 x2

x=

j 
.
, =
0j
0j

The variation of this profile with normalised frequency x for different values
of the ratio /0 is shown in figure 6.1. For strong damping ( > 0.1 0 ) the
profile is wide and asymmetric, for weak damping ( 0 ) the profile is narrow
and nearly symmetric.

6.1.5

Lorentz profile

In most applications one has to deal with the case of weak damping 0
where the width of the profile l(x) is small. Often then also the oscillators
are non-overlapping in the sense that the frequency difference of neighbouring
resonant frequencies 0j on the -axis is much bigger than the width of the
profile l(x). Let = 0 . Then one has 02 2 = (2 + ) =

6.1. Lorentz model for insulators

135

1
/ 0 = 1

/ 0 = 0.1
/ 0 = 0.01
/ 0

Figure 6.1. Normalised absorption profile of the Lorentz damped oscillator model

Figure 6.2. Lorentz profile

2 + ()2 . For a narrow profile one usually has only to consider such
values of which satisfy . Then one can neglect quadratic terms in
and has 02 2 2. With this approximation one obtains for the real
and imaginary part of the dielectric function of a single oscillator
p2
0
20 (0 )2 +

r () =

1+

i () =

p2

40 (0 )2 +

2
4

2
4

(6.24)
(6.25)

This describes the variation of the dielectric function in the vicinity of an absorption line and is sometimes preferred over the more general expressions (6.20)
and (6.21) to describe relatively small absorption lines. If there is more than
one kind of oscillators, one clearly has to sum over the contributions from all
oscillators.

136

Chapter 6. Classical Models for the Dielectric Functions

The integral of i over an absorption line is proportional to

Z +
Z +

dw
d
=
2
2
2
2
(0 ) + 4

w + 1
Here the integration is extended over the whole real -axis, since for a narrow
line the integral over the negative part of the -axis gives an negligible contribution. Further the new variable w = 2( 0 )/ has been introduced. Next
one transforms to a new variable by my means of w = tan . Then
Z
Hence one has
Z

dw
=
w2 + 1

+
2

d = .

(0 )2 +

2
4

= 2 .

(6.26)

(6.27)

Define a profile function

() =

2 ()2 +

2
4

This profile is called the Lorentz-profile or dispersion profile. It is normalised as

follows
Z +
d () = 1 .
(6.28)

The profile is shown in figure 6.2. The half width of the profile is /2.
In terms of the Lorentz profile one can write the absorption coefficient in the
region of a narrow absorption line of a transparent medium, equation (6.23), as
=
6.1.5.1

p2
() .
2 c nr

(6.29)

Limit cases

Consider first the limit case of frequencies large compared to the resonance
frequency 0 of a single oscillator. From equations (6.20) and (6.21) one obtains
for 0 and a single oscillator
r () =

p2
2

(6.30)

i () =

p2
3

(6.31)

The real part of () approaches unity from below and the imaginary part
vanishes as 1/ 3 . For the complex index of refraction it follows from equations

6.1. Lorentz model for insulators

137

(6.22) for transparent media and a single oscillator

nr

= 1

p2
2 2

(6.32)

p2
.
(6.33)
2 3
On the high frequency side well outside the central peak of the profile the
contribution of an oscillator to the dielectric function or the complex index of
refraction vanishes. For very high frequencies, above all resonant frequencies
0j , every material becomes transparent.
In the opposite limit 0 one obtains from equations (6.20) and (6.21)
for a single oscillator correct up to second order in
ni

p2
p2 (02 2 ) 2
+

02
06

r () =

1+

i () =

p2
.
04

(6.34)
(6.35)

In the limit 0 the real part of () approaches a constant value > 1 and the
imaginary part vanishes proportional to . For the complex index of refraction
of a single oscillator one has
p2
p2 (02 2 ) 2
+

2
20
206
p2
.
202 (02 + p2 )

nr

= 1+

(6.36)

ni

(6.37)

On the low frequency side, well outside the central peak of the profile, the
contribution of an oscillator to the imaginary part of the dielectric function or
the complex index of refraction vanishes proportional to . There is, however,
a non-vanishing contribution to the real part of the dielectric function or index
of refraction.
If there exists more than one oscillator and if the oscillators are well separated, the dielectric function or the complex index of refraction for the matter
at a certain frequency therefore is determined by the contribution of the oscillator with the closest lying resonant frequency 0 and by the low frequency tails
of all other oscillators having resonant frequencies on the high frequency side of
the -axis. The contribution of all other oscillators having resonant frequencies
on the low frequency side of the -axis can be neglected.
For frequencies well below the resonant frequencies of all oscillators one
obtains
2
X pj
(6.38)
r () = 1 +
2
0j
j
i () =

2
X j pj
j

4
0j

(6.39)

138

Chapter 6. Classical Models for the Dielectric Functions

In the limit 0 the real part of approaches the static dielectricity constant
0 of the material. One therefore has the following sum rule for all oscillators
0 = 1 +

2
X pj
j

2
0j

(6.40)

For the absorption coefficient one obtains from equations (??) and (6.22)
for transparent media in the low frequency limit
=

2
2ni
i
4c X j pj
1
.
=
=
4
c
nr c
0 j 0j 2

(6.41)

This approximation predicts a dependence of on wavelength proportional

to 2 at very long wavelengths. It describes the absorption behaviour of many
crystalline solids in the far IR and in the sub-mm wavelength very well, since
they have all their vibrational modes in the mid-IR wavelength region. Deviations from this behaviour are however observed, in particular for amorphous
materials, where often varies as p with p < 2.
6.1.5.2

Approximation for the optical and IR region

All solid materials have a number of strong oscillators for high frequencies associated with electronic transitions in the material. Their resonant frequencies
are usually located in the far UV to X-ray wavelength region. Insulating materials have additionally strong oscillators with resonant frequencies in the far IR
wavelength region, associated with vibrational modes in the lattice or with certain atomic groups in the matter. In the intermediate region from the near IR
to the optical wavelength region most insulating materials are transparent and
have no or only weak oscillators with resonant frequencies in this region. The
dielectric function of such materials in the optical to far IR wavelength region
may be approximated by replacing the oscillators with resonant frequencies in
the UV and higher frequency region by their low frequency limit and retaining
all other oscillators as they are. The real part of the dielectric function then is


2
2
2
X pj
X pj
0j
2
+
(6.42)
r () = 1 +

2
02
2 2 + 22
j
j
0j

where the first sum runs over all oscillators with high resonant frequencies and
the second over all oscillators with resonant frequencies in the IR. In the optical
region this simplifies to the constant value
r = 1 +

2
X pj
j

02

= n2 = .

This corresponds to the value of the real part of the dielectric function
or to the square of the real part of the index of refraction n2 in the transparent

6.2. Brendel model for amorphous material

139

optical region. Hence one may approximate for insulators the real part of the
dielectric function in the optical to IR region by


2
2
X pj
0j
2
.
(6.43)
r () = +


2 2 2 + 22
0j
j
j
For the imaginary part of the dielectric function one may write
i () =

2
X
pj
j

+


2 2 2 + 22
i
0j
j
j

(6.44)

with some constant i which is related in principle to the high frequency oscillators in the UV and higher frequencies by
i1 =

2
X j pj
j

4
0j

(6.45)

The summation here is over all high frequency oscillators.

Equations (6.43) and (6.44) are frequently used to describe the optical behaviour of insulators in the optical and infrared wavelength region. The constants entering into the expressions for r () and i () are determined by fitting
results of laboratory measurements of or n to these analytical expressions.

6.2

Brendel model for amorphous material

For amorphous materials the usually regular geometric arrangement of atoms

and atomic groups in the lattice of a solid is distorted and the type of mutual
bonding of the atoms may also be changed. If the type of bonding of the
atoms for most atoms is the same as in the crystal and if at least the relative
arrangement of the atoms remains similar to that in the crystal state one expects
the number and type of the lattice vibrations that contribute to the dielectric
function remains essentially unchanged, except that the resonance frequencies
0 and possibly the damping constants change because bond lengths and bond
angles are modified.
While in a crystalline state there exists for each type of oscillator a very
big numbers of individual oscillators of a given type with identical resonance
frequencies, one expects that in the corresponding amorphous material the resonance frequencies of the individual oscillators are now spread over a certain
range of values around that of the corresponding crystalline material. A simple model for this has been proposed by Brendel (1992). The model accounts
for the amorphous structure of a material by assuming that the different lattice vibrational modes in the infrared can be represented by Lorentz-oscillators
with randomly shifted resonance frequencies that are distributed according to
a Gaussian probability distribution.

140

Chapter 6. Classical Models for the Dielectric Functions

The dielectric function in the infrared spectral region where it is determined

by the lattice vibrations then is assumed in this model to be given by

() = +

N
X
j=1

1
p
2j

dz e(z0,j )

/2j2

2
p,j
.
z 2 2 ij

(6.46)

It consists of a dielectric background accounting for the contribution of

modes with resonance frequencies at much higher frequencies and N distributions of Lorentz oscillators with resonance frequencies 0,j , damping constants
j , and plasma frequencies p,j with standard deviations j of the Gaussian
probability distributions.
In practical applications this model is fitted to measured data. It is then
often not possible to determine the damping constants j with any precision
because of the large widths j of the Gaussian probability distribution. Then it
seems to be a good compromise to fix j at some fixed value, e.g. j = 5 cm1 ,
and to fit only the remaining parameters.

6.3

Drude-model for conductors

Electric conductors have quite different optical properties than insulators because the electrons in the solid may move freely under the action of an applied
external field. A very simple model for the dielectric properties of conductors
has been developed by Drude (1900). This model is frequently used to represent
the optical properties of conductors.

6.3.1

Conductivity

The first basic assumptions of the model of Drude is that in a conductor there
exist n0 electrons per unit volume which may move freely under the action of an
external field. The positively charged particles in the material are assumed to
be immobile. The second basic assumption is that the electrons are subject to a
drag force due to scattering processes if they move through the material. This
drag force is assumed to be given by F drag = (1/ )v with some characteristic
time constant . The equation of motion of the electrons then is
1
mv = v + eE ,

(6.47)

where m is the mass and e is the charge of the mobile charge carriers. Experience
shows that this simple approximation often (but not always) yields quite a good
representation of the properties of conductors.
In a stationary state (i.e., v = 0) one has
v=

e
E.
m

6.3. Drude-model for conductors

141

The electric current corresponding to a motion of the charge carriers with this
velocity is
n0 e 2
E.
J = en0 v =
m
Comparison with Ohms law, equation (??), yields for the electric conductivity
n0 e 2
.
m

(6.48)

This corresponds to the d.c. conductivity of the material. Since the scattering
processes which determine are strongly temperature dependent, depends
significantly on temperature.
For the case of time dependent fields one performs a Fourier decomposition
of v(t) and E(t) and obtains by standard methods from the equation of motion
the following relation between the spectral components of v() and E()
v() =

e

E() .
1
m
i

(6.49)

A comparison with Ohms law suggests to introduce a quantity

() =

0
,
1 i

(6.50)

where 0 equals the d.c. conductivity given by equation (6.48). The complex
quantity () is denoted as optical conductivity. In the limit 0 it approaches
the d.c. conductivity, for 6= 0 it is complex valued.

6.3.2

Dielectric function for conductors

A comparison with equation (??) shows that within the frame of the Drude
model the dielectric function () of electric conductors is
() = 1 +

i p2
,
1 i

(6.51)

4n0 e2
m

(6.52)

where
p2 =

is the plasma frequency, as it is defined in elementary plasma physics. The real

and imaginary parts of the dielectric function are
p2
2 + 2
p2
.
( 2 + 2 )

= 1

(6.53)

(6.54)

142

Chapter 6. Classical Models for the Dielectric Functions

The quantity is defined as

=

6.3.3

1
.

(6.55)

Drude-Lorentz model

These results for r and i become identical to the results of the Lorentz oscillator
model, equations (6.20) and (6.21), if there is only a single oscillator and the
resonance frequency equals 0 = 0. One can therefore extend the Lorentz model
to a Lorentz-Drude model by allowing one or more of the resonance frequencies
in equations (6.20) and (6.21) to equal zero. Such a model describes materials
which show some degree of electric conductivity and at the same time there
exist oscillators which may be excited by an external field. Also for fitting
the extinction properties of metals in the UV-region the pure Drude model of
free electrons usually fails and one requires additionally some strongly damped
oscillators to describe the contribution of the bound electrons.