NONLINEAR OPTICS I

Martti Kauranen
Department of Physics
Tampere University of Technology

2
1. Introduction
1.1 Origin of the nonlinear optical response
We can define nonlinear optics as the field of study where the response of a material to an op-
tical field deviates from linear. In this respect, nonlinear optics is similar to many other fields,
where the response of a system to an external perturbation becomes nonlinear when the per-
turbation is strong enough. For example, when a spring is stretched (Fig. 1.1), the displace-
ment x and applied force F are related by
+ + =
2
2 1
x k x k F (1.1)
When the force is small, only the term x k
1
is important. However, it is clear that when the
force is sufficiently increased, the displacement cannot be increased in the same proportion.
Hence, we can expect that the higher-order
terms become important when the displace-
ment is large.
A model of the type of Fig. 1.1 with an
electron attached to the end of a spring can be
used to predict nonlinear optical properties
from a purely classical model.
Most materials respond strongest to the electric component of the electromagnetic field.
This response is described by the material polarization, i.e., dipole moment per unit volume.
When two equal and opposite charges of magnitude q are displaced by vector d, the dipole
moment is
q = p d . (1.2)
If the number density (number per unit volume) of such dipoles is N the polarization of the
medium is then
N = P p . (1.3)
Linear optical response
The linear optical response of a material is described by its linear susceptibility. We assume
for now that vectorial quantities, such as the electric field and polarization, can be approxi-
mated as scalars. Within this approximation, the relation between the polarization and the
electric field is
( ) 1
0
P E c _ = , (1.4)
where
0
c is the permittivity (dielectric constant) of vacuum and
( ) 1
_ is the linear susceptibil-
x
F = k
1
x
F = k
1
x+k
2
x
2
+...

Figure 1.1. Nonlinear response of a spring to an
applied force.
3
ity. When the electric field has harmonic time dependence, the polarization will oscillate at the
same frequency. Oscillating polarization acts as a source of radiation, which will modify the
optical field. This effect is described by the index of refraction of the material, whose real part
changes the phase of the optical field and the imaginary part leads to absorption.

Nonlinear optical response
The nonlinear optical response is described by including higher powers of the electric field in
the polarization
(1) (2) 2 (3) 3
0
P E E E c _ _ _
(
= + + +

, (1.5)
where
( ) n
_ represent material susceptibilities of different orders and form the basis for tradi-
tional nonlinear optics. Usually, the second- (n=2) and third- (n=3) order effects are by far the
most important, as they can lead to qualitatively different types of effects.
This course will focus on effects that can be treated using the susceptibility formalism
as described by Eq. (1.5). However, not all nonlinear effects can be described using this for-
malism. For example, when the optical field is strongly resonant with the material, the effec-
tive susceptibility is sometimes of the form
s
I I +
=
1
0
_
_ , (1.6)
where I is the intensity of the optical field and
s
I is a quantity characteristic to the material,
known as saturation intensity. It is clear that when the field intensity exceeds the saturation
intensity, the power expansion of the type of Eq. (1.5) does not converge. Such saturation ef-
fects are not covered in these lectures. Other important topics not covered are: stimulated scat-
tering processes (e.g., Brillouin, Raman, and Rayleigh scattering), acousto-optic effects, and
photorefractive effects.
The course is based on SI units. However, a significant fraction of the publications on
nonlinear optics use the Gaussian units, including the earlier versions of the textbook of Boyd.
The conversion between different unit systems can be cumbersome and requires care.
Notationally, this course follows closely the latest edition of the textbook of Boyd. This
book is quite accessible to a researcher entering the field of nonlinear optics. Nevertheless, it
also covers topics that are quite advanced.
1.2 Complex notation
In order to treat nonlinear effects and optical effects in general, we will use the complex nota-
tion. Of course, all physical quantities are real, but their mathematical description can often be
greatly facilitated by using the complex notation.
A real time-harmonic electric field can be written as
4
( ) ( ) ( )
( ) ( )
0 0
0
1 1
2 2
cos
i i i t i t
E t E t
E e e E e e
e e
e e
e e

' = +
' ' = +

, (1.7)
where the tilde ~ denotes a quantity that oscillates rapidly in time and ( ) E e ' is the real-
valued amplitude of the field at frequency e. From here, we define the complex amplitude of
the field at this frequency and its complex conjugate as
( ) ( ) ( ) ( )
0 0
1 1
2 2
,
i i
E E e E E e

e e e e
-
' ' = = , (1.8)
so that the original real-valued field is
( ) ( ) ( )
i t i t
E t E e E e
e e
e e
-
= +

. (1.9)
The factors of 1/2 are thus included in the definition of the complex amplitude. This is not a
universal convention, and the factors are sometimes explicitly shown. The polarization at fre-
quency e is expressed analogously as
( ) ( ) ( )
i t i t
P t P e P e
e e
e e
-
= +

. (1.10)
Note that ( ) ( ) E E e e
-
= which allows us to simplify the notation even further if the
oscillating frequency is allowed to be either positive or negative. The complex field at a given
frequency can often include components that propagate in different spatial directions, e.g.,
( ) ( ) ( )
1 2
1 2
i i
E A e A e e

= +
k r k r
r r . (1.11)
1.3 Examples of second-order effects
The second-order effects arise from the square of the field of the form of Eq. (1.9). Dropping
the frequency arguments to simplify the notation, the field is taken to be
( )
i t i t
E t Ee E e
e e -
= +

, (1.12)
and its square is
( )
( )
2
2 2 2 2
2
i t i t
E t E e E e EE
e e - -
= + +

. (1.13)
By assuming that we need not worry about the possible frequency dependence of the
nonlinear response, the second-order polarization is obtained as
( )
( )
( )
( )
2 2 2
0
P t E t c _ =

and
includes the following frequency components
5
( )
( )
( ) ( )
( )
( )
( )
( )
2 2
0
2
2
0
2
0
2
2 2
0 2
P E
P P E
P EE
e c _
e e c _
c _
- -
-
=
= =
=
. (1.14)
This result already shows that one needs to be careful in book keeping, because the zero fre-
quency component has an additional factor of 2 compared to the oscillating components.

Optical rectification
The last component of Eq. (1.14) is known as optical rectification. It leads to a nonvanishing
voltage across a second-order material as light is incident on the material. However, it in-
volves no new radiation fields as only oscillating polarization can radiate.

Second-harmonic generation
The two first components of Eq. (1.14) are
complex conjugates of each other. They there-
fore correspond to the same process known as
optical second-harmonic generation or frequen-
cy doubling. Because the oscillating polariza-
tion acts as a source of radiation at its frequen-
cy, the results suggest that materials with a second-order response can give rise to light at new
frequencies, here at the doubled frequency compared to the incident field (Fig. 1.2). In terms
of photons, two photons are destroyed from the incident field at the fundamental frequency
and a photon is generated at the second-harmonic frequency. This way energy is conserved in
the process.

Sum- and difference-frequency generation
We next consider the case where the optical field contains two frequencies
( )
1 1 2 2
1 1 2 2
i t i t i t i t
E t E e E e E e E e
e e e e - -
= + + +

. (1.15)
Squaring this, we find that it is convenient to express the polarization as a sum of several fre-
quency components, negative and positive
( )
( )
( ) ( )
2 2
0
n
i t
n
n
P t E t P e
e
c _ e

= =

, (1.16)
and we find the following frequency components from the second-order polarization

Figure 1.2. Second-harmonic generation (a) and its
photon diagram (b).
e
e
2e
_
(2)
e
e
2e
(a) (b)
6
( )
( )
( )
( )
( )
( )
( )
( )
( )
( )
( )
( )
( )
( )
( )
( )
( )
( )
( )
( )
( )
2
2 2 2
1 0 1 1 0 1
2
2 2 2
2 0 2 2 0 2
2 2
1 2 0 1 2 1 2 0 1 2
2 2
1 2 0 1 2 1 2 0 1 2
2
0 1 1 2 2
2 2
2 2
2 2
2 2
0 2
P E P E
P E P E
P E E P E E
P E E P E E
P E E E E
e c _ e c _
e c _ e c _
e e c _ e e c _
e e c _ e e c _
c _
-
-
- -
- -
- -
= =
= =
+ = =
= + =
= +
. (1.17)
Here, the components on each line are com-
plex conjugates of each other and they corre-
spond to the same physical process. We
therefore conclude that second-order effects
can lead to new frequencies that correspond
to the sums and differences of the original
frequencies (Fig. 1.3).
Due to the two ways of describing a
given process, we can always choose the
frequency of the polarization to be positive.
For
1 2
e e > , sum- and difference-frequency generation are thus described by the terms
( )
( ) 2
1 2 0 1 2
P E E e e c _ + = , (1.18)
( )
( ) 2
1 2 0 1 2
P E E e e c _
-
= , (1.19)
respectively. The processes are described by the photon diagrams also shown in Fig. 1.3.
1.4 Photon diagrams
The photon diagrams for second-harmonic generation and sum- and difference-frequency
generation (Figs. 1.2 and 1.3) follow certain conventions. First, real atomic or molecular en-
ergy levels are indicated by solid horizontal lines. The dashed horizontal lines correspond to
virtual states, which can be thought of as being created at locations where the photons drive
the system. Their lifetime is very short, implying that the system can stay there only for an
instant. The interaction is therefore essentially instantaneous.
In addition, an input field at the positive frequency (complex field) drives the material
system up in energy (upward thick arrow). A photon is thus destroyed. An input field at nega-
tive frequency (conjugate field), on the other hand, drives the system down (downward thick
arrow). A photon is created at this frequency, thus providing a gain mechanism. The polariza-
tion always has positive frequency and it provides a source for radiation. A photon at the fre-

Figure 1.3. Sum- (a) and difference- (c) frequency gen-
eration and their photon diagrams (b) and (d).
(a) (b)
e
1
+e
2
_
(2)
e
1
e
2 e
1
e
2
e
1
+e
2
(c) (d)
e
1
e
2
_
(2)
e
1
e
2
e
2
e
1
e
1
e
2
7
quency of polarization is thus created (downward thin arrow) and the system returns to the
ground state.
1.5 Parametric oscillators
Difference-frequency generation provides a
gain mechanism also for the lower input fre-
quency. In addition, the nonlinear polariza-
tion produces a photon at the difference fre-
quency. In some cases and with the aid of an
optical cavity, the gain may be sufficiently
high that the system starts oscillating, very much like a laser (Fig. 1.4), when the system is
pumped at the higher input frequency. This is known as parametric oscillation.
Parametric oscillation is usually described by the convention where the generated fre-
quencies are
1
e and
2
e , whereas the pump frequency is
3 1 2
e e e = + . The pump photons
thus split into the two lower-frequency photons, which are known as signal and idler.
Parametric oscillators can provide tunable sources of radiation, where the frequency se-
lection is achieved by cavity resonance and/or phase matching.
1.6 Third-order effects
In order to consider third-order effects, we again take the optical field to contain only one fre-
quency and to be of the form
( )
i t i t
E t Ee E e
e e -
= +

. (1.20)
By calculating the third power of this expression, we find that the third-order polarization
contains the following frequencies
( )
( )
( )
( )
( )
( )
( )
( )
( )
( )
3
3 3 3
0 0
2
3 3 2
0 0
3 , 3
3 , 3
P E P E
P E E P E E
e c _ e c _
e c _ e c _
-
- -
= =
= =
. (1.21)
Third-harmonic generation
The terms in the first line correspond to third-
harmonic generation (Fig. 1.5), which is con-
ceptually a very similar process to second-
harmonic generation. Three photons at the orig-
inal frequency e are annihilated and a photon at
frequency 3e is created.


3
e
1
e
2
e
3
e
2
e
3
e
pump
signal
idler
1
e

Figure 1.4. Parametric oscillator, where pump photons
at e
3
split into lower-frequency photons at e
1
and e
2
.

Figure 1.5. Third-harmonic generation (a) and its
photon diagram (b).
e
e
3e
_
(3)
(a) (b)
e
3e
e
e
8
Nonlinear index of refraction
The terms in the second line of Eq. (1.21) show that the
third-order response also oscillates at the original frequen-
cy of the incident field. This is conceptually different to
harmonic generation because no new frequencies are gen-
erated. The positive frequency part of the response at e
can be written as
( )
( )
( ) 2 3 3
0
3 P E E e c _ = , (1.22)
which is represented by the photon diagram shown in Fig. 1.6.
The response at the original frequency can be interpreted as a contribution to effective
linear polarization at frequency e. The total effective polarization is thus as
( )
( )
( )
( )
( )
( ) ( )
( ) ( )
1 3
2 1 3
0 0
1 3
0
0 0
3
3
2
P P P
E E E
I E
n c
e e e
c _ c _
c _ _
c
= +
= +
| |
= +
|
\ .
, (1.23)
where we have used the result that the intensity of the field in our notation is
2
0 0
2 I n c E c = , (1.24)
and c is the speed of light and
0
n

is the linear index of refraction. This result suggests that the
third-order response modifies the index of refraction experienced by a sufficiently strong
beam. The nonlinear index of refraction
2
n

is defined as
0 2
n n n I = + , (1.25)
so that it is
( ) 3
2
2
0 0
3
4
n
n c
_
c
= . (1.26)
Self-focusing and defocusing
Typical laser beams are not plane waves but
often have a Gaussian intensity distribution.
The intensity is then higher at the center and
the nonlinear change in the index varies
across the beam profile. For materials with

Figure 1.6. Photon diagram for the
third-order response at frequency e.
e
e
e
e

Figure 1.7. Self-focusing (a) and self-defocusing (b).
(a) (b)
_
(3)
>0 _
(3)
<0
9
positive nonlinear index of refraction, the situation is like for a positive lens, which tends to
focus the beam (Fig. 1.7). For negative nonlinear index of refraction, the material looks like a
negative lens and the beam diverges.

Wave mixing
We next consider the case where the field at frequency e consists of two plane waves that
propagate in different directions and is thus of the form
1 2
1 2
i i
E A e A e

= +
k r k r
. (1.27)
The nonlinear polarization then contains the following term form
( )
1 2
2 (3) (3) 2 *
0 1 2
( ) 3
i
P A A e e c _

=
k k r
. (1.28)
This result shows that nonlinear processes can also
generate beams that propagate at new directions
compared to the incident beams (Fig. 1.8). The
strength of the important new beams depends on
the relative strength of the incident beams.
1.7 Parametric processes
Nonlinear processes are often said to be para-
metric. This means that the quantum state of the
material remains unchanged after the interac-
tion. A simple example is the index of refrac-
tion, which can be described by the photon dia-
gram shown in Fig. 1.9a. A simple example of a
non-parametric process is linear absorption,
where a photon is absorbed and the material becomes excited Fig. 1.9b. Second-harmonic
generation is usually thought to be parametric. However, if one of the virtual states is close to
a real state of the material, the material can become excited and the process is not parametric
any more (Fig. 1.9c). Parametric processes are described by susceptibilities that are real val-
ued. The energy of the photons is then conserved.





Figure 1.8. Generation of a new beam through
third-order nonlinearity.
k
1
k
1
k
2
k
2
2k
1
- k
2
_
(3)

Figure 1.9. Photon diagrams for index of refraction
(a), absorption (b), and two-photon-resonant second-
harmonic generation.
e
e
2e
(a)
(c)
e
e
e
(b)
10
1.8 Definition of nonlinear optical susceptibility
We assume that the optical field can be expressed as a sum over contributions at several dis-
crete frequencies as
| |

+ =

n
t i
n
t i
n
n n
e e t
e e
e e ) ( ) ( ) (
*
E E E , (1.29)
where the summation is over positive frequencies. By extending the summation to include
also the negative frequencies, this can be expressed as

=
n
t i
n
n
e t
e
e ) ( ) ( E E . (1.30)
The reality of the total field requires that the amplitudes of the negative and positive frequen-
cy parts are related by
) ( ) (
*
n n
e e E E = . (1.31)
Analogous to Eq. (1.30), we also expand the material polarization as a sum over different fre-
quency components as

=
n
t i
n
n
e t
e
e ) ( ) ( P P . (1.32)
The electric field and polarization are vectorial quantities. In addition, the susceptibili-
ties are expected to depend on the frequencies of the optical field.

Second-order susceptibility
We define the second-order susceptibility tensor formally as
(2)
0
( )
( ) ( ; , ) ( ) ( )
i n m n m n m j n k m
ijk
jk nm
P E E e e c _ e e e e e e + = +

, (1.33)
where the indices ijk refer to the Cartesian coordinates and the notation (nm) indicates that
n m
e e + is to be held fixed in summation over n and m. We next consider the consequences
of this definition for two special cases.

Sum-frequency generation
The input fields are
1
( ) e E and
2
( ) e E , and the sum frequency is denoted by
3 1 2
e e e = + .
By performing the summation over n and m explicitly, we obtain
11
(2)
3 0 3 1 2 1 2
(2)
3 2 1 2 1
( ) ( ; , ) ( ) ( )
( ; , ) ( ) ( )
i j k
ijk
jk
j k
ijk
P E E
E E
e c _ e e e e e
_ e e e e e

=

(
+

. (1.34)
Here, j and k go through all possible values and can be interchanged in the second term, which
becomes
(2)
3 2 1 2 1
( ; , ) ( ) ( )
k j
ikj
E E _ e e e e e . (1.35)
Furthermore, the susceptibility can be defined to have the intrinsic permutation symmetry
(2) (2)
3 2 1 3 1 2
( ; , ) ( ; , )
ikj ijk
_ e e e _ e e e = , (1.36)
so that Eq. (1.34) becomes
(2)
3 0 3 1 2 1 2
( ) 2 ( ; , ) ( ) ( )
i j k
ijk
jk
P E E e c _ e e e e e =

. (1.37)
For example, when both input fields are x-polarized, this becomes
(2)
3 0 3 1 2 1 2
( ) 2 ( ; , ) ( ) ( )
i x x ixx
P E E e c _ e e e e e = . (1.38)
Note that these expressions include an overall factor of two, which arises from the two possi-
ble permutations of the input frequencies and is known as the degeneracy factor. This factor,
however, is sometimes included in the definition of the susceptibility, which complicates the
comparison of the results from different sources.

Second-harmonic generation
The input field is
1
( ) e E . We still denote the generated frequency by
3 1
2 e e = . Now, there is
nothing to be summed in the frequencies and we obtain
(2)
3 0 3 1 1 1 1
( ) ( ; , ) ( ) ( )
i j k
ijk
P E E e c _ e e e e e = . (1.39)
For the x-polarized input, this becomes
(2) 2
3 0 3 1 1 1
( ) ( ; , ) ( )
i x ixx
P E e c _ e e e e = . (1.40)
This expression does not include the factor of two, because there are no frequency permuta-
tions for second-harmonic generation. Note that the susceptibility for is still a continuous
quantity as
2 1
e e in
(2)
3 1 2
( ; , )
ijk
_ e e e .

12
Third-order susceptibility
The third-order susceptibility is defined in a way analogous to Eq. (1.33) as
(2)
0
( )
( )
( ; , , ) ( ) ( ) ( )
i o n m
o n m o n m j o k n l m
ijkl
jkl onm
P
E E E
e e e
c _ e e e e e e e e e
+ +
= + +

. (1.41)
Number of tensor components
The second-order susceptibility is a third-rank tensor. Hence, it has in general 3
3
= 27 inde-
pendent complex-valued components. Similarly, the third-order susceptibility has 3
4
= 81 in-
dependent components. The number of independent components can fortunately be reduced
by symmetry arguments.
1.9 Lorentz model
The Lorentz model is a classical model that allows cer-
tain optical properties of materials to be described. The
model assumes that an electron bound to an atom is
equivalent to a harmonic oscillator (Fig. 1.1), i.e., a
mass (electron) attached with a spring to a much heav-
ier mass (nucleus). The model is usually one-
dimensional where the electron moves only in x direc-
tion, but this is valid for isotropic materials.
An ideal harmonic oscillator is described by the parabolic potential (Fig. 1.10)
( )
2 1
2
U x kx =

. (1.42)
When the electron is displaced from its equilibrium by amount x, the spring thus provides a
restoring force
dU
F kx
dx
= =

. (1.43)
The potential of realistic systems, however, is not completely parabolic. In certain mate-
rial systems, the potential may exhibit asymmetry between the positive and negative x direc-
tions. This can be described by adding a cubic correction to the potential (Fig. 1.10)
( )
2 3 1 1
2 3
U x kx Ax = +

, (1.44)
so that the restoring force becomes

Figure 1.10. Potentials for a harmonic os-
cillator (blue) and an anharmonic oscillator
in a noncentrosymmetric material (red).
x
harmonic
real
U(x)
13
2
dU
F kx Ax
dx
= =

. (1.45)
When the electron is driven by an oscillating optical field, the total force includes the
force from the field on the electric charge e. By using the tilde ~ again to denote quantities
that may exhibit fast temporal variations, we obtain the equation of motion for the electron
2
mx eE kx Ax =



, (1.46)
where m is the mass of the electron. This equation does not yet include damping due to fric-
tional forces, which are proportional to velocity. By adding this term and recalling the reso-
nance frequency of the system
2
0
k m e = , we end up with the final equation of motion using
rather standard notation
2 2
0
2 x x x ax eE m e + + + =



, (1.47)
where is the damping rate and a = A/m.
We consider the case where the atom interacts with an optical field that consists of two
distinct frequencies
( )
1 2
1 2
. .
i t i t
E t E e E e c c
e e
= + +

(1.48)
Perturbation theory
Unfortunately, there is no easy way to solve the above problem in the general case and in
closed form. We therefore resort to perturbation theory, where the magnitude of the optical
field is scaled by a parameter , i.e., ( ) ( ) E t E t

. Furthermore, we seek solution as a pow-
er series with respect to
( ) ( ) ( ) 1 2 3 2 3
... x x x x = + + + . (1.49)
Due to the fact that is an arbitrary scaling constant, the equation we obtain must hold sepa-
rately for each power of lambda. We then obtain separate equations for the different powers
( ) ( ) ( ) 1 1 1 2
0
2 x x x eE m e + + =



, (1.50)
( ) ( ) ( ) ( )
2
2 2 2 1 2 2
0
2 0 x x x a x e
(
+ + + =
(




. (1.51)
Note that Eq. (1.50) for the first-order problem is now exactly the same as would be ob-
tained for a standard harmonic oscillator without the anharmonic term. The second-order
problem [Eq. (1.51)], on the other hand, depends only on the solution to the first-order prob-
lem. The idea of perturbation theory is thus to first solve the main problem using standard
14
techniques and then use this solution to seek for the next higher-order corrections that de-
scribe the actual problem even better.

First-order solution
We solve the first-order problem by assuming that the electron tends to follow the driving
field. Our trial solution is thus of the form
( )
( )
( )
( )
( )
( )
1 2
1 1 1
1 2
. .
i t i t
x t x e x e c c
e e
e e

= + + , (1.52)
where the amplitudes are constant. The first and second time derivatives of this expression are
needed, but they are easy to calculate. When Eq. (1.52) and its derivatives are inserted in Eq.
(1.50), we end up with a rather complicated expression. However, this expression must hold at
all times and therefore it must hold for each frequency component separately. Focusing our
attention to the terms that oscillate at frequency
1
e , we obtain
( )
( ) ( )
( )
( )
( )
( )
1 1 1 2 2
1 1 1 1 0 1 1
2 x i x x eE m e e e e e e + + = . (1.53)
This equation is straightforward to solve and we express the solution in terms of the
resonance denominator
( )
2 2
0
2 D i e e e e = . (1.54)
The first-order solution is thus found to be
( )
( )
( )
1
1
1
1
E e
x
m D
e
e
= , (1.55)
with similar solutions for any other possible frequency component.
It is well known that atoms left at rest do not possess a dipole moment although the
electron is not at the same location as the nucleus. The dipole moment therefore arises from
the displacement of the electron from the equilibrium. The dipole moment is thus ex =
and the polarization P Nex = , where N is the number density of atoms. The linear polariza-
tion oscillating at frequency
1
e is thus
( )
( )
( )
( )
( )
2
1 1
1
1 1
1
E Ne
P Nex
m D
e e
e
= = . (1.56)
On the other hand, this is by definition
( )
( )
( )
( )
1 1
1 0 1 1
P E e c _ e = , where the linear suscepti-
bility at frequency
1
e is
15
( )
( )
( )
2
1
1
0 1
N e m
D
_ e
c e
= . (1.57)
It is evident that the linear response of the atom to the field is resonantly enhanced when the
frequency of the field is close to the resonance frequency of the atom.

Second-order solution
When we seek solution to the second-order equation [Eq. (1.51)], the last term can be inter-
preted as an effective driving force and it depends on the square of the first-order solution of
the form of Eq. (1.52). The square is seen to contain the following different frequencies
1
2e ,
2
2e ,
( )
1 2
e e , and 0. We will focus on the frequency component
1
2e , whose amplitude
is obtained from Eq. (1.55) as
( )
( )
( )
2 2
2
1
1
1
2 2
1
E e
x
m D
e
e
(
=
(

. (1.58)
The equation corresponding to Eq. (1.53) now becomes
( )
( )
( ) ( )
( )
( )
( )
( )
( )
2 2 2 2 2
1 1 1 1 0 1
2 2
1
2 2
1
2 2 2 2 2
0
x i x x
E e
a
m D
e e e e e e
e
+ +
+ =
, (1.59)
from where we obtain the second-order solution at frequency
( )
( )
( ) ( )
2 2
2
1
1
2 2
1 1
2
2
E e
x a
m D D
e
e e
= . (1.60)
The second-order polarization oscillating at frequency
1
2e is thus
( )
( )
( )
( )
( ) ( )
2 3
2 2
1
1 1
2 2
1 1
2 2
2
E Nae
P Nex
m D D
e e
e e
= = , (1.61)
yielding the second-order susceptibility for second-harmonic generation in the form
( )
( )
( )
( ) ( )
3 2
2
1 1 1
2
0 1 1
2 , ,
2
N e m a
D D
_ e e e
c e e
= . (1.62)
This expression is seen to exhibit resonances when either the fundamental frequency or se-
cond-harmonic frequency is close to the resonance of the Lorenz atom.
In the same way, we obtain the second-order susceptibilities for sum- and difference-
16
frequency generation of the input fields
( )
( )
( )
( ) ( ) ( )
3 2
2
1 2 1 2
0 1 2 1 2
, ,
N e m a
D D D
_ e e e e
c e e e e
+ =
+
, (1.63)
( )
( )
( )
( ) ( ) ( )
3 2
2
1 2 1 2
0 1 2 1 2
, ,
N e m a
D D D
_ e e e e
c e e e e
=

. (1.64)

Millers rule
The results of the Lorentz model can be used to find an interesting relation between the linear
and nonlinear optical response. From Eq. (1.57) and similar expressions for the linear suscep-
tibilities at frequencies
2
e and
1 2
e e + as well as Eq. (1.63) for the second-order susceptibil-
ity, we obtain
( )
( )
( )
( )
( )
( )
( )
( )
2
2
1 2 1 2 0
2 3 1 1 1
1 2 1 2
, , ma
N e
_ e e e e c
_ e e _ e _ e
+
=
+
. (1.65)
This quantity has experimentally been found to be almost constant for almost all solid-
state materials that have a second-order response. This can be true only if the ratio
2
a N is
constant. The lattice constant d of most crystals varies only a little and is about 3 , implying
that the atomic number density N is also almost constant. The anharmonicity constant a can
be estimated by assuming that the linear and nonlinear contributions to the restoring force are
comparable when the electron is displaced by the lattice constant, i.e.,
2 2
0
d ad e = , (1.66)
yielding
2
0
a d e = . (1.67)
This quantity is indeed almost constant, because the resonance frequencies
0
e

of most crys-
tals also fall in the range of
16
0
10 rad s e .

Estimate for second-order susceptibility
A similar approach can be used to estimate the magnitude of the second-order susceptibility
for the case where all frequencies are much smaller than the resonance frequency. The reso-
nance denominator is then dominated by the quantity
2
0
e , and we find that the susceptibility
for any nonlinear process is approximately
17
( )
( ) ( )( )( )
3 2 3 3 2 2
0
2
6 6
0 0 0 0
1
7pm V
N e m a d e m d e
_
c e c e
= = ~ , (1.68)
where the numerical value was obtained by using the above estimates for the lattice constant
and resonance frequency. This number has the correct order of magnitude for real materials.
1.10 Symmetry
We consider the interaction between three different frequencies that satisfy the condition
3 1 2
e e e = + . If all three fields are sufficiently strong, they can interact in all possible ways
through sum- and difference frequency generation. In order to describe this fully, we therefore
need to know all the components of the following tensors
( ) ( )
( ) ( )
( ) ( )
2 2
3 1 2 3 2 1
2 2
1 3 2 1 2 3
2 2
2 3 1 2 1 3
( ; , ) ( ; , )
( ; , ) ( ; , )
( ; , ) ( ; , )
_ e e e _ e e e
_ e e e _ e e e
_ e e e _ e e e


. (1.69)
In addition, we need six more tensors where each frequency is replaced by a frequency with
opposite sign. We therefore need a total of 12 tensors. Each of these tensors is a third-rank
tensor, i.e., it has three indices i, j, k, which all can have as values all possible spatial direc-
tions x, y, z. Each tensor thus has 27 components and we would need 12*27=324 complex
numbers to fully describe the interaction.
Fortunately, the number of independent tensor components can be greatly reduced by
symmetry principles. We do not get deep into the symmetry principles in this course, but do
outline the most important ones in the following.

Reality of the physical field
The time-dependent real-valued optical field is again expressed as a sum over contributions at
several discrete frequencies as
*
0
( ) ( ) ( ) ( )
n n n
n n
i t i t i t
n n n
t e e e
e e e
e e
e e e

>
(
= = +


E E E E . (1.70)
The reality of the total field requires that
) ( ) (
*
n n
e e E E =

. (1.71)
The nonlinear polarization can be expressed a similar form, which results in the condition

18
*
( ) ( )
n m n m
e e e e = + P P . (1.72)
By considering the definition of the second-order susceptibility
(2)
0
( )
( ) ( ; , ) ( ) ( )
i n m n m n m j n k m
ijk
jk nm
P E E e e c _ e e e e e e + = +

, (1.73)
we can immediately conclude that the reality requires that the following symmetry property is
satisfied
*
(2) (2)
( ; , ) ( ; , )
p q p q p q p q
ijk ijk
_ e e e e _ e e e e
(
= +

. (1.74)
This result immediately reduces the number of independent quantities by a factor of two.

Intrinsic permutation symmetry
The intrinsic permutation symmetry was already mentioned and used to get from Eq. (1.34) to
Eq. (1.37). It requires

) , ; ( ) , ; (
) 2 ( ) 2 (
q p q p
ijk
p q q p
ikj
e e e e _ e e e e _ + = + . (1.75)
There is no fundamental reason for this symmetry. However, it is useful for book keeping and
the susceptibility is usually defined so that this symmetry is fulfilled. This symmetry implies
that the frequencies
n
e and
m
e can be freely interchanged as long as the indices j and k are
simultaneously interchanged.

Lossless materials
The nonlinear material can be assumed to be lossless when the optical frequencies are detuned
far from the material resonances. In addition, the damping rate is usually much smaller than
the resonance frequency. The resonance denominator of the Lorentz model can then be ap-
proximated as
( )
2 2 2 2
0 0
2 D i e e e e e e = ~ , (1.76)
and, consequently, the susceptibility is a real-valued quantity for lossless media. In the same
limit, the susceptibility can be shown to satisfy the full permutation symmetry, which implies
that all frequencies can be freely permuted as long as the indices i, j, and k are simultaneously
permuted. In applying this rule, we have to remember that the output frequency is always the
sum of the two input frequencies. For example
19
(2) (2)
3 1 2 1 2 3
(2) (2)
1 2 3 1 2 3
( ; , ) ( ; , )
( ; , ) ( ; , )
ijk jki
jki jki
_ e e e _ e e e
_ e e e _ e e e
-
=
(
= =

, (1.77)
where the last step followed from the reality of the susceptibility in lossless media.

Kleinman symmetry
The optical frequencies are often much smaller than any of the resonance frequencies of the
material, i.e.,
0
e e . The resonance denominator is then further simplified to the form
( )
2
0
D e e ~ , (1.78)
and the susceptibility is then independent of the optical frequencies. In this case, the indices i,
j, and k can be permuted without permuting the frequencies, i.e.,
) 2 ( ) 2 ( ) 2 ( ) 2 ( ) 2 ( ) 2 (
kji kij jik jki ikj ijk
_ _ _ _ _ _ = = = = = . (1.79)
This symmetry is known as Kleinman symmetry and it can greatly simplify the description of
nonlinear optical interactions. However, one must be very careful in justifying it in practical
situations, which is often not done properly

Spatial symmetry
The nonlinear susceptibility tensors must also satisfy conditions that reflect the spatial point-
group symmetry of the material. The most powerful symmetry rule is the requirement of non-
centrosymmetry for second-order (and other even-order) processes.
Centrosymmetric materials have spatial inversion as a symmetry operation. This opera-
tion implies that the material looks the same when the sign of each spatial coordinate is in-
verted about some central atom at origin, i.e.,
r r . (1.80)
The electric field and polarization transform similar to the position vector r. Consequently,
0 : ) ( :
) 2 ( 2 ) 2 ( 2 ) 2 ( ) 2 (
= = = = P E E P _ _ . (1.81)
The second-order susceptibility tensor must therefore vanish in all centrosymmetric materials.
This is a very powerful rule because 11 of the 32 point groups are centrosymmetric.
In addition, the nonlinear tensors must be compatible with any other spatial symmetry
operations of the material. The results for the various point groups are tabulated.

20
1.11 Contracted notation
We have described the second-order response using a very general susceptibility defined in
Eq. (1.33) or Eq. (1.37). However, there is another traditional way to describe the response
when the symmetry
(2) (2)
ikj ijk
_ _ = is valid without changing the frequencies. Note that this is
always valid for second-harmonic generation.
A new tensor
ijk
d is defined as
(2) (2)
1
2
ijk ijk
d _ = , (1.82)
so that the nonlinear polarization becomes
0
( )
( ) 2 ( ) ( )
i n m ijk j n k m
jk nm
P d E E e e c e e + =

. (1.83)
However, this tensor is usually defined by replacing jk by just one index l as follows
, , ,
1 2 3 4 5 6
jk xx yy zz yz zy zx xz xy yx
l
. (1.84)
In addition, for the first index i, the spatial coordinates x, y, z are replaced by the numbers 1, 2,
3. The tensor
il
d thus has eighteen components and can be organized in the matrix
11 12 13 14 15 16
21 22 23 24 25 26
31 32 33 34 35 36
il
d d d d d d
d d d d d d d
d d d d d d
| |
|
=
|
|
\ .
. (1.85)
The components of the nonlinear polarization are then obtained from
( )
( )
( )
( )
( ) ( )
( ) ( )
( ) ( )
( ) ( ) ( ) ( )
( ) ( ) ( ) ( )
( ) ( ) ( ) ( )
0
4
x n x m
y n y m
x n m
z n z m
y n m il
y n z m z n y m
z n m
z n x m x n z m
x n y m y n x m
E E
E E
P
E E
P d
E E E E
P
E E E E
E E E E
e e
e e
e e
e e
e e c
e e e e
e e
e e e e
e e e e
| |
|
|
| |
+
|
|
|
+ =
|
+ |
|
| +
\ .
+
|
|
+
\ .
. (1.86)
Note that, for second-harmonic generation, the factor of 4 must be replaced by 2.
It is often the case, that the polarizations of the interacting fields are prescribed. In that
case, only a few of the components
il
d play role and they occur in a combination that de-
pends on the geometry of the experiment. In such cases, the interaction can usually be de-
scribed by only one number, the effective nonlinear coefficient
eff
d .