Nonlinear Optics

Manfred Wöhlecke Klaus Betzler Mirco Imlau

University of Osnabrück Summer Term 2003, rev. 2005

“Physics would be dull and life most unfulfilling

if all physical phenomena around us were linear.
Fortunately, we are living in a nonlinear world.
While linearization beautifies physics, nonlinear-
ity provides excitement in physics.”

Y. R. Shen in The Principles of Nonlinear Optics

ii U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

apl. Prof. Dr. Manfred Wöhlecke

http://www.physik.uni-osnabrueck.de/optmat
mailto:[email protected]
apl. Prof. Dr. Klaus Betzler
http://www.physik.uni-osnabrueck.de/kbetzler
mailto:[email protected]
Jun.-Prof. Dr. Mirco Imlau
http://www.physik.uni-osnabrueck.de/photonik
mailto:[email protected]

Fachbereich Physik
Universität Osnabrück
Barbarastraße 7
49074 Osnabrück

http://www.physik.uni-osnabrueck.de

© Fachbereich Physik, Universität Osnabrück 2003, 2004, 2005

C ONTENTS iii

Contents
1 Introduction 1

2 Tools to describe crystals 2

2.1 Two cubic crystals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.2 Point symmetry operations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.3 Crystal systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.4 The 14 Bravais Lattices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.5 Point groups and lattices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.5.1 Notations of point groups . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.5.2 Cyclic point groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.5.3 Dihedral groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.5.4 Cubic groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.5.5 Centrosymmetric groups . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.5.6 Subgroups of the centrosymmetric groups . . . . . . . . . . . . . . . . . 16
2.6 Space groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.6.1 Nonsymmorphic symmetry operations . . . . . . . . . . . . . . . . . . . 19
2.6.2 Space group notations . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.6.3 Space group examples . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

3 Symmetry induced tensor properties 29

3.1 Neumann’s principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.2 Relationships for tensor components . . . . . . . . . . . . . . . . . . . . . . . . 30
3.3 Direct group theoretical approach . . . . . . . . . . . . . . . . . . . . . . . . . 30

4 Nonlinear Optical Susceptibilities 37

4.1 Nonlinear Polarization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
4.2 The Phase-Matching Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.3 Mechanisms for the Nonlinear Polarization . . . . . . . . . . . . . . . . . . . . 39
4.4 The Anharmonic Oscillator as a Qualitative Model . . . . . . . . . . . . . . . . 40
4.5 Structural Symmetry of Nonlinear Susceptibilities . . . . . . . . . . . . . . . . . 42
4.6 Permutation Symmetry of Nonlinear Susceptibilities . . . . . . . . . . . . . . . 43
4.7 Example: Strontium Barium Niobate . . . . . . . . . . . . . . . . . . . . . . . . 43
4.8 Contraction of Indices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

5 Harmonic Generation 45
5.1 Second-Harmonic Generation . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
5.2 Phase Matching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
5.3 Quasi Phase Matching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
5.4 Walk-Off . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
5.5 High-Order Harmonic Generation . . . . . . . . . . . . . . . . . . . . . . . . . 55
iv U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

6 Measurement of Nonlinear Optical Properties 60

6.1 Powder Technique . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
6.2 Maker Fringe Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
6.3 Absolute Measurements by Phase-Matched SHG . . . . . . . . . . . . . . . . . 65
6.4 Z-Scan Technique . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

7 Non-Collinear Harmonic Generation 70

7.1 Induced Non-Collinear Frequency Doubling . . . . . . . . . . . . . . . . . . . . 70
7.2 Spontaneous Non-Collinear Frequency Doubling . . . . . . . . . . . . . . . . . 73
7.3 Non-Collinear Scattering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
7.4 Conical harmonic generation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
7.5 Domain-Induced Non-Collinear SHG . . . . . . . . . . . . . . . . . . . . . . . 78

8 CW Lasers with intra-cavity second harmonic generation 84

8.1 Fundamentals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
8.1.1 Absorption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
8.1.2 Spontaneous emission . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
8.1.3 Induced emission . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
8.1.4 3-level system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
8.1.5 4-level system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
8.1.6 Optical resonator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
8.1.7 Pump processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
8.2 Cavity design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
8.2.1 Optical resonator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
8.2.2 Laser medium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
8.2.3 Losses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
8.2.4 Dimensions of the laser rod . . . . . . . . . . . . . . . . . . . . . . . . 91
8.2.5 Estimation of the cavity parameter τc . . . . . . . . . . . . . . . . . . . 92
8.2.6 Reduction of unwanted Eigenmodes . . . . . . . . . . . . . . . . . . . . 93
8.2.7 Cavity design with intra-cavity second harmonic generation . . . . . . . 93
8.2.8 Losses by the non-linear crystal . . . . . . . . . . . . . . . . . . . . . . 94
8.2.9 Selection of the non-linear crystal . . . . . . . . . . . . . . . . . . . . . 95

A Matrices for symmetry operations 97

B A tiny group theory primer 100

C Some completions for point groups 102

D Some completions for space groups 106

Textbooks on Nonlinear Optics 107

Index 108
1 I NTRODUCTION 1

1 Introduction

Linearity is one of the basics of classical optics. Light waves usually do not interact. In other fields
of electricity and magnetism, yet, nonlinearities are known since scientists have begun to study
the phenomena in more detail. Saturation effects at high (static) electric or magnetic fields and
nonlinear electrical characteristics of devices like vacuum tubes, semiconductor diodes, and even
resistors are quite familiar examples. In the field of optics, however, nonlinear effects became a
subject of interest only after the invention of the laser.
To measure the nonlinear response of matter to electromagnetic waves in the optical region, in
general high fields are necessary, starting at about 1 kV/cm. The corresponding light intensities
of some kW/cm2 necessitate laser beams. As laser physics started with the ruby laser with its
high pulse intensities, it took only few years after the invention of the laser [1] that many classical
experiments in nonlinear optics were successfully performed. Among the first were the second
order processes like the experiments on second harmonic generation by Franken et al. [2] in 1961,
on sum frequency generation by Bass et al. [3] in 1962, and on optical rectification by Bass et
al. [4] in 1962.
Since that time Nonlinear Optics has become a rapidly growing field in physics. Nonlinearities
are found everywhere in optical applications. Presently, many optical materials are of special
interest in information technologies, photonics as supplement and extension of electronics plays
a steadily increasing role. Nonlinearities in the properties of these optical materials are often of
significant relevance for the technological application – sometimes useful, sometimes hampering.
To understand these nonlinearities – and to use them for new effects – will be of basic importance
for the further development of photonic applications.
These lecture notes cover some basic topics in nonlinear optics, they accompany lectures held for
the Ph. D. students in the graduate college Nonlinearities of Optical Materials.
The first part gives a short introduction to the physics of crystals and the treatment of symmetry-
dependent properties. Then the nonlinear susceptibility is shortly discussed followed by a sec-
tion about harmonic generation with an emphasis put on second-order and high-order processes.
Thereafter various techniques for the measurement of nonlinear optical properties of crystals are
described. A subsequent chapter deals with non-collinear harmonic generation processes and some
of their applications.

References
[1] T. H. Maiman. Stimulated Optical Radiation in Ruby. Nature 187, 493–494 (1960).
[2] P. A. Franken, A. E. Hill, C. W. Peters, G. Weinreich. Generation of Optical Harmonics. Phys.
Rev. Lett. 7, 118 (1961).
[3] M. Bass, P. A. Franken, A. E. Hill, C. W. Peters, G. Weinreich. Optical Mixing. Phys. Rev.
Lett. 8, 18 (1962).
[4] M. Bass, P. A. Franken, J. F. Ward, G. Weinreich. Optical Rectification. Phys. Rev. Lett. 9,
446 (1962).
2 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

2 Tools to describe crystals

2.1 Two cubic crystals

Usually an introduction to point and space groups starts with a definition of symmetry elements
and a lattice. We will begin with an inspection of two cubic crystals (see Figure 1) and ask the
following question:
How can we describe the crystal structure in a simple and elegant way?

Figure 1: Crystal structure of CsCl (left), structure of diamond (right)

The CsCl crystal looks quite simple, eight Cs ions are at the corners and one Cl in the cube
center. Diamond looks much more complicate and we hope that we do not have to write down
the coordinates of ions in all these many positions. One solution would be to generate with some
operator all ions in similar positions, simply by applying a translational operation. But what means
similar position? In CsCl we may regard the Cs ions as being in a similar position. If we extend this
to the Cl ions, then we have a procedure to describe the crystal. But this idea immediately fails or
becomes very complicate in crystals like diamond, where we need four or five such sets of similar
positions depending on the translational operations allowed. A much more clever approach uses a
lattice and some basic pattern, often called the basis and sometimes motif. A lattice is related to
our sets of similar positions, but not identical.
A lattice is a collection of mathematical points arranged in such a way that each lattice point has
the same environment in the same orientation.
If we apply this idea to CsCl , we see that the centered Cl ion is not qualified as a lattice point,
because a movement to the right brings us sooner or later to another Cl ion, but if we do the same
starting from a Cs ion we meet another Cs, thus we have no equal environment moving in the same
direction. Similar violations are faced with diamond. Here are some warnings:

(a) There are no constraints that ions have to be placed on lattice points!
2 T OOLS TO DESCRIBE CRYSTALS 3

(b) Ions may occupy positions which are not lattice points!

With respect to the warnings (a) and (b) we may collect some ions and regard them as belonging
to one lattice point. Let us consider the Cs plus the Cl ion as one unit and regard them as a basis.
Then the lattice points coincide with the positions of the Cs ions. With this choice everything
looks fine, but what about the diamond? If we think of the same kind of lattice, fixed to one C ion
in a cube corner, the basis grows up and consists of one C in the origin, three C in the center of the
cube faces (squares) and four C in the octants. This looks quite complicate. An easier description
is obtained if we construct a lattice with lattice points in the corner of the cube and the center of
the faces. Then our basis reduces to C in the origin and one C in the octant at a position ( 14 , 14 , 14 ).
This finding can be generalized as follows:
A crystal structure is made up of a lattice and a basis.
To get an idea how many lattices exist in 3 dimensions we derive first the crystal systems and then
the 14 Bravais lattices. For the first step we need the symmetry operations leaving one point fixed.

2.2 Point symmetry operations

A point-symmetry operation is a symmetry operation that keeps one point in space fixed. A set
of such operations may form a group. For a mathematical description we take three vectors, a,
b, and c, measured from a common origin, in such a way that a and b are not collinear and c
is not coplanar with the ab-plane. These three vectors act as axes of reference and need not be
orthogonal. There are basically two ways of describing the effect of symmetry operations with
an operator R. We can set up a symmetry operator which moves all points or position vectors
of space with all vectors referred to a set of fixed axes. Alternatively, the symmetry operator can
be made to move the axes of reference leaving all points in space, and hence position vectors,
unmoved. The former type of operator is called an active operator Ractive , and the latter, which is
its inverse, is called a passive operator Rpassive . We shall use only active operators here. The two
−1
conventions are linked by Ractive = Rpassive , but do not forget the change of the multiplication
order. In active mode we have Ractive × Tactive = Sactive but in passive mode the order is reversed
Tpassive × Rpassive = Spassive .
Under a point-symmetry operation R a point P with position vector r will be moved to a point P 0
with position vector r0 governed by an equation of the form
 0    
x r11 r12 r13 x
r0 = R r or  y 0  =  r21 r22 r23   y  (2.1)
z 0 r31 r32 r33 z

An alternative is the scheme x0i and xi with (i = 1, 2, 3)

 0    
x1 r11 r12 r13 x1 X3
 x02  =  r21 r22 r23   x2  or x0i = rij xj or x0i = rij xj (2.2)
x03 r31 r32 r33 x3 j=1
4 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

and the Einstein convention of calculating the sum if a suffix occurs twice.
Besides the identity, that means do nothing with the object, there are two important point-symmetry
operations, the inversion and the rotation. There are two types of notations for point-symmetry
operations, the International System (preferred by crystallographers) and the Schönflies System
(often used in Solid State Physics).

point-symmetry International Schönflies

element notation notation

identity 1 E
inversion 1̄ i
rotation n[uvw] Cn [uvw]

mirror reflection m σ
σh : mirror reflection in the horizontal
plane
σv : mirror reflection in the vertical
plane
σd : mirror reflection in the vertical di-
agonal plane
rotation inversion n̄ Sn

Table 1: Types of point-symmetry elements and the different notations

In Table 1 several entries have to be explained. We start with the elements n or Cn [uvw] which
describe a rotation by an amount 2π/n, where n is called the order of the rotation, which is per-
formed according to the counter-clockwise convention. In crystals we have only n = 1, 2, 3, 4, 6.
This rotation is often called a pure or proper rotation. Since rotations are taken about certain
rotation axes we have to specify these axes. There is no generally accepted way of doing this,
often the axis is described by a vector with components u, v, and w with respect to a, b, and c.
Usually crystallographers write brackets [uvw], because parentheses are reserved for planes. We

Figure 2: Inversion operation

point out a convention about the sequence of two operations: if two operations A B are applied to
an object, the operation on the right B is considered to be applied first and A second. Finally we
2 T OOLS TO DESCRIBE CRYSTALS 5

emphasize that the inversion (inversion through a center) changes right hands to left hands and we
illustrate this in Figure 2.
All the operations below the horizontal line in Table 1 consist of a rotation n (Cn ) followed by an
inversion 1̄ (i) that means 1̄ n with the short-hand notation n̄ and the special case 1̄ 2 = 2̄ = m (σ).
The operations are called improper rotations . What about the notation of these point-symmetry
elements in the Schönflies scheme? Its name is Sn , but we have to be very careful, because of
the following property n̄ 6= Sn . The reason for this difference is the different approach in the
two systems. As explained above the International System starts with a rotation followed by an
inversion. The Schönflies system starts with a rotation, too, but then a horizontal mirror operation
σh follows. Since σh = i C2 we have Sn = iC2 Cn = 1̄2n with one more 2-rotation than in the
International System. For an illustration we use Figure 3

Figure 3: Illustration of the oper-

ation 4̄ and S4 . Multiple applica-
tions of 4̄ yields:
4̄ 4̄ 4̄ 4̄
A =⇒ B =⇒ C =⇒ D =⇒ A
in analogy we find for S4 :
S4 S4 S4 S4
A =⇒ D =⇒ C =⇒ B =⇒ A

2.3 Crystal systems

Because of its translational invariance a lattice is generated by applying primitive translations Tn

to an object. A primitive translation vector tn is described by

tn = n1 a + n2 b + n3 c (2.3)

where ni is an integer and the vectors are those defined in section 2.2. If we construct a paral-
lelepiped with the vectors a, b, and c according to a · (b × c), the space is completely filled by
translations tn applied to this object. Such a volume a · (b × c) is called a unit cell. If it only
contains one lattice point, it is called a primitive unit cell. What means this last definition? If
the origin of a unit cell is chosen on a lattice point, then there are eight lattice points involved
(one at each corner). However, because each lattice point belongs to eight parallelepipeds, in one
primitive cell there is only 8 × 81 = 1 lattice point. The message is even more obvious, if we do
not put the origin in a lattice point. Thus the origin can arbitrarily be chosen anywhere and we add
that all primitive unit cells have the same volume. For an illustration see Figure 4
With the demand that the lattice is invariant with respect to the translations and point-symmetry
operations, relations hold between the vectors a, b, and c and the angles α(b, c), β(c, a), γ(a, b)
between them. The general procedure for deriving the relations is quite simple. One starts with
6 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

r r r r r

P










r
r
r
r
NP r




P


r r r
r r


A P A Figure 4: Examples for primitive and

A A
r r
r Ar Ar non-primitive unit cells
A

 NP A
r r
 r Ar r
A 
A  
r r r
A r r

the most simple point-symmetry operations 1 and 1̄ and applies this to the vectors a, b, and c of a
unit cell. We take the appropriate matrices from Appendix A and obtain the equations
 
1 0 0
r0 = {1}r =  0 1 0  r = x0 a + y 0 b + z 0 c = xa + yb + zc (2.4)
0 0 1
 
−1 0 0
r0 = {1̄}r =  0 −1 0  r = −x0 a − y 0 b − z 0 c = −xa − yb − zc (2.5)
0 0 −1
In both cases the coordinates x, y, z are attached to the same axes as before the symmetry opera-
tion. Thus no restrictions are put on the unit cell shape, we call this system triclinic , the relations
can be read from Table 2.
In a next step we derive the relations for the monoclinic system. We add a rotation 2(C2 ) and
/ or a mirror operation m(σ). In solid state physics the first setting is preferred ( 2 k c), while
in crystallography 2 k b (second setting) is used. Let us rotate a point inside the unit cell with
coordinates x, y, z by
r 0 = 2[001]r = −xa − yb + zc.

The brackets after the 2 describes the orientation of the rotation axis. With r we obtain for the
scalar product xza · c, while r 0 yields −xza · c. Because 2 is a symmetry operation, we have
xz|a||c| cos β = −xz|a||c| cos β with cos β = − cos β and β = 90◦
which means a ⊥ c. In the same manner we find b ⊥ c. Thus the triclinic system requires
a 6= b 6= c, α = β = 90◦ γ 6= 90◦ .
Often for the unit cell γ > 90◦ is taken, but this is simply a convention.
Next we add two more twofold rotations or mirror planes and apply them to r
2[100]r = xa − yb − zc (2.6)
2[010]r = −xa + yb − zc (2.7)
2[001]r = −xa − yb + zc (2.8)
2 T OOLS TO DESCRIBE CRYSTALS 7

With similar arguments as in the monoclinic system we deduce from equations (2.6 - 2.8) the
relations for the orthorhombic system

a ⊥ b und a ⊥ c from (2.6) b⊥a and b ⊥ c from (2.7)

Since no operation interchanges the components with respect to a, b, c , we yield again for the
unit vectors
a 6= b 6= c and α = β = γ = 90◦ because of (2.8)
.
To derive the conditions for the tetragonal system one allows besides the identity or inversion the
operations 4 or 4̄.

r0 = {4 [001]}r = −ya + xb + zc (2.9)

0 3
r = {4 [001]}r = ya − xb + zc (2.10)

Again, the opposite signs show that all axes are mutually perpendicular and x interchanges with
y, thus a = b.
The hexagonal system will be skipped because of special problems, the reader should consult
chapter 2 in the book of Burns and Glazer [1], where the following discussion of the cubic system
is taken from. This crystal system is familiar to solid state physicists and chemists, and it is
the system with the highest symmetry. However, despite its familiarity we need to be careful
about how we define it; it is not sufficient to use the criterion that all axes are equal and all
angles 90◦ . As we continue to emphasize, the symmetry is the important thing in determining
the crystal system. The symmetry conditions our choice of axes, not the other way round. What,
then, are the important symmetry elements in the cubic system? Surprisingly perhaps, they are
not the three mutually perpendicular 4-fold axes so readily observed in a cube, but rather, the
four 3-fold axes corresponding to the body-diagonals, h111i, of the cubic unit cell. (Note that
angular brackets signify the set of symmetry-equivalent directions. In this case h111i means the
set [111], [1̄1̄1̄], [1̄11], [11̄1̄], [1̄1̄1], [111̄], [1̄11̄], and [11̄1].) It is possible to have a cubic crystal
without any 4-fold axes of symmetry. It is also possible to prove either by group theory or by
spherical trigonometry that if a crystal contains more than one 3-fold axis, then it must contain
four altogether, each one making an angle of 109◦ 280 with any other. The equal length of all axes
follows from the fact that a {3 [111]} operation interchanges all axes.

2.4 The 14 Bravais Lattices

The seven crystal systems of Table 2 define six different parallelepipeds, because the conditions for
the trigonal and the rhombohedral system are equal. These lattices are called P-Lattices, because
they are based on primitive unit cells. But as we will see, there are more possibilities to fill the
whole space by translational operations of some unit. To do this, one has to allow centering of a
cell. Centering means that we put a lattice point somewhere into a cell or on one or more of its
surfaces. After doing this we have to ask:
8 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

1 or 1̄ triclinic a 6= b 6= c
α 6= β 6= γ

2 or 2̄ monoclinic a 6= b 6= c
α = β = 90◦ 6= γ (1st setting)
α = γ = 90◦ 6= β (2nd setting)

two 2 or 2̄ orthorhombic a 6= b 6= c
α = β = γ = 90◦

4 or 4̄ tetragonal a = b 6= c
α = β = γ = 90◦

four 3 or 3̄ cubic a=b=c

α = β = γ = 90◦

6 or 6̄ hexagonal a = b 6= c
α = β = 90◦ ; γ = 120◦

3 or 3̄ trigonal a = b 6= c
α = β = 90◦ ; γ = 120◦
(rhombohedral) a = b = c; α = β = γ 6= 90◦

Table 2: The seven crystal systems. The first column contains the generating point-symmetry
operations. The last one gives the relations for the lengths of the axes and the angles.

a) Is this new arrangement of points really a lattice?

b) Have we produced a new lattice?

As a result of centering eight new lattices with new names show up.

I: body centering
F: face centering
A,B,C: one-face centering (base centering).

Of course, the corresponding centered parallelepipeds are not unit cells, but they have the advan-
tage that they have the symmetry of the underlying P-cell and thus the same axes can be used.
Later we will see, that even for I, F, A, B, C cells primitive unit cells are possible, but with the
disadvantage that they do not display the point symmetry.
Again we follow the book by Burns and Glazer [1] and discuss the idea of centering in some
selected crystal systems.
2 T OOLS TO DESCRIBE CRYSTALS 9

Figure 5: Different aspects of centering a monoclinic lattice. Taken from [1]

Figure 5 shows several trials for centering a monoclinic lattice. We put the 2-axis along c and from
part (b) of this figure we conclude that a centering of the C-face (a, b-plane) does not produce a
new lattice. If we introduce a new a and γ according to the dashed cell, we obtain again a primitive
cell of the monoclinic system. But centering a B-face (a, c-plane) results in a new lattice. The
dashed parallelepiped in part (c) does not fulfill the conditions (length and angle) for a monoclinic
lattice, that means we have no primitive monoclinic lattice. A B-lattice occurred with unit cell
vectors (bold arrows) as shown in part (c) of Figure 5. In the same manner an A-lattice can be
formed, but the B-lattice is preferred. Our next example deals with the tetragonal system. The

Figure 6: Centering of two independent

faces can never form a lattice, since the en-
vironment of all points (as indicated, for ex-
ample, by the dashed lines) is not the same
no matter how one picks the translation vec-
tors. Taken from [1].

existence of a fourfold axis requires a centering of both A- and B-planes, but as explained in the
caption of Figure 6 this does not form a lattice, where each point has the same environment.
Next we try a centering of the C-face. An inspection of the left part of Figure 7 shows that we
have produced a P-lattice, which is rotated by 45◦ . With a body centering we have more luck and
really obtain a new lattice, where each point has the same environment, see Figure 7 right side. A
face centering is possible in the tetragonal system, too, but it turned out, that such a centering can
be described by a rotated body-centered cell, see right part of Figure 7.
10 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

Figure 7: Trials of centering a tetragonal lattice

The remaining Bravais lattices are shown together with those just discussed in Figure 8. Again
we skip the details of hexagonal and trigonal lattices, a detailed discussion is given in the book by
Burns and Glazer [1]. For the cubic system often I-lattices are called bbc = body-centered cubic,
while F-lattices are denoted by fcc = face-centered cubic.
2 T OOLS TO DESCRIBE CRYSTALS 11

Figure 8: All 14 Bravais lattices, taken from [2].

12 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

Primitive unit cells of the Bravais lattices are discussed now. Such cells are of importance for
calculations of electronic energy bands or the dispersion of lattice vibrations. The typical primitive
unit cells for I, F, and C-lattices are shown in Figure 9.

Figure 9: The conventional primitive unit cells for I, F, and C-lattices, taken from [1].

We proceed now with some general remarks on Wigner-Seitz unit cells, taken from Burns and
Glazer [1]. Occasionally some special unit cell is chosen to emphasize certain special aspects of
the crystal structure. For example, there might be a change of structure at some temperature (a
phase transition) from a very simple high-temperature structure to a much more complicated low-
temperature structure. The unit cell chosen in the simple structure could be relatively complicated
(contain many lattice points centered at various positions), but might be picked to show how the
low-temperature structure follows in a natural way from the high-temperature structure. Thus,
the orientation of the axes in the high-temperature structure might be chosen so that the same
orientation of axes above and below the phase transition is maintained. This choice would make
visualization of the phase transition easy, although it may make the choice of the unit cell in the
high-temperature structure unusual.
Besides this type of special choice of unit cell there is another primitive unit cell of more general
use, particularly in electronic band theory. This is known as the Wigner-Seitz cell . This cell is
2 T OOLS TO DESCRIBE CRYSTALS 13

obtained by starting at any lattice point, called the origin, and drawing vectors to all neighboring
lattice points. Planes are then constructed perpendicular to and passing through the midpoints of
these vectors. The Wigner-Seitz cell is the smallest-volume cell about the origin bounded by these
planes.

Figure 10: Some steps of constructing the Wigner-Seitz cell of a cubic I-lattice. The lattice points
of the origin and the next nearest ones are shown.

Figure 10 shows a step-by-step construction of the Wigner-Seitz cell for a body-centered cubic (I-
cubic) lattice. Note that the Wigner-Seitz cell contains just one lattice point and displays the full
rotational symmetry of the crystal system as do the conventional unit cells of the 14 Bravais lattices
(Fig. 8). Of course, the Wigner-Seitz cell does not form a parallelepiped but it is, nevertheless,
a perfectly acceptable unit cell. Figures 11 (a) and (b) show the Wigner-Seitz cells for I- and F-

Figure 11: Wigner-Seitz cells of the cubic I- and F-lattices, taken from [1].

cubic lattices, where the full cubic symmetry is again obvious. The Wigner-Seitz cell construction
is often used in reciprocal space, or k-space, where the resulting cell is given the name, the first
Brillouin zone. Furthermore, to those readers who are familiar with Brillouin zones we remind
them that an I-cubic lattice in space (direct space) has a reciprocal lattice that is an F-cubic lattice.
Thus, the first Brillouin zone of an I-cubic lattice looks like the Wigner-Seitz cell a direct space
F-cubic lattice and vice versa.
14 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

The Wigner-Seitz cell has been the last topic in our discussion of the lattices. Although there are
still many details missing the present knowledge should be sufficient to understand the description
of crystals.

2.5 Point groups and lattices

On page 3 we have learnt that a crystal is composed of a lattice and the basis. The symmetry
induced properties of the lattice have been treated in the previous two sections. Now we investigate
the influence of the symmetry on the basis. The same is meant, when we ask for point groups
compatible with a given lattice.
To answer this question we give first a short derivation of the 32 point groups allowed in crystals.
In principal there is no need to use group theoretical concepts for such a derivation, but first there
is an intimate connection between these point groups and mathematical groups, and second the
application of group theoretical methods allow a much deeper insight into the structure of point
groups. The details of mathematical groups are considered in appendix B.

2.5.1 Notations of point groups

Schönflies Notation Like for point group symmetry elements, the notations are different for the
groups, too.

Cn : cyclic groups, n = 1, 2, 3, 4, 6
Ci : group composed of inversion and identity
Dn : Cn plus a two-fold rotation axis perpendicular to the main one
Cnh : Cn enlarged by σh
Dnh : Dn enlarged by σh
Cnv : Cn enlarged by σv
Dnd : Dn enlarged by σd
Sn : n-fold rotary reflection n = 2,4,6
T, Th , Td , O, Oh : cubic groups

Table 3: Schönflies notation for point groups

International Notation If we omit the cubic groups, three basic types of the labelling scheme are
apparent.
2 T OOLS TO DESCRIBE CRYSTALS 15

inversion 2-fold axis ⊥ n

? ?
n̄ 22
n-fold rotation axis 6 6 2-fold diagonal axis ⊥ n

vertical mirror plane

?
n̄ mm
6 diagonal mirror plane

vertical mirror plane

?
n/m mm
horizontal mirror plane 6 6 diagonal mirror plane

2.5.2 Cyclic point groups

Cyclic groups occur when there is only one axis of rotation. Because of n= 1,2,3,4,6 we obtain
five groups, which are called according to their generators n; Cn . Do not be confused that here
the labels stay for both the symmetry elements and the group.

2.5.3 Dihedral groups

We add now 2-fold rotation axes (to keep the principal axis) perpendicular to the principal axis.
The direct product of C2 = {E, C2 [100]} and Cn = {E, Cn [001], ...} yields the groups Dn

Cn × {E, C2 [100]} = Dn .

Since C2 contains two elements the order of Dn is twice of that of the cyclic subgroups Cn . We
obtain four new groups (not five, because C1 ≡ E and C1 × C2 = C2 is already known).

Int. Not.: 12346 222 32 422 622

cyclic groups dihedral groups

Schönflies Not.: C1 C2 C3 C4 C6 D2 D3 D4 D6

2.5.4 Cubic groups

A cubic group does not necessarily has a 4-fold axis as already discussed. Two symmetry elements
are sufficient, a 3-fold rotation about a space diagonal and a 2-fold rotation. The products of these
two generators produce altogether 12 rotations and we call the group 23 (T ); please note: 23 (T )
6= 32 (D3 ).
16 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

Adding 4-fold axes to 23 (T ) yields the group 432 (O). Besides 4, 43 about [100], [010], and
[001], six more twofold axes along h110i directions show up. The group order of 432 (O) is 24.
All groups derived sofar consist of pure rotations and we have found altogether 11 of them.

2.5.5 Centrosymmetric groups

Eleven more groups are generated by the direct product with 1̄ = {1, 1̄} = {E, I}. The order of
these groups is twice as that of the subgroups of pure rotations.

International Notation

1 2 3 4 6 222 32 422 622 23 432

1̄ 2/m 3̄ 4/m 6/m mmm 3̄m 4/mmm 6/mmm m3 m3̄m

2.5.6 Subgroups of the centrosymmetric groups

The eleven centrosymmetric groups contain ten subgroups which are unknown to us. Below we
list them together with their subgroups they originate from.

2/m mmm 4/m 4/mmm 3̄m 6/m 6/mmm m3̄m

m mm2 4̄ 42m 4mm 3m 6̄ 6̄m2 6mm 4̄3m
C2h D2h C4h D4h D3d C6h D6h Oh
C1h C2v S4 D2d C4v C3v C3h D3h C6v Td

Thus all the 32 point groups have been introduced. (C1h ≡ Cs ≡ S1 )

Remark: Laue pattern always appear to be centrosymmetric, even so the crystal has no inversion
center. Thus we can determine by simple X-ray diffraction only 11 point groups, more details have
to follow from other physical properties like polar axes.
For those who are more interested in the derivation of the 32 point groups we mention two other
methods. Starting with a proper point group P one searches for an invariant subgroup Q of index
2. Then P can be written P = Q + RQ for some rotation R. Then the group P 0 = Q + I · RQ is
also a point group. (For details see [2])
Probably the most elegant way of deriving the 32 point groups is to start from the symmetry
of the 7 holohedries [1]. These are simply the point symmetries of the lattices of the 7 crystal
systems, and such a derivation highlights the central importance of the lattice itself. To do this,
we just write down the holohedral point goups and list the subgroups that remain in the crystal
system. The holohedral point groups are given as the last entry in the first column of Table 7 of
the corresponding crystal system, see appendix C. We add the following remarks: the lattice has
the highest symmetry, the basis can keep it, or may reduce it to a subgroup, being a member of
the crystal system, but never raises the number of symmetry elements. The basis must have at
2 T OOLS TO DESCRIBE CRYSTALS 17

least the symmetry of the first entry for each crystal system in column 1 of Table 7, because these
symmetry operations are used to derive the relations for the unit vectors (length and angles).
Thus, starting in the triclinic system, the holohedral group is 1̄ and its subgroups are 1 and 1̄ itself.
These are the two point groups of the triclinic system. In the monoclinic system, the holohedral
group is 2/m and its subgroups are 2, m and 2/m (we do not add 1 and 1̄, since they are in the
triclinic system). In the orthorhombic system, the subgroups of mmm are 222, mm2, m2m, 2mm
and mmm itself. In this case mm2, m2m and 2mm are equivalent since they only arise from an
interchange of axes, and so in the orthorhombic system there are just three point groups. In the
tetragonal system, we start from 4/mmm and obtain the subgroups 4, 4̄, 422, 4̄2m, 4m2, 4mm, 4/m
and 4/mmm. Here again, 4̄2m and 4m2 are equivalent and so there are seven unique tetragonal
point groups. For the remaining systems see Table 7 in appendix C, which contains in its second
column the full notation of the International System. For example, n/mmm is the short form of
n/m2/m2/m.
We finish this section with two questions. For example, can we put an object with just 1(C1 )
point symmetry on a cubic Bravais lattice? Conversely, can we put an object with m3̄m(Oh )
point symmetry on a triclinic Bravais lattice? The answer to both these questions is no. The
reason is that, physically, the potential energy of the crystal compatible with a cubic Bravais
lattice can be expanded about a lattice point in combinations of spherical harmonics that have
cubic symmetry. There are no combinations that have lower symmetry. Or putting it in an other
way, since the symmetry operations must act through each lattice point, addition of a non-cubic
object automatically forbids overall cubic symmetry. Thus, in a cubic Bravais lattice, the object
must have one of the cubic point symmetries. Similarly, for a triclinic Bravais lattice the potentials
expanded about a lattice point have triclinic symmetry and so these potentials would distort the
object of higher symmetry and make it adopt triclinic point symmetry. However, remember that
the distortion may be small and difficult to observe experimentally.

2.6 Space groups

Symmetry operations of space groups are combinations of point group operations and translations.
For a description the Seitz-Operator {R|t} is used.

r0 = Rr + t ≡ {R|t}r (2.11)

We quickly show that Seitz-Operators form a group. With {R|t} and {S|u} the product is an
element of the set,

{R|t}{S|u}r = {R|t}(Sr + u) = RSr + Ru + t = {RS|Ru + t}r (2.12)

because Ru is again a translation. The identity is {E|0} ≡ {1|0}. The inverse element of {R|t}
is denoted {R|t}−1 , with
{R|t}−1 = {R−1 | − R−1 t}. (2.13)
The pure translations form an invariant subgroup of the space group.
18 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

Later on we will see, that symmetry operations of the form {R|τ }, with a translation τ being a
fraction of a primitive translation, are possible. Operations like {R|τ } are not essential in all space
groups. Therefore we distinguish

symmorphic Space groups: τ = 0

nonsymmorphic Space groups: with at least one τ 6= 0.

In symmorphic groups elements like {R|τ } may exist, but they are not essential, they are not
needed as a generator, while in nonsymmorphic groups they are essential elements and qualify as
generator.
The point group of a space group G is a set of all elements {R|0}. We denote it Pg and obtain all
its elements from those of G, by setting all tn and τ zero. This group is often called the isogonal
point group.
There exist 230 space groups, 73 of them are symmorphic, the rest nonsymmorphic. This needs
some explanation. Why not 14*32=448 space groups? Remember at the end of the last section we
discussed that the basis cannot have an arbitrary symmetry. We add the following group theoretical
arguments, by inspecting the hierarchy of the crystal systems: each system develops from on (two)
by small distortions.

hexagonal

 

 orthorhombic 


triclinic monoclinic 
J
J
J tetragonal HH
J H
J HH
J H

 cubic
J 
J 
J trigonal 

The symmetry of the unit cell of a crystal system is governed by the following rules:

a) all symmetry operations applied to the basis must produce a result which is compatible with
occupied lattice points. The symmetry operations have to be elements of Pg .

b) The set of symmetry elements of the basis must contain at least one element which does not
belong to the system one level lower in the hierarchy.
2 T OOLS TO DESCRIBE CRYSTALS 19

2.6.1 Nonsymmorphic symmetry operations

We start with the screw operation, or more specific the screw rotation which is a combination of a
proper rotation and a nonprimitive translation τ parallel to the axis of rotation. This axis is called
the screw axis. The order of performing the operations is unimportant. A n-fold rotation axis can
be combined with nonprimitive translation of the form τ = (m/n)tn with m = 1, 2, 3, . . . , n − 1.
Figure 12 illustrates all possible screw operations in crystals including their names and symbols.
The vector tn is always parallel to the rotation axis.

Figure 12: Allowed crystallographic screw operations, taken from [1].

The next operation to be discussed is the glide operation, which is a combination of a mirror
operation and a nonprimitive translation. There are three types of glide planes known. In the axial
glide the magnitude of the translation vector τ is one half of a unit-cell translation parallel to the
20 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

reflection plane. We refer to the axial glide as an a-, b-, or c-glide according to the axis along
which the translation is done. Figure 13 shows a b-glide.

Figure 13: Two b-glide operations, taken

from [1].

A diagonal glide, often called a n-glide, is a combination of a reflection and translations in two or
three directions, e.g., (a + b)/2; (b + c)/2; (a + c)/2 or (a + b + c)/2. The last operation works
only in tetragonal and cubic crystals. In this case the direction of the translation has a component
perpendicular to the mirror plane, while in all other cases the translation is parallel to the mirror
plane. Figure 14 shows two n-glide operations, the mirror plane is perpendicular to c.

Figure 14: Two n-glide operations, taken

from [1].

The diamond or d-glide has translations like (a ± b)/4; (b ± c)/; (a ± c)/4. For tetragonal and
cubic systems the translations are (a ± b ± c)/4.

2.6.2 Space group notations

Schönflies notation for space groups In this scheme the point groups of the space groups are
used. Since a crystal system can have several space groups with the same point group Pg , they
are distinguished solely by a serial number. It is impossible to derive space group operations or
lattices.

International notation for space groups, Herman-Mauguin This drawback is fixed in the Interna-
tional Notation. The general form is
Γ. . .

Γ describes the type of lattice (P, A, B, C, I, F, R). The dots are place-markers for generators which
correspond in monoclinic and orthorhombic systems to the three vectors a, b, c. A C2 rotation
about a is denoted by a 2 at the position of the first point. An a-glide, having its mirror plane
perpendicular to c, is denoted by an a at the position of the third point. In tetragonal, trigonal, and
2 T OOLS TO DESCRIBE CRYSTALS 21

hexagonal systems the principal axis is along c and replaces the first dot. Table 8 in appendix D
gives a summary of the conventions used for directions of symmetry elements and some space
group examples.
Like for point groups there exist a short and a full version, for details consult the second edition
of the International Tables [3].

2.6.3 Space group examples

In this section we continue to explain what kind of information can be taken from the knowledge
of the space group of a crystal. Figure 15 shows the relevant parts of the pages of a tetragonal
space group from the International tables. We start with an explanation of the first line.

Tetragonal 4/m P4/m No. 83 P4/m

? number of the C14h

point group space group
? ? ?
crystal system Int. Symbol Schönflies
full Notation
?
identical with
Int. Short
short form
for space group

Figure 16 belongs to space group number 85. The symbol P4/n points directly to a nonsymmorphic
space group, because it is a generator and 4/n means a n-glide perpendicular to the fourfold rotation
axis. Below the first line some diagrams are drawn with axes following the conventions explained
in Table 8 of appendix D. The left illustrates the action of symmetry operations on a general point,
which is denoted by a small circle. The diagrams display the unit cell, which is identical with the
primitive cell for P-lattices only. The diagram at the right demonstrates the symmetry elements
within the cell. Please notice that the positions in the center of the cell for P4/n does not mean a
centering of the lattice and thus a contradiction to the P-lattice. It is simply a consequence of the
n-glide as a generator. Inside a unit cell the number of circles (bisected circles count twice) has
to be equal to the order of the point group. In the conventional cells of the I-, F-, A-, B-, C- or
R-lattices the number of circles is by a factor 2, 4, 2 or 3 larger according to the volume relations to
the corresponding primitive cells. In the examples P4/m and P4/n we observe eight circles, since
Pg = {1, 4, 2, 43 , 1̄, 4̄, m, 4̄3 }.
The meaning of the left diagram should be clear, otherwise have a look to Figure 3 and the cor-
responding discussion. For an understanding of the right diagram we need some explanation of
the graphical symbols. In appendix D we give a selection from the International Tables [3], for
details see pages 7 - 10 therein. We find inversion centers, fourfold-rotation axes and 4̄ operations,
as well as a diagonal n-glide. Before discussing an additional tetragonal lattice (space group 100)
with some more reflection planes we continue to discuss the information given in the pages of
22 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

Figure 15: The space group P4/m

the International Tables. Below the diagrams the origin is given. Often pages with information
on positions corresponding to different origins are presented. This, for example, is true for space
2 T OOLS TO DESCRIBE CRYSTALS 23

Figure 16: The space group P4/m

group 85. There are rules for the choice of the origin.

a) in symmorphic groups the origin has the symmetry of the point group

b) in nonsymmorphic groups the position with the highest site symmetry is chosen as origin.
24 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

Each possible site is denoted by a letter, starting with a for the site with the highest symmetry
(however note, several sites can have equal site symmetry), and sequentially going through the
alphabet until the lowest site symmetry 1 of a general point is reached. These letters are called
Wyckoff notation. There is a number close to each Wyckoff letter which describes the number of
symmetry-related positions that belong to this particular site. This number is often referred to as
multiplicity and is placed to the left of the Wyckoff position. To its right information on the point
group of this site is given. These point group labels are expanded by some dots, for details see the
explanation in the International Tables [3]. In the next columns coordinates of positions follow.
A site with multiplicity 1 can have only one position while a site with n-fold multiplicity has n
different positions. If one applies space group operations to a certain position, it will permute
with one of the other, never leaving this set. The last sentence sounds strange. We apply space
group operations but characterize this Wyckoff position by a point group. The solution is simple.
The point group of a Wyckoff position is obtained by taking this point as an origin and neglecting
all fractional translations. Therefore the positions (0,0,0) and (0,0, 12 ) in P4/m establish the site
symmetry 4/m, while those in P4/n only the group 4̄. All equivalent positions can be occupied by
atoms or even molecules. But please keep in mind

(a) there is no need to fill all possible Wyckoff positions with atoms

(b) it is allowed to fill one type of Wyckoff positions with different atoms, but they have to have
different values for x, y, z

(c) all positions of a particular Wyckoff site have to be occupied by the same kind of atoms.

In groups with nonprimitive unit cells, e. g., Γ 6= P several positions are given under the line
Coordinates. This means that all Wyckoff positions have to be combined with them.
As a last example we discuss the group P4bm (C24v ). This is a nonsymmorphic tetragonal space
group with details in Figure 17. This group consists of the following 8 symmetry operations:
the identity 1, the 2-fold rotation about c, a 4-fold rotation 4+ and its reverse 4− , both along
c, axial glides (a x, 14 , z) and (b 14 , y, z), a mirror plane (m x + 12 , x̄, z) and a glide reflection
(no standard type, thus no typical label). From this the symmetry operations can be derived and
applied to a general point (x, y, z) to produce the coordinates listed for the Wyckoff position (d).
The matrices for the operations (1) 1; (2) 2 0, 0, z; (3) 4+ 0, 0, z; (4) 4− 0, 0, z; can be taken
directly from appendix A. The next is more complicate. Operation (5) a x, 41 , z means an a-glide
with a fractional translation τ = ( 12 , 0, 0) and a mirror plane through (0, 14 , 0) perpendicular to b
or parallel to (010) planes.
   1  
x + 12
 
1 0 0 x 2
1
{m[010]|τ ( , 0, 0)}r =  0 −1 0   y  +  0  =  −y 
2
0 0 1 z 0 z

with m[010] from the appendix A. If we compare this with the entry (5) for the Wyckoff position
d, a disagreement is observed. The problem is easily fixed, because in appendix A the matrices
are given with respect to the origin (0 0 0), but we have to operate at position (0, 41 , 0). For a
2 T OOLS TO DESCRIBE CRYSTALS 25

Figure 17: The space group P4bm

correction we have to add this shift twice, thus

x + 12 x + 12
     
0
 −y  + 2  1  =  −y + 1  .
4 2
z 0 z

The element (6) b 14 , y, z denotes b-glide with a fractional translation τ = (0, 12 , 0) and a mirror
plane through ( 14 , 0, 0) perpendicular to a or parallel to (100) planes.
 1  
−x + 12
     
−1 0 0 x 0 4
1
{m[100]|τ (0, , 0)}r =  0 1 0   y  +  21  + 2  0  =  y + 12 
2
0 0 1 z 0 0 z
26 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

Element (7) m x + 12 , −x, z describes a mirror operation with a mirror plane bisecting the a and
-b axes and intersecting a at ( 12 , 0, 0). Thus the mirror plane is perpendicular to [1, 1, 0] (for the
corresponding matrix see the appendix A) and with the help of Figure 17 we deduce its shortest
distance from the origin to be in ( 41 , 41 , 0).
 1  
−y + 12
   
0 −1 0 x 4
 −1 0 0   y  + 2  1  =  −x + 1 
4 2
0 0 1 z 0 z

Finally we interpret the operation (8) g( 12 , 12 , 0) x, x, z. The first part means a fractional translation
of τ and the second a mirror plane bisecting a and b and going through the origin.
   1  
y + 12
 
0 1 0 x 2
1 1
{m[11̄0]|τ ( , , 0)}r =  1 0 0   y  +  12  =  x + 12 
2 2
0 0 1 z 0 z

In section 11.2 and 11.3 of the International Tables procedures are discussed to calculate the ma-
trices from the information given under the heading Symmetry Operations. Therefore the entries
of the coordinates for each Wyckoff site are not superfluous but very helpful for a quick inspection
and sufficient if no group theoretical representations but only structural knowledge is involved.
Realizations of this group look somewhat complicate. This is do to the tricky structure which asks
for a large number of atoms in a formula unit. Typical building motifs are tetrahedra (Ce3 Si6 N11 ,
nitridosilicate) or octahedra (unfilled tungsten-bronze structure, like SBN) . In structure data bases
information on SBN (Srx Ba1−x Nb2 O6 ) is available, e. g., Table 4:

label atom multiplicity Wyckhoff x y z occupancy

Nb1 Nb5+ 2 b 0.0 0.5 0.024 1.0
Nb2 Nb5+ 8 d 0.0745(1) 0.2114(1) 0.0080(3) 1.0
Sr1 Sr2+ 2 a 0.0 0.0 0.5000(4) 0.725(4)
Sr2 Sr2+ 8 d 0.1532(6) 0.6864(8) 0.5168(23) 0.202(8)
Ba1 Ba2+ 4 c 0.1729(1) 0.6729(1) 0.5002(9) 0.487(3)
O1 O2− 8 d 0.3435(3) 0.0057(3) -0.0224(27) 1.0
O2 O2− 8 d 0.1393(3) 0.0687(3) -0.0286(27) 1.0
O3 O2− 4 c 0.2819(3) 0.7819(3) -0.0192(23) 1.0
O4 O2− 4 c -0.0130(8) 0.4870(8) 0.4864(23) 0.5
O5 O2− 8 d 0.3056(6) 0.4053(8) 0.4836(20) 0.5
O6 O2− 8 d 0.2847(7) 0.4458(9) 0.4842(25) 0.5

Table 4: Selected structure data for SBN

A short inspection of Table 4 shows that there are general Wyckoff sites of type (d) which are
occupied by all kind of elements except Ba. Only one Sr has site symmetry 4, the rest is placed
in mirror planes. Taking this data the following Figure 18 can be prepared using commercial
programmes or asking Mr. Betzler for help. The main motif is the Nb-O6 octahedra. Two kind of
such crystallographically independent octahedra exist and are linked via oxygen corners forming a
2 T OOLS TO DESCRIBE CRYSTALS 27

Figure 18: Illustration of the structure of SBN by a projection onto the c-plane using realistic
fractions for the positions of the atoms.

three-dimensional network. As a result three kinds of empty channels (or cavities) along the polar
axis c appear. The 12-coordinated cavity with 4-fold symmetry is either empty or occupied by Sr.
The 15-fold-coordinated one has a pentagon shape and is again empty or occupied by Ba or Sr.
The small 9-fold-coordinated cavity is always empty. Comparing Figure 17 and Figure 18 allows
to become more familiar with the structure of SBN. For example, the mirror plane belonging to the
symmetry operation (7), e. g., the solid line in the upper left of Figure 17 is easily recognized in the
real crystal. Even the glide operation (8, dashed diagonal line in Figure 17) can be demonstrated
by taking the projected Sr, Ba atoms at the lower left corner of the unit cell, reflect them at the
diagonal mirror plane, applying a fractional translation ( 12 , 12 , 0) to come to the position at the
upper right of the unit cell, with a change of the handedness. Further details should be explored
by the reader.

References

[1] G. Burns, A. M. Glazer. Space Groups for Solid State Scientists. Academic Press, 2nd ed.
edition, 1990.
28 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

[2] C. J. Bradley, A. P. Cracknell. The Mathematical Theory of Symmetry in Solids. Clarendon,

Oxford, 1972.

[3] T. Hahn. International Tables for Crystallography. Kluwer Acad. Publishers, Dordrecht, 2.
edition, 1989.
3 S YMMETRY INDUCED TENSOR PROPERTIES 29

3 Symmetry induced tensor properties

A property tensor may exhibit intrinsic symmetry, often expressed by a permissible change of
two or more indices. Such behavior follows from the physical property and very general relations
introduced by Onsager or Kleinman, for the last see section 4.6.

3.1 Neumann’s principle

In contrast to the intrinsic symmetry the crystal symmetry governs tensor properties, too. This is
based on Neumann’s well-known principle:

If a system has a certain group of symmetry operations, G, then any physical

observable of that system must also possess these symmetry operations.

An alternative formulation is given in the book by Burns and Glazer [1] by emphasizing that
the macroscopic (tensor) properties of a crystal have at least the symmetry of the point group
of the space group. The physical reason for this is that in measuring a macroscopic property
one would not expect to be able to detect the effect of a translation that is only a fraction of a
primitive unit cell. On the other hand, the rotational part of the symmetry operation will relate
points within the crystal that are separated by macroscopic distances and hence will have an effect
on the macroscopic properties. Somewhat special is the optical activity, because such a property
adds in phase from one unit cell to the next to result in an overall macroscopic effect.
Note the use of the words at least in the statement of Neumann’s principle. It is important to
understand that the point group of the space group is generally of lower symmetry than the sym-
metry of the physical properties. For example, if we consider properties specified by second-rank
tensors, we find that cubic crystals (those having cubic point groups) are isotropic, i.e., the diag-
onal elements of the tensor are equal and nonzero, while the nondiagonal elements are zero. This
means that such properties are the same in all directions and therefore have spherical symmetry.
Clearly, all of the symmetry operations of the cubic point groups are contained by such spherical
symmetry, so that we can say that the cubic point groups are subgroups of a spherically symmetric
group. With properties that are described by tensors of higher rank, the situation is more involved
[1].
According to Neumann’s principle the tensor representing any physical property has to be invariant
with regard to all symmetry operations of the crystal. This condition reduces the number of the in-
dependent tensor components because of relationships between the tensor components. There are
different ways known to extract the information from such relations, one deals with solving a set
of equations with the tensor components as unknowns, the second uses a reducible representation
of the group, its reduction into irreducible representations and projection operator techniques.
30 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

3.2 Relationships for tensor components

A tensor of rank m is transformed according to

t0ijk...m = aip ajq . . . amu tpqr...u . (3.1)
where the aij direction cosines are the elements of the (aij ) matrix. The (aij ) matrix connects
the original and the new co-ordinates, and are identical in our case with the rij . The invariance
changes this general equation 3.1 to
tijk...m = rip rjq . . . rmu tpqr...u . (3.2)
where r is the matrix corresponding to a particular symmetry operation of the point group. The
last equation holds for a polar tensor, but has to be modified to
tijk...m = |rij |rip rjq . . . rmu tpqr...u . (3.3)
with |rij | denoting the determinant, whose value is (+1) if the transforming operation consists of
a pure rotation and (-1) if an inversion (as part of a mirror operation) is involved changing the
handedness. Note that |rij | is a number and not a transformation matrix.
It is not difficult to find out whether the tensor representing any physical property is polar or axial,
since this can be easily decided by the eq. (2.1) defining the physical property in question. If only
one of the physical quantities in (3.1) is axial (for example the magnetic field) then the property
tensor will be axial, in every other case the tensor is polar.
An immediate consequence of equations (3.3) and (3.2) is that axial tensors of even rank (optical
activity) and polar tensors of odd rank (pyroelectricity, second harmonic generation) must vanish
identically in centrosymmetrical crystals with the inversion as a single symmetry operation. This
forces |rij | to be (-1) and thus tijk...m = −tijk...m for axial tensors. If the rank p is odd, then polar
tensors behave like (−1)p = −1, while axial ones like (−1)p ∗ |rij | = 1. For an even rank we find
for polar tensors (−1)p = 1, but (−1)p ∗ |rij | = 1 ∗ (1−) = −1 for axial tensors.
Application of all symmetry operations to both (3.2) and (3.3) produce a set of |g| equations, with
|g| being the order of the group. Of course not all of these equations are independent, because
we know that a group is described by a set of few generators. These are sufficient to deduce rela-
tionships, but sometimes equations obtained from additional operations may make the derivation
easier. Such approaches have been used to calculate the properties for all tensors up to rank four
for nonmagnetic and magnetic groups (not covered here), for details see the books by [2, 3, 4] and
have a look to section 4.7.

3.3 Direct group theoretical approach

A much more direct approach uses group theoretical methods, but one has to apply all symmetry
operations (not only the generators) to form a representation of the tensor. We rewrite Eq. (3.2)
XX X
Ug Ti1 i2 ...im = ... Dj1 i1 (g) Dj2 i2 (g) . . . Djm im (g)Tj1 j2 ...jm = Ti1 i2 ...im ∀ g ∈ G
j1 j2 jm
(3.4)
3 S YMMETRY INDUCED TENSOR PROPERTIES 31

Ug is the operator which transforms the tensor Ti1 i2 ...im according to the point group symmetry
element g. The direct product of the m three-dimensional representations Djk ik (g) forms the 3m -
dimensional tensor representation DT (g). Like in section 3.2 we can distinguish between polar
representations D(p) = D− and axial D(a) = D+ ones. An example for a polar vector, a polar
tensor of rank 1 is r, thus D− is a representation of the coordinates, while D+ is that of the angular
momentum.
The invariance condition (3.4) means that the independent tensor components transform accord-
ing to the totally symmetric (trivial) irreducible representation. To calculate the number of such
independent components one has to reduce the reducible representation DT (g) into its irreducible
ones and and count how often the trivial irreducible representation shows up. This number is given
by
1 X
Z = mα=1 = 1 · χT (g) (3.5)
|g|
g∈G

where χT (g) denotes the character of the reducible representation of the symmetry element g
and the factor 1 represents the character of the trivial repesentation. Then with the help of the
projection operator
1 X
P (α=1) Ti1 i2 ...im = Ug Ti1 i2 ...im = Ti1 i2 ...im (3.6)
|g|
g∈G

the independent tensor components can be calculated. The action of Ug on Ti1 i2 ...im is described
by the corresponding transformation of the coordinates, the rij .
For an example we consider the point group 4mm (C4v ), which has the following symmetry ele-
ments, if we take the principal axis along z: 1, 4, 43 , 42 ≡ 2, σv ≡ 2̄[100], σv0 ≡ 2̄[010], σd ≡
2̄[110], σd0 ≡ 2̄[11̄0]. From the corresponding matrices in appendix A and in Table 5 we im-
mediately construct the representation for a polar vector, because this is identical with the set of
matrices. To calculate Z for a polar vector (see Eq. 3.5) we need the characters of this representa-
tion, they are given in Table 5, too. We obtain
1 8
Z = (1 · 3 + 1 · 1 + 1 · 1 + 1 · (−1) + 1 · 1 + 1 · 1 + 1 · 1 + 1 · 1) = = 1
8 8
which means that a polar tensor of rank 1 exists in the point group C4v . But this is not new for us,
because we know that SBN is a ferroelectric. Using the projection operator P A1 we can calculate
which component transforms like the trivial representation A1 . Eq. 3.6 reduces to
1X
P A1 Ti = Ug Ti
8
g∈G

with i = x, y, z. The action of Ug on Ti can be taken from the corresponding matrices in Table 5.
For g = C4 we read x → −y, y → x, z → z and for g = σd we read x → −y, y → −x, z → z,
respectively. Applying P A1 to all three components results in
1
P A1 x = (x + (−y) + y + (−x) + (−x) + x + (−y) + y) = 0
8
32 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

C4v E C4 C43 C2
1 4 43 2
       
(x, y, z) 1 0 0 0 −1 0 0 1 0 −1 0 0
or  0 1 0   1 0 0   −1 0 0   0 −1 0 
       

(Rx , Ry , Rz ) 0 0 1 0 0 1 0 0 1 0 0 1
χ(p) (g) 3 1 1 -1
χ(a) (g) 3 1 1 -1
χ(p) (g 2 ) 3 -1 -1 3
χ(a) (g 2 ) 3 -1 -1 3

C4v σv σv0 σd σd0

2̄[100] 2̄[010] 2̄[110] 2̄[11̄0]
       
(x, y, z) −1 0 0 1 0 0 0 −1 0 0 1 0
or  0 1 0   0 −1 0   −1 0 0   1 0 0 
       

(−Rx , −Ry , −Rz ) 0 0 1 0 0 1 0 0 1 0 0 1

χ(p) (g) 1 1 1 1
χ(a) (g) -1 -1 -1 -1
χ(p) (g 2 ) 3 3 3 3
χ(a) (g 2 ) 3 3 3 3

Table 5: Group theoretical information for the point group C4v . The first column denotes the
transformation properties of the components of polar (x, y, z) and axial (Rx , Ry , Rz ) vectors or
the characters of the corresponding representation.

1
P A1 y = (y + x + (−x) + (−y) + y + (−y) + (−x) + x) = 0
8
1
P A1 z = (z + z + z + z + z + z + z + z) = z,
8
which means that the spontaneous polarization will be aligned along the c-axis. If we calculate Z
for an axial vector, using χa , we obtain
1
Z = (1 · 3 + 1 · 1 + 1 · 1 + 1 · (−1) + 1 · (−1) + 1 · (−1) + 1 · (−1) + 1 · (−1)) = 0,
8
thus no spontaneous magnetization can be expected in crystals with point group C4v . At a first
glance this looks surprising, but if we remember that the magnetization originates from ring cur-
rents, then we immediately understand this results, because mirror planes like σ reverse the current
and thus no magnetization can occur.
As a next example we discuss a symmetric tensor of rank 2 with both components transforming
according to polar vectors. (electrical and thermal conductivity, dielectric constant). Here the
3 S YMMETRY INDUCED TENSOR PROPERTIES 33

tensor representation to be reduced is not D(p) (g) ⊗ D(p) (g) = DT (g) with g ∈ G, but
h i
T
D+ (g) = D(p) (g) ⊗ D(p) (g) g∈G (3.7)
+

where the index + denotes the symmetric product representation, because the components i, j are
interchangeable. The number Z for such a symmetric representation follows from

2. rank 1 X 1 h (p) 2 (p) 2

i
Zsym = (χ (g)) + χ (g ) . (3.8)
|g| 2
g∈G

For our favorite point group C4v we obtain with the help of Table 5

2. rank 11 2
Zsym = [(3 + 3) + (12 + (−1)) + (12 + (−1)) + ((−1)2 + 3) +
82
+ ((−1)2 + 3) + (12 + 3) + (12 + 3) + (12 + 3)] = 2,

we expect two independent tensor components. Does this mean that two components are nonzero
and the others are zero? The answer is no! There are two independent components, but the number
of nonzero components can be calculated only with the projection operator!
With
1 1
P A1 xx = (x2 + y 2 + y 2 + x2 + x2 + x2 + y 2 + y 2 ) = (x2 + y 2 )
8 2
and
P A1 yy = P A1 xx but P A1 zz = z 2
we find T11 = T22 6= T33 and P A1 , ij = 0 with i, j = x, y, z; i 6= j, a result well-known to most
of us.
The extension to a third rank tensor is straight forward. The piezoelectric tensor dk,ij has three
components each one transforming as a polar vector, but the last two are interchangeable. Thus
Eq. 3.7 reads now
h i
DT (g) = D(p) (g) ⊗ D(p) (g) ⊗ D(p) (g) g∈G (3.9)
+

Notice that the representations of those components which are commutable have been written after
the non-commutable ones. This will help us later when we project the nonzero components from
introducing a commutability which is not valid in general. The number of independent components
follows from
1 X (p) 1 h (p) i
Z 3. rank = χ (g) (χ (g))2 + χ(p) (g 2 ) . (3.10)
|g| 2 +
g∈G

and for C4v symmetry we calculate

11
Z 3. rank = [3(9 + 3) + 1(1 − 1) + 1(1 − 1) + (−1)(1 + 3) +
82
+ 1(1 + 3) + 1(1 + 3) + 1(1 + 3) + 1(1 + 3)] = 3
34 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

Now we start a tiresomely lengthy work, because up to 33 = 27 times we have to apply the
projection operator to a component, with 8 terms in each sum, so a Mathematica notebook would
be a nice tool, but unfortunately no student has presented such one to the author. Lets hope for the
future! As pointed out above, we have to obey the order of the components which do not commute,
therefore we write the first with capital letters and the second and third, which commute with small
letters. For the notation we use both the tensor notation and the reduced matrix notation, for details
see section 4.8.

1
dXxx = d11 = P A1 Xxx = (Xx2 − Y y 2 + Y y 2 − Xx2 − Xx2 + Xx2 − Y y 2 + Y y 2 ) = 0
8
1
dXyy = d12 = P A1 Xyy = (Xy 2 − Y x2 + Y x2 − Xy 2 − Xy 2 + Xy 2 − Y x2 + Y x2 ) = 0
8
1
dXzz = d13 = P A1 Xzz = (Xz 2 − Y z 2 + Y z 2 − Xz 2 − Xz 2 + Xz 2 − Y z 2 + Y z 2 ) = 0
8
1
dXyz = d14 = P A1 Xyz = (Xyz − Y xz − Y xz + Xyz − Xyz − Xyz + Y xz + Y xz) = 0
8
dXzy = d14 =0
1
dXxz = d15 = P A1 Xxz = (Xxz + Y yz + Y yz + Xxz + Xxz + Xxz + Y yz + Y yz)
8
1
= (Xxz + Y yz)
2
1
dXzx = d15 = (Xzx + Y zy) because i, j commute
2
1
dXxy = d16 = P A1 Xxy = (Xxy + Y yx − Y yx − Xxy + Xxy − Xxy − Y yx + Y yx) = 0
8
dXyx = d16 = 0
1
dY xx = d21 = P A1 Y xx = (Y x2 + Xy 2 − Xy 2 − Y x2 + Y x2 − Y x2 − Xy 2 + Xy 2 ) = 0
8
1
dY yy = d22 = P A1 Y yy = (Y y 2 + Xx2 − Xx2 − Y y 2 + Y y 2 − Y y 2 − Xx2 + Xx2 ) = 0
8
1
dY zz = d23 = P A1 Y zz = (Y z 2 − Xz 2 + Xz 2 − Y z 2 + Y z 2 − Y z 2 − Xz 2 + Xz 2 ) = 0
8
1
dY yz = d24 = P A1 Y yz = (Y yz + Xxz + Xxz + Y yz + Y yz + Y yz + Xxz + Xxz)
8
1
= (Y yz + Xxz) = d15
2
dY zy = dY yz = d24
1
dY xz = d25 = P A1 Y xz = (Y xz − Xyz − Xyz + Y xz − Y xz − Y xz + Xyz + Xyz) = 0
8
dY zx = dY xz = d25 = 0
1
dY xy = d26 = P A1 Y xy = (Y xy − Xyx + Xyx − Y xy − Y xy + Y xy − Xyx + Xyx) = 0
8
dY yx = dY xy = d26 = 0
1
dZxx = d31 = P A1 Zxx = (Zx2 + Zy 2 + Zy 2 + Zx2 + Zx2 + Zy 2 + Zy 2 )
8
1 2 2
= Z(x + y )
2
3 S YMMETRY INDUCED TENSOR PROPERTIES 35

1
dZyy = d32 = P A1 Zyy = (Zy 2 + Zx2 + Zx2 + Zy 2 + Zy 2 Zx2 + Zx2 + Zx2 )
8
1
= Z(x2 + y 2 ) = d31
2
1
dZzz = d33 = P A1 Zzz = (Zz 2 + Zz 2 + Zz 2 + Zz 2 + Zz 2 + Zz 2 + Zz 2 + Zz 2 ) = Zz 2
8
1
dZyz = d34 = P A1 Zyz = (Zyz + Zxz − Zxz − Zyz + Zyz − Zyz − Zxz + Zxz) = 0
8
dZzy = dZyz = d34 = 0
1
dZxz = d35 = P A1 Zxz = (Zxz − Zyz + Zyz − Zxz − Zxz + Zxz − Zyz + Zyz) = 0
8
dZzx = dZxz = d35 = 0
1
dZxy = d36 = P A1 Zxy = (Zxy − Zyx − Zyx + Zxy − Zxy − Zxy + Zyx + Zyx) = 0
8
dZyx = dZxy = d36 = 0

In matrix notation our calculations result in

 
0 0 0 0 d15 0 d15 = 21 (Xxz + Y yz)
 0 0 0 d15 0 0  with d31 = 12 Z(x2 + y 2 ) (3.11)
 

d31 d31 d33 0 0 0 d33 = Zz 2 .

As predicted by Eqn. 3.10 we obtained three independent components but five nonzero compo-
nents. If we consider instead of the piezoelectric tensor that of the second harmonic, then due to
Kleinman symmetry the elements d15 = d31 are equal, because we do not distinguish between
capital and small letters, in agreement with the results in section 4.7.
Finally we comment shortly on polar 4. rank tensors (electrostriction, piezooptic effect, quadratic
electrooptic effect). Commutability like ij, kl = ji, kl = ij, lk = ji, lk yields
h i h i
DT (g) = D(p) (g) ⊗ D(p) (g) ⊗ D(p) (g) ⊗ D(p) (g) g∈G (3.12)
+ +

with

1 X
 h
1 i 2
(p) 2 (p) 2
Z = (χ (g)) + χ (g ) (3.13)
|g| 2
g∈G
1 X 1 h (p) i
= (χ (g))4 + 2(χ(p) (g))2 χ(p) (g 2 ) + χ(p) (g 2 ) . (3.14)
|g| 4
g∈G

If we assume permutation with respect to index pairs ([ij], [kl] = [kl], [ij]) the representation
reads
    
T (p) (p) (p) (p)
D (g) = D (g) ⊗ D (g) ⊗ D (g) ⊗ D (g) g∈G (3.15)
+ + +
36 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

and
 
1 X 1  (p⊗p) 2 (p⊗p) 2
Z = χ (g) + χ (g ) (3.16)
|g| 2
g∈G
( )
1  (p) 2 1 (p) 2 2
 
1 X1 1  (p) 2 2
= χ (g) + χ (g ) + χ (g ) + χ(p) (g 4 ) (3.17)
|g| 2 2 2 2
g∈G
 
1 X 1  (p) 4 
(p)
2
(p) 2

(p) 2
2
(p) 4
= χ (g) + 2 χ (g) χ (g ) + 3 χ (g ) + 2χ (g ) (3.18) .
|g| 8
g∈G

So far we have not discussed axial tensors of higher rank. There is nothing special with them, ex-
amples are given in [5] (in German), together with Tables of independent polar tensor components
for all point groups. Even more extended Tables are presented in the famous book by Birss [3],
where magnetic groups with i-tensors (invariant with respect to time reversal, electric polarization)
and c-tensors ( change sign tensors with respect to time reversal, magnetization) are discussed.

References

[1] G. Burns, A. M. Glazer. Space Groups for Solid State Scientists. Academic Press, 2nd ed.
edition, 1990.

[2] J. F. Nye. Physical Properties of Crystals. Oxford UP, 1956.

[3] R. Birss. Symmetry and Magnetism. North-Holland, Amsterdam, second edition, 1966.

[4] W. A. Wooster. Tensors and group theory for the physical properties of crystals. Clarendon
Pr., Oxford, 1973.

[5] M. Böhm. Symmetrien in Festkörpern. Wiley-VCH, Berlin, 2002.

4 N ONLINEAR O PTICAL S USCEPTIBILITIES 37

4 Nonlinear Optical Susceptibilities

All electromagnetic phenomena are governed by the Maxwell equations for the electric and mag-
netic fields E, D, B, and H:

∇D = ρ (4.1)

∇B = 0 (4.2)
∂B
∇×E = − (4.3)
∂t
∂D
∇×H = j+ (4.4)
∂t
where ρ and j are electric charge and current.
The fields are related by
D = 0 E and B = µµ0 H (4.5)
where  and µ, the dielectric constant and the relative permittivity, describe material properties. In
general, both are second rank tensors. 0 and µ0 are the permittivity and the permeability of the
vacuum, fundamental physical constants.
Approximations often introduced to simplify calculations include

 µ=1 for non-magnetic materials,

  = 1, µ = 1 for vacuum,
 j=0 for isolators,
 ρ=0 for vanishing charges,
∂
 ∂t = 0 for the static case.

A more detailed overview is given in the lecture notes on Linear Response Theory by P. Hertel [1]
and in numerous textbooks in the field.
In the linear case, the polarization P may be written in a simple form
Z ∞
P(r, t) = 0 χ(1) (r − r0 , t − t0 ) · E(r0 , t0 )dr0 dt0 (4.6)
−∞

where χ(1) is the linear susceptibility of the medium. Usually monochromatic plane waves are
assumed, E(k, ω) = E(k, ω) exp(ik · r − iωt), then a Fourier transformation applied to Eq. 4.6
yields
P(k, ω) = 0 χ(1) (k, ω)E(k, ω) (4.7)
with Z ∞
χ(1) (k, ω) = χ(1) (r, t) exp(−ikr + iωt)drdt . (4.8)
−∞
The dependence of χ on k is only weak, in nearly all practical cases it can be neglected.
38 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

4.1 Nonlinear Polarization

In the nonlinear case, P can be expanded into a power series of E – at least as long as E is
sufficiently weak

Z ∞
P(r, t) = 0 χ(1) (r − r0 , t − t0 ) · E(r0 , t0 )dr0 dt0
−∞
Z ∞
+0 χ(2) (r − r1 , t − t1 ; r − r2 , t − t2 )E(r1 , t1 )E(r2 , t2 )dr1 dt1 dr2 dt2
−∞
Z ∞
(4.9)
(3)
+0 χ (r − r1 , t − t1 ; r − r2 , t − t2 ; r − r3 , t − t3 )E(r1 , t1 )
−∞

× E(r2 , t2 )E(r3 , t3 )dr1 dt1 dr2 dt2 dr3 dt3

+ ...

where χ(n) is the nth-order nonlinear susceptibility. As in the linear case, the problem can be
Fourier transformed. Yet, for E now a sum of monochromatic plane waves should be assumed
X
E(r, t) = E(ki , ωi ) , (4.10)
i

yielding for the polarization

P(k, ω) = P(1) (k, ω) + P(2) (k, ω) + P(3) (k, ω) + . . . (4.11)

with

P(1) (k, ω) = 0 χ(1) (k, ω) · E(k, ω) ,

P(2) (k, ω) = 0 χ(2) (k = ki + kj , ω = ωi + ωj )E(ki , ωi )E(kj , ωj ) ,

(4.12)
P(3) (k, ω) = 0 χ(3) (k = ki + kj + kl , ω = ωi + ωj + ωl )

E(ki , ωi )E(kj , ωj )E(kl , ωl ) .

The χ(n) (k, ω) can be expressed in a similar way as in the linear case as integrals over the respec-
tive χ(n) (r, t). Again, the dependence on k can be neglected.
χ(n) is an (n + 1)st-rank tensor representing material properties. Using Einstein’s summation
convention, the above equations may be rewritten in component form, e. g.

(2) (2)
Pk (ω) = 0 χkmn (ω = ωi + ωj )Em (ωi )En (ωj ) . (4.13)
4 N ONLINEAR O PTICAL S USCEPTIBILITIES 39

4.2 The Phase-Matching Problem

We have arrived now at the nonlinear polarization of a medium. The fundamental waves generate
an oscillating polarization through the medium which oscillates with ω. The phases at different
locations are defined and connected by the fundamental waves travelling through the medium.
That means that the polarization wave travels through the medium at a velocity v(ωi , ωj ) for the
fundamental frequencies ωi , ωj .
The local polarization at every location acts as a source of electromagnetic dipole radiation. The
generated free waves, yet, travel through the medium at a velocity v(ω) characteristic for their
own frequency ω.
The velocities are defined by the respective refractive indices and – due to the dispersion present
in all materials – generally are different. In an extended medium the two relevant waves – polar-
ization wave and generated free wave – thus come out of phase after a typical distance commonly
referred to as coherence length. The sum free wave is amplified due to constructive interference up
to this coherence length, then attenuated due to destructive interference. No efficient generation
of nonlinear radiation seems to be possible. Yet, there are some solutions to the problem.

4.3 Mechanisms for the Nonlinear Polarization

As for the linear polarization in matter, various mechanisms are responsible for the nonlinear po-
larization, too. Depending on the frequencies of the applied fields and of the resulting nonlinear
polarizations the possible mechanisms may contribute more or less. At comparably low electro-
magnetic fields all of these mechanisms (excepts for the last one) can be regarded as being strictly
linear, nonlinearities show up when the fields are increased.

Electronic polarization: The distortion of the outer-shell electronic cloud of atoms, ions, and
molecules, respectively, in gases, liquids, or solids, compared to the undisturbed state. This
mechanism has very fast response time (< 10−15 s). Most optical frequency mixing effects
such as second harmonic and third harmonic generation, sum-frequency mixing, optical
parametric oscillation, four-photon parametric interaction use this mechanism.

Ionic polarization: The contribution from an optical-field induced relative motion (vibration, ro-
tation in molecules, optical phonons in solids) between nuclei or ions. The response time
of this mechanism is around 10−12 seconds. Examples: Raman resonance-enhanced four-
wave-mixing effects, Raman enhanced refractive index change.

Molecular reorientation: It denotes the additional electric polarization contribution from an op-
tical-field induced reorientation of anisotropic molecules in a liquid. The response time
of this process is dependent on the rotational viscosity of molecules in the liquid and is
approximately 10−12 –10−13 seconds. Examples: Stimulated Kerr scattering, Kerr-effect
related refractive index change.
40 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

Induced acoustic motion: It is the polarization contribution from an optical induced acoustic
motion related to the so-called electrostriction interaction. The response time of this mech-
anism is around 10−9 –10−10 seconds depending on the medium. Examples: Brillouin scat-
tering, self focusing, optical breakdown.

Induced population change: The contribution of electrons to the polarization depends on their
eigenstates. Their populations are changed by one-photon or two-photon absorption and by
other resonant interactions (e. g. in Raman processes). The response time strongly depends
on the respective electronic transition, but is in general slower than in the above discussed
mechanisms. Examples are all resonance-enhanced nonlinear processes.

Spatial redistribution of electrons: Excited charge carriers in solids – electrons or holes – can
be spatially redistributed due to a spatially modulated light pattern. This is a major effect
in all so-called photorefractive materials. The response time depends on the mobility of the
carriers and on the internal electric field, in general it is slow compared to the response times
discussed up to here. Examples are all processes which can be summarized under the term
Photorefractive Nonlinearity.

Spatial redistribution of ions: There are some materials where not electrons but – also or instead
– ions are redistributed by a spatially modulated light pattern. Of course this effect again is
considerably slower. It is only of minor importance within the photorefractive materials.

4.4 The Anharmonic Oscillator as a Qualitative Model

As a crudely qualitative but nevertheless vivid model for the nonlinear polarization one can use the
classical anharmonic oscillator. Physically, the oscillator describes an electron bound to a core or
an infrared-active molecular vibration. The potential may exhibit anharmonicities of odd or even
symmetry as sketched in Fig. 19.

V(x) V(x) V(x)

x x x

Figure 19: Potential forms for the anharmonic oscillator. Left: harmonic potential Vh (x) = a2 x2 ,
middle: odd-symmetric anharmonicity Vo (x) = Vh (x) + 3b x3 , right: even-symmetric anharmonic-
ity Ve (x) = Vh (x) + 4c x4 . The dashed curve denotes the respective harmonic part.

The equation of motion for the oscillator in the presence of a driving force F can be written as

d2 x dx
2
+γ + ax + bx2 + cx3 = F . (4.14)
dt dt
4 N ONLINEAR O PTICAL S USCEPTIBILITIES 41

For the harmonic case b = c = 0, for an odd-symmetric anharmonicity b 6= 0, for an even-

symmetric c 6= 0. Both b and c are assumed to be small so that they can be treated as perturbations.
As driving force we consider an applied electric field with Fourier components at the frequencies
±ω1 and ±ω2
q 
E1 e−iω1 t + eiω1 t + E2 e−iω2 t + eiω2 t .



F = (4.15)
m
q and m are charge and mass of the oscillating particle (electron, ion, etc.).
When we neglect the anharmonic perturbations b and c, we get the first order solution x(1) for x
X (q/m)Ei
x(1) = x(1) (ωi ) , x(1) (ωi ) = e−iωi t (4.16)
i
ω02 − ωi2 − iωi γ

where ω02 = a.
For a density of N such classical anharmonic oscillators per unit volume the induced electric
polarization is simply
P = N qx . (4.17)

Higher order solutions are obtained by substituting lower order solutions for the nonlinear terms
in Eq. 4.14, e. g. bx(1) 2 for bx2 .
First we look at the second order solution in the presence of an odd-symmetric anharmonicity only
(b 6= 0, c = 0). Omitting the first order solution, we use −bx(1) 2 as driving force

d2 x dx
+γ + ax = −bx(1) 2 . (4.18)
dt2 dt

−bx(1) 2 introduces terms with frequencies 2ωi , ωi + ωj , ωi − ωj , ωi − ωi = 0. Thus we have

included second-harmonic generation, sum-frequency and difference-frequency generation, and
optical rectification. A typical solution (here for second harmonic generation) is of the form

−b(q/m)2 Ei2
x(2) (2ωi ) = e−i2ωi t . (4.19)
(ω02 − ωi2 − iωi γ)2 (ω02 − 4ωi2 − i2ωi γ)

Second we assume that only an even-symmetric anharmonicity is present which means that b = 0,
c 6= 0. We now have to use −cx(1) 3 as driving force

d2 x dx
2
+γ + ax = −cx(1) 3 . (4.20)
dt dt
Obviously the driving force now introduces only terms with an odd number of ωs, e. g. 3ωi ,
2ωi − ωi = ωi , ωi + ωj + ωk , ωi + ωj − ωk . Thus third-harmonic generation, nonlinear refraction
and similar effects are described. Even-symmetric anharmonicities are present in all types of
materials, even in isotropic ones like liquids and gases. From the above we can conclude that such
materials are only suited for odd-harmonic generation and other odd-order effects.
42 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

From Eqs. 4.16, 4.19, and 4.17 we can roughly estimate the ratio between linear and second order
nonlinear polarization. If we assume that we are far from any resonance, i. e. ω0  ωi , we find
for this ratio  
 P (2)   qbE 
 (1)  ≈   . (4.21)
 
P  mω 4 
0

For the limit that for a bound electron harmonic and anharmonic force, mω02 x and mbx2 , are of
the same order of magnitude, one can assume that both are of the order of magnitude of the total
binding force of the electron |qEat | (one can show that this is only valid for large anharmonicities
b)
|qEat | ≈ mω02 x ≈ mbx2 (4.22)

or, eliminating x,
mω04
|qEat | ≈ . (4.23)
b
Eq. 4.21 then becomes
P (2) /P (1) ≈ E/Eat (4.24)

and for the susceptibilities

χ(2) /χ(1) ≈ 1/Eat . (4.25)

This can be generalized to

P (n+1) /P (n) ≈ E/Eat and χ(n+1) /χ(n) ≈ 1/Eat . (4.26)

The inner-atomic fields Eat are in the order of 3 × 1010 V/m [2], thus with χ(1) ≈ 3 we arrive
at 10−10 m/V for the second order nonlinear susceptibility. Some typical measured values are
listed in Table 7.1 of Ref. [2]. They range from approximately 10−12 m/V for materials with low
(2)
anharmonicities (Quartz: χxxx = 0.8 × 10−12 m/V) up to 10−10 m/V for typical nonlinear optical
(2)
materials (LiNbO3 : χzzz = 0.8 × 10−10 m/V).

4.5 Structural Symmetry of Nonlinear Susceptibilities

The susceptibility tensors must remain unchanged upon symmetry operations allowed for the
medium. This reduces the number of independent and nonzero elements. The most important
conclusion from this property is that for all centrosymmetric crystals and for all isotropic me-
dia (gases, liquids, amorphous solids) all tensor elements of the even-order susceptibility tensors
(χ(2) , χ(4) , . . . ) must be zero. This has been already shown qualitatively for the model of the
anharmonic oscillator in section 4.4. Thus, e. g., no second harmonic generation can be observed
in such media. Odd-order susceptibility tensors, yet, will be non-zero and will provide nonlinear
effects. Using gases or metal vapors, e. g., only odd-order harmonics can be produced.
4 N ONLINEAR O PTICAL S USCEPTIBILITIES 43

4.6 Permutation Symmetry of Nonlinear Susceptibilities

When tensors are multiplied with vectors, usually the order of the vector multiplication can be
changed. In nonlinear optics it should not matter which of the fundamental fields is the first to
be multiplied. From this, permutation symmetry for the nonlinear susceptibilities follows, for the
second order
(2) (2)
χijk (ω1 , ω2 ) = χikj (ω2 , ω1 ) , (4.27)
or for the third order susceptibility
(3) (3) (3) (3)
χijkl (ω1 , ω2 , ω3 ) = χiklj (ω2 , ω3 , ω1 ) = χiljk (ω3 , ω1 , ω2 ) = χijlk (ω1 , ω3 , ω2 ) = . . . (4.28)

Besides this trivial one, a more general permutation symmetry can be defined due to time reversal
symmetry resulting in relations like
(2) (2) (2)
χijk ∗ (ω = ω1 + ω2 ) = χjki (ω1 = −ω2 + ω) = χkij (ω2 = ω − ω1 ) . (4.29)

Time reversal symmetry can be applied as long as absorption can be neglected.

If the dispersion of χ can also be neglected, then the permutation symmetry becomes independent
of the frequencies. Consequently, then a very general permutation symmetry exists between dif-
ferent elements of χ: elements remain unchanged under all permutations of the Cartesian indices.
This so-called Kleinman’s conjecture or Kleinman symmetry [3] reduces the number of indepen-
dent elements further. Yet, it should be noted that it’s a good approximation only at frequencies
far from resonances such that dispersion really can be neglected.

4.7 Example: Strontium Barium Niobate

Strontium Barium Niobate is a crystal which is in a ferroelectric phase at room temperature, its
point symmetry group is 4mm. The symmetry operations present in the point group include
     
x → y x → −x x → −y
4:  y → −x   y → −y   y → x 
     

z → z z → z z → z

   
x → −x x → x
m1 :  y → y   y → −y  (4.30)
   

z → z z → z

   
x → y x → −y
m2 :  y → x   y → −x  .
   

z → z z → z
44 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

The tensor elements transform like products of the respective coordinates, they must remain un-
changed under all the transformations listed. The mirror plane m1 changes x into −x or y into
−y, thus all elements with an odd number of indices 1 or an odd number of indices 2 have to be
zero. The mirror plane m2 transform x to y and y to x, thus elements where 1s are replaced by 2s
have to be equal.
For the second order susceptibility tensor for second harmonic generation, e. g., we arrive at the
nonzero elements

χ311 = χ322 , χ333 , χ131 = χ113 = χ232 = χ223 . (4.31)

All other elements must be zero. Kleinman symmetry further reduces the number of independent
elements to two (χ311 and equivalent, and χ333 ).

4.8 Contraction of Indices

Especially for the susceptibility tensor for second harmonic generation it is common to write it
in a different form. As the last two indices can be exchanged, there are 18 different elements left
from the full set of 27. These 18 are written as a 2-dimensional matrix dij , the last two indices kl
of the elements χikl are contracted to one index j such that

11 → 1 , 22 → 2 , 33 → 3 , 23, 32 → 4 , 31, 13 → 5 , 12, 21 → 6 .

(4.32)
Using this matrix form of the susceptibility tensor, the second harmonic polarization is written as
 
Ex2
  Ey2 
 
  
Px d11 d12 d13 d14 d15 d16 
2

  Ez 
 
 Py  = 0  d21 d22 d23 d24 d25 d26  ×  . (4.33)
  

 2Ey Ez 
Pz d31 d32 d33 d34 d35 d36 
 2E E 

 z x 
2Ex Ey

References

[1] P. Hertel. Linear Response Theory. University of Osnabrück, 2001.

http://www.physik.uni-osnabrueck.de/virtual-campus.html.

[2] Y. R. Shen. The Principles of Nonlinear Optics. John Wiley & Sons, Inc., 1984.

[3] D. A. Kleinman. Nonlinear Dielectric Polarization in Optical Media. Phys. Rev. 126, 1977
(1962).
5 H ARMONIC G ENERATION 45

5 Harmonic Generation

One of the most important nonlinear optical processes for technical applications is the generation
of harmonics from laser light. We will discuss here second-harmonic generation, widely used for
producing visible and near ultraviolet coherent light, and the generation of higher harmonics in
gases, used for EUV (extreme ultraviolet) light sources.

5.1 Second-Harmonic Generation

Second-harmonic generation (SHG) was the first experiment in the history of nonlinear optics
carried out by Franken et al. [1] soon after the invention of the Ruby laser [2]. Presently it is one
of the main applications of nonlinear optics, maybe the only really important one. In the preceding
chapter we already discussed some important points concerning the nonlinear susceptibility. The
general symmetry arguments have to be adopted in a suitable way for SHG. The responsible tensor
is of third rank, materials for SHG thus must be non-centrosymmetric. For practical reasons,
usually the d-tensor described is used instead of the more general χ-tensor. Because of a different
definition, most authors use the convention d = χ/2 for the tensor elements.
The local second harmonic polarization can be calculated according to Eq. 4.12. For the generated
second-harmonic intensity, yet, we face the phase-matching problem shortly discussed. Fig. 20
visualizes the principle.

E(2) E(2)
(2) (2)
E E
(2) (2)
E E

P(2) P(2)
(1) (1)
E E
x1 x2 x3 x1 x2 x3

Figure 20: Fundamental wave E (1) , induced second-harmonic polarization P (2) , and second-
harmonic waves E (2) , generated at the positions x1 , x2 , and x3 in a nonlinear material for two
different cases. Left: second-harmonic waves travel at the same velocity as the fundamental wave,
all are in-phase throughout. Right: different velocities, the usual case, mismatch between the
phases of the second-harmonic waves E (2) .

Due to dispersion present in all materials, waves of different frequencies travel at different veloc-
ities, yielding a phase-mismatch between second-harmonic waves generated at different positions
in a nonlinear material. To get the total second-harmonic intensity produced, we have to integrate
over the generated waves taking into account the different velocities. For simplicity we omit all
pre-factors and all rapidly oscillating factors and calculate only the phase-factors with respect to
46 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

x = 0. For E (1) (x) and P (2) (x) we can write

E (1) (x) = E (1) (0) · e−ik1 x , (5.1)

(2) (1) (1) (1) (1) −i2k1 x
P (x) = χE (x)E (x) = χE (0)E (0) · e . (5.2)

Taking P (2) as driving force in a wave equation for E (2) yields

E (2) (x) = K 0 · P (2) (x) = K · E (1) (0)E (1) (0) · e−i2k1 x (5.3)

where the K contains all necessary constants like nonlinear susceptibility or refractive indices.
E (2) now travels through the material with a velocity characteristic for the frequency ω2 = 2ω1
and wave vector k2 . Thus at an arbitrary position x0 where we could measure the second-harmonic
0 0
E (2) (x0 ) = E (2) (x) · e−ik2 (x −x) = K · E (1) (0)E (1) (0) · e−ik2 x e−i(2k1 −k2 )x . (5.4)

Assuming homogeneous material for 0 < x < L, we have to integrate

Z L
(2) −ik2 x0
Etotal (x0 ) = K ·E (1)
(0)E (1)
(0) · e e−i(2k1 −k2 )x dx
0

1 h i∆kL 0
i
= K · E (1) (0)E (1) (0) · e−ik2 x e −1
i∆k
0 ∆k 1 h i ∆k L ∆k
i
= K · E (1) (0)E (1) (0) · e−ik2 x ei 2 L e 2 − e−i 2 L
i∆k
0 ∆k sin(∆k L/2)
= K · E (1) (0)E (1) (0) · e−ik2 x ei 2
L
· (5.5)
∆k/2

with
2π 2π 4π
∆k = k2 − 2k1 = n(ω2 ) − 2 n(ω1 ) = (n(ω2 ) − n(ω1 )) . (5.6)
λ2 λ1 λ1
λ1 and λ2 = λ1 /2 are the wavelengths of the fundamental and second harmonic waves, respec-
tively, in vacuum.
Often a characteristic length, the so-called coherence length Lc , is defined. Yet one has to be
careful as two different definitions are used – the length after which the sine reaches its maximum
or the length after which the sine changes sign. Thus it may be defined as
π 2π
either Lc = or Lc = . (5.7)
∆k ∆k
The generated second-harmonic intensity depends mainly on the phase mismatch ∆k, and of
course on the square of the input intensity and the tensor elements involved. For the latter of-
ten a so-called effective tensor element is used which is a suitable combination for the geometry
considered
sin2 (∆k L/2)
I (2) = C · d2eff · I (1) 2 · . (5.8)
(∆k/2)2
5 H ARMONIC G ENERATION 47

If one is interested in calculating numerical results for I (2) , an appropriate constant C may be
adopted from textbooks on nonlinear optics.
As already discussed, due to dispersion, ∆k in Eq. 5.8 generally is non-zero, the intensity oscillates
in a sine-square way. If, however, ∆k approaches zero, we have to calculate the limit
sin(∆k L/2)
lim =L. (5.9)
∆k→0 ∆k/2
In this case, the second-harmonic intensity increases quadratically with L – at least as long as
we are in the limit of low second-harmonic intensities where I (1) is unchanged (undepleted fun-
damental wave approximation). The spatial variation of second-harmonic intensities for some
characteristic values ∆k are sketched in Fig: 21.
SHG Intensity [a.u.]

0 0.2 0.4 0.6 0.8 1

Position L in the Crystal (Relative to Size)

Figure 21: Second-harmonic intensities as a function of the position in the nonlinear material for
different ∆k.

5.2 Phase Matching

For an efficient generation of second-harmonic light it is highly desirable to achieve phase match-
ing, ∆k = 0. Usually the refractive indices are governed by normal dispersion which means that
in Eq. 5.6 the difference n(ω2 ) − n(ω1 ) is larger than zero, revealing ∆k > 0. One way out is
to utilize the birefringence which is present in crystals of all symmetry classes except the cubic
one. Uniaxial classes with two different principal refractive indices include the tetragonal, hexag-
onal and trigonal ones; biaxial classes, where all three principal indices are different, include the
orthorhombic, monoclinic and triclinic ones.
The refractive index of a material is derived from the linear susceptibility, a second rank tensor.
This tensor can be visualized by a general ellipsoid – general means that all three axes of the ellip-
soid are of different lengths and that the orientation is arbitrary. However, this ellipsoid has to be
48 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

compatible with the point symmetry of the material regarded. That means that certain symmetry
elements may fix the orientation of the ellipsoid and may force two or all three axes to be equal.
This reveals the above classification. In all uniaxial classes, the orientation of the ellipsoid is fixed,
and the ellipsoid is rotationally symmetric. In the biaxial classes where all three axes are different
in length, the orientation is fixed for orthorhombic crystals, one axis is fixed for monoclinic crys-
tals, and the orientation is completely free for triclinic ones. For the latter two cases, moreover,
the orientation is wavelength dependent.
The k-vector of light propagating in the material defines a plane perpendicular to it through the
center of the ellipsoid. This plane intersects the ellipsoid yielding an ellipse as intersection curve.
The directions of the major and minor axes of this ellipse define the two polarization directions
allowed, the length of these axes determine the respective refractive indices. These two different
indices for every crystallographic direction can be plotted as index surfaces which reveal the two
refractive indices as intersections with the respective k-vector direction.
This directional dependence of the refractive indices for the two cases – uniaxial and biaxial – is
schematically shown in Fig. 22. For every direction of the wave vector in an uniaxial or biaxial
crystal two different refractive indices are found which are valid for the two light polarizations
possible. The two refractive indices define the two possible velocities of light – a maximal and a

z z

ne
n
o

x y x y

Figure 22: Refractive index surfaces in an uniaxial crystal (left) and in a biaxial one (right). The
two surfaces indicate the refractive indices for the respective crystallographic directions.

minimal one – for every propagation direction. Two fixed polarization directions inside the crys-
tal, perpendicular to each other, are connected with the two refractive indices. There are obvious
distinct exceptions to this general rule of two different refractive indices. For the uniaxial case in
the left drawing light propagating along the crystallographic z-axis finds only one refractive index.
The same is valid in the biaxial case for light travelling along the direction denoted by the gray line
in the right drawing. For these special propagation directions arbitrary light polarizations are pos-
sible. These crystallographic directions are called the optic axes. There is one in uniaxial crystals
– the z-axis – and there are two in biaxial crystals – the gray line and its symmetry equivalent.
Utilizing the birefringence of a material, it may be possible to find propagation directions where
the velocities of fundamental and harmonic waves are identical. Drawing the index surfaces for
fundamental and harmonic frequencies, these directions are found as the intersection curves be-
5 H ARMONIC G ENERATION 49

tween the index surfaces. Fig. 23 shows this for an uniaxial material, one of the simplest cases.
(1)
The index surfaces for the ordinary index at the fundamental frequency, no , and for the extraor-
(2)
dinary index at the harmonic frequency, ne , are sketched, the intersection curve is a circle, all
propagation directions with a fixed angle Θ versus the z-axis are phase-matched.

z
Figure 23: Refractive index surfaces for the or-
(1)
dinary index at the fundamental frequency, no ,
and for the extraordinary index at the harmonic
(2)
frequency, ne in a uniaxial material with so-
Θ called negative birefringence (ne < no ). The
n(2) n(1) gray intersection curve (circular in the uniaxial
e o
case) determines the phase-match angle Θ.
x y

The idealized conditions sketched in Fig. 23, which enable phase matching, may be reality for
certain materials, yet they need not. To check whether phase matching is really possible, one has to
consider the dispersion behavior of the material. Typical dispersion curves for uniaxial crystals are
sketched in Fig. 24. A fundamental wavelength of 1000 nm, consequently a harmonic at 500 nm
are assumed. Low birefringence (left) inhibits phase matching, higher birefringence (right) allows
it. Or – to put it in other terms – every birefringent material has a certain restricted wavelength
range with a characteristic short-wavelength limit, in which phase-matching is possible.

2.45 2.45
2.40 2.40
Refractive Index

Refractive Index

2.35 2.35
2.30 2.30
2.25 n 2.25 no
o
2.20 ne 2.20
2.15 2.15 ne

400 600 800 1000 1200 400 600 800 1000 1200
Wavelength [nm] Wavelength [nm]

Figure 24: Dispersion of the refractive indices in uniaxial crystals. Left: low birefringence, right:
higher birefringence. The refractive index for the ordinary fundamental wave is fixed, the index
for the extraordinary harmonic wave can be angle-tuned along the vertical lines drawn.

The refractive index of the harmonic beam is defined as a function of the angle Θ as
1 cos2 Θ sin2 Θ
= + . (5.10)
n2e (Θ) n20 n2e
50 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

From Eq. 5.10, in turn the phase-matching angle Θ can be deduced demanding a value ne (Θ) at
the harmonic wavelength to be equal to no at the fundamental wavelength. A real solution for Θ
then indicates that we are inside the wavelength range where phase matching is possible.
The above considerations assume that the two relevant fundamental waves are identical. This is
referred to as Type I phase matching. Instead, two different fundamentals can be combined which
usually are split from one incident wave. We then speak of Type II phase matching.
The angle Θ also determines the effective tensor element deff used in Eq. 5.8. A suitable combi-
nation defined by the polarization directions involved has to be used.
Besides the phase-matching issues discussed, some more conditions have to be fulfilled to make a
material suitable for efficient second-harmonic generation:

Absorption: The material considered must not absorb both at the fundamental and the harmonic
wavelength. This is usually automatically fulfilled as near the absorption edge of a material
the refractive indices rise considerably and thus prevent phase matching.

Susceptibility Tensor: Trivially, the point symmetry of the crystal must allow for at least one
nonzero tensor element contributing to the geometry necessary for phase matching.

5.3 Quasi Phase Matching

Already in one of the first theoretical publications on nonlinear optics [3], Bloembergen and
coworkers discussed a different method to achieve phase matching for nonlinear optical processes,
especially for second-harmonic generation. They proposed to reverse the sign of the respective
tensor element periodically after an appropriate crystal thickness. In ferroelectric materials this
can be done by an antiparallel poling of crystal regions, ferroelectric domains. The geometry for
a typical example (lithium niobate or lithium tantalate) is sketched in Fig. 25.

Figure 25: Periodically poled domain structure for second-harmonic generation in materials like
lithium niobate or lithium tantalate.

The usage of such periodically poled structures is commonly referred to as quasi phase match-
ing. The momentum conservation law is fulfilled with the help of the additional vector K which
describes the periodicity of the antiparallel domains:

k2 = k1 + k01 + K . (5.11)
5 H ARMONIC G ENERATION 51

The second-harmonic intensity achieved through the periodically poled geometry is depicted in
Fig. 26. The intensity dependencies are calculated for phase-matched, quasi-phase-matched, and
SHG Intensity [a.u.]

Figure 26: Intensities of phase-matched

(dark gray parabola), quasi-phase-matched
(black curve), and non-phase-matched SHG
(light gray), assuming identical tensor ele-
ments and identical beam geometries.

0 1 2 3 4 5
Crystal Position

non-phase-matched conditions under the assumption of identical tensor elements d involved.

For real SHG materials, however, the situation often can be dramatically improved when large
tensor elements can be used which do not suit conventional phase matching. Let us look at lithium
niobate as an example. For phase-matched SHG from 1000-nm light the tensor element d31 is
used with an absolute value of about 4.3 pm/V [4]. The effective d has approximately the same
value, as the phase matching angle is nearly 90°. In a suitable periodically poled domain pattern,
yet, quasi phase matching can be attained using the tensor element d33 with an absolute value of
approximately 27 pm/V [4]. For quasi phase matching an effective d may also be defined using
the approximation drawn as dashed parabola in Fig. 26, it is the original d multiplied by 2/π. Thus
we arrive at deff of approximately 17 pm/V, four times the value of d33 , yielding a sixteen fold
second harmonic intensity. Fig. 27 shows the two dependencies.

LiNbO
3
Figure 27: Case study lithium niobate:
Comparison of the intensities of phase-
SHG Intensity [a.u.]

matched and quasi-phase-matched SHG, us-

d33 − QPM
ing d31 and d33 , respectively. For the cal-
culations ideal conditions are assumed: a
phase-matching angle of 90° for the PM,
and an exact periodically poled domain pat-
d
31
− PM tern without any deterioration due to domain
walls for the QPM second-harmonic inten-
0 5 10 15 20 25 sity.
Crystal Position

Fig. 27 clearly demonstrates the attractiveness of quasi-phase-matching geometries. They gained

increasing interest in the recent years because of several reasons:

• Successful techniques for the fabrication of periodically poled structures have been devel-
52 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

oped [5].

• Nonlinear optical materials – especially lithium niobate and lithium tantalate – have been
improved to facilitate poling.

• The demand for doubling of low light intensities has increased due to the rapid development
of semiconductor lasers.

• Quasi phase matching extends the wavelength range for nonlinear optical processes up to
the full transparency range of the material.

It should be emphasized that the technique is only applicable to ferroelectric nonlinear optical
materials, thus is not suitable for a number of classical materials.
A periodically poled structure is mathematically described by a square function, and in the fourier
transform of such a square function all odd harmonics of the base periodicity are present. Thus
a periodically poled structure is also usable in higher order [6]. Besides odd harmonics of the
square function, even harmonics can be reached by changing the ‘duty cycle’ appropriately. For
higher orders, K in Eq. 5.11 has to be replaces by mK where m is the order. Compared to first
order, the effective d is reduced by this factor m. Therefore higher orders are only used when it is
not possible to fabricate structures for first order.

5.4 Walk-Off

A well-known effect in birefringent materials is visualized in Fig. 28: Unpolarized light propagat-
ing in an arbitrary direction is refracted in two different ways (double refraction).

Figure 28: Double refraction: The left picture shows the propagation of unpolarized light through
an optically isotropic (left) and an anisotropic crystal (right) – calcite. In the right picture two
polarizers are used to select ordinary and extraordinary light polarization. Picture taken from
Ref. [7].

Fig. 28 shows this double refraction for calcite, a crystal with point group 32/m – thus optically
uniaxial. Inside the crystal, the light is split into two parts for the two possible polarizations. The
ordinary light passes straightly, the extraordinary one is distinctly displaced.
5 H ARMONIC G ENERATION 53

As discussed in the subsection about phase matching, for second-harmonic generation birefrin-
gent crystals are used. Ordinary and extraordinary polarizations have to be applied for the two
waves, fundamental and harmonic, to match the relevant refractive indices. Thus we do suffer the
described problem of double refraction which is called walk-off in the field of nonlinear optics
as it causes a geometric walk-off of one beam from the other one. Fig. 29 shows such a walk-off
geometry for an ordinary fundamental and an extraordinary harmonic beam in an arbitrary crystal
direction of a uniaxial crystal. The effect of the walk-off is a reduction of that interaction volume

Figure 29: Walk-Off: Ordinarily polarized fundamental and extraordinarily polarized harmonic
beam. The regime of quadratic intensity increase is restricted to the overlapping volume between
the two beams. It decreases when the fundamental beam is strongly focussed. As simplification
only the part of the harmonic beam generated in the entrance region of the crystal is drawn, the
contribution of the successive regions is omitted.

where the second-harmonic intensity increases quadratically as a function of crystal length. The
regime of quadratic increase is restricted to the overlapping volume between fundamental and har-
monic beam, i. e. to an effective length Le . For a crystal of length L the total intensity then scales
with L · Le instead of L2 (see Eqs. 5.8 and 5.9).
For a qualitative description of the walk-off, the Maxwell equations have to be concerned:

∇ × E = −Ḃ , ∇ × H = Ḋ + J , ∇D = ρ , ∇B = 0 . (5.12)

Assuming monochromatic plane waves

E(r, t) = E0 ei(ωt−kr) , H(r, t) = . . . , D(r, t) = . . . , B(r, t) = . . . (5.13)

and no charges and currents, we arrive at

∇D = k · D = 0, ∇B = k · B = 0 (5.14)

and
∇ × E = k × E = −Ḃ = −iωB , ∇ × H = k × H = Ḋ = iωD . (5.15)
Assuming further that we are not disturbed by magnetics, i. e. that the relative permeability is
µ = 1, thus B = µ0 H, from Eqs. 5.14 and 5.15 follows that k, D and B are perpendicular to each
other. Due to Eq. 5.15 (left) B is perpendicular to E, thus k, E and D are lying in the same plane
perpendicular to B. D and E are connected by the permittivity 

D = 0 E (5.16)
54 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

where  is a second rank tensor of the form

 
11 0 0
 =  0 22 0  . (5.17)
 

0 0 33

For optically isotropic materials, 11 = 22 = 33 , thus always E k D. For uniaxial materials,
11 = 22 , thus E k D for ordinary polarization and E ∦ D for extraordinary polarization.
The direction of energy flow is defined by the Poynting vector

S=E×H (5.18)

which for extraordinary polarization thus is not parallel to the k-vector – we have walk-off. This
uniaxial situation is sketched in Fig. 30.

Figure 30: Light propagation in a uniaxial material, the optical axis is in z-direction. Left: Ordi-
nary polarization, E and D (not shown) are parallel to each other and perpendicular to k. Middle:
Extraordinary polarization, E not perpendicular to k, thus S not parallel to k – walk-off. Right:
only xz-plane shown. ψ is the walk-off angle.

The walk-off angle usually is in the order of some degrees. Quantitative formulas are given in
many articles and textbooks for the various doubling geometries. For the case of negative birefrin-
gent materials (ordinary fundamental, extraordinary harmonic wave) and the usual case of Type I
phase matching,e. g., Boyd et al. [8] give the formula
 
1 o 2 1 1
tan ψ = (nω ) − sin 2Θ . (5.19)
2 (ne2ω )2 (n02ω )2

Eq. 5.19 shows that there is no walk-off, i. e. the walk-off angle ψ will be zero, for Θ = 0 and for
Θ = 90◦ . In uniaxial crystals that means propagation along and perpendicular to the optical axis,
respectively. With k along the optical axis of course no phase matching is possible. However,
it might be possible for k perpendicular to it if the magnitude of the birefringence suits. Often
this can be tuned within certain limits by varying the temperature. This sort of phase matching is
known as 90◦ phase matching or temperature phase matching. As the two relevant refractive index
surfaces in this case do not intersect, instead are tangent to each other, thus allowing for a larger
angle uncertainty, it is also referred to as Non Critical Phase Matching.
5 H ARMONIC G ENERATION 55

There is a second type of geometries where walk-off is completely absent – that’s in all quasi-
phase-matching schemes. The periodically poled structures there are always made for a wave
propagation along a highly symmetric crystal direction. To make use d33 in lithium niobate, e. g.,
the beams propagate perpendicular to the c-direction of the crystal, allowing the polarization of
both, fundamental and harmonic wave, respectively, to be in c-direction. This complete absence of
walk-off problems is an important additional advantage of quasi-phase-matching configurations.

5.5 High-Order Harmonic Generation

For an efficient generation of harmonic light commonly crystals are used which show large non-
linear susceptibilities. For the generation of even harmonics, e. g. the second harmonic, these
crystals, in addition, have to be acentric. A very crucial condition, however, is good transparency,
the absence of absorption, at both the fundamental and harmonic wavelengths. In solids, this can
be accomplished down to approximately 150 nm. For shorter wavelengths, therefore, one has to
use other arrangements.
Besides the scientific interest, shorter wavelengths are important at least in two fields of current
optical applications:

• Lithographical techniques for the fabrication of integrated circuits are limited by the wave-
length of light employed. Presently excimer laser light of 192 nm is used in combination
with silica optics, the next step will be 157 nm in combination with calcium fluoride optics.
This will be the limit of excimer lasers and conventional optics. Beyond this limit, new light
sources (and new optical concepts) are in demand.

• For many studies – especially in biological systems – one would like to have single short
pulses of X-rays. A very interesting X-ray wavelength region is the so called ‘water window’
(3–4 nm) where water and carbon have a reduced absorption. This allows diffraction and
absorption imaging of biological systems on a molecular scale, and – if pulses can be used
– with an extremely good time resolution.

To accomplish the generation of harmonic light well below 150 nm, media transparent in this
region – gases or clusters – have to be used. Atoms, molecules, clusters in general are centrosym-
metric or even isotropic, thus only odd harmonics are generated. For a good conversion efficiency,
the light of a pulsed high-power laser is focused onto the gaseous medium. Doing this, the elec-
tromagnetic field becomes of the same magnitude as the Coulomb field, which binds a 1s electron
in a Hydrogen atom (5.1 × 109 Vm−1 ). At such high fields various nonlinear phenomena can
happen [9], three typical processes are sketched in Fig. 31:

• Electrons initially in the ground state absorb a large number of photons, many more than
the minimum number required for ionization, thus being ionized with a high kinetic energy.
This process, shown for the first time in 1979 [10], is called Above Threshold Ionisation
(ATI).
56 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

Figure 31: Excitation processes in atoms in strong laser fields [9]. ATI: above threshold ionization,
MI: multiple ionization, HHG: high-order harmonic generation.

• Not only one, but many electrons can be emitted from atoms subject to strong laser fields.
They can be emitted one at a time, in a sequential process, or simultaneously, a mechanism
called direct, or non-sequential. Double ionization of alkaline earth atoms was observed as
early as in 1975 [11] and the first evidence for non-sequential ionization of rare gas atoms
was first demonstrated in 1983 [12].

• Finally, efficient photon emission in the extreme ultraviolet (EUV) range, in the form of
high-order harmonics of the fundamental laser field (HHG), shown for the first time in
1987 [13, 14], can occur.

The described processes are mutually competing, all are scaling with a high power of the incident
light intensity. Only the third one (HHG) leads to the generation of coherent EUV light.
About the spectrum generated, Anne L’Huillier, one of the pioneers in this field, writes [9]: A
high-order harmonic spectrum consists of a sequence of peaks centered at frequency qω, where q is
an odd integer. Only odd orders can be observed, owing to inversion symmetry in an atomic gas. A
HHG spectrum has a characteristic behavior: A fast decrease for the first few harmonics, followed
by a long plateau of harmonics with approximately constant intensity. The plateau ends up by a
sharp cut-off. Most of the early work on harmonic generation concentrated on the extension
of the plateau, i. e. the generation of harmonics of shorter wavelength. Today, harmonic spectra
produced with short and intense laser pulses extend to more than 500 eV, down to the water window
below the carbon K-edge at 4.4 nm. A large effort has been devoted to optimize and characterize
the properties of this new source of EUV radiation. A milestone in the understanding of HHG
processes was the finding by Kulander and coworkers in 1992 [15] that the cut-off position in
the harmonic spectrum follows the universal law Emax ≈ Ip + 3Up . This result was immediately
interpreted in terms of the simple man’s theory, and led to the formulation of the strong field
 5 H ARMONIC G ENERATION 57

approximation (SFA). A realistic description of HHG involves, however, not only the calculation
of the single atom response, but also the solution of propagation (Maxwell) equations for the
emitted radiation.
Simplified, the above expression for Emax means that the maximum energy in the generated har-
monic spectrum corresponds to the maximum energy imposed on a quasi-free electron by the
electromagnetic field of the incident laser pulse. A schematic sketch of this strong field approxi-
mation is given in Fig. 32. In the strong electromagnetic field of the focused laser beam the atomic

Figure 32: Potential distortion in an extremely

strong light field. An electron is accelerated
in the strong field and produces X-rays when
falling back to the ionic core (picture taken from
Ref. [16]).

potential is highly distorted, an electron is accelerated. When the field reverses, the electron can
fall back to the ionic core and emit photons during the collision process. The result is a burst
of X-rays. This process repeats itself many times over the duration of the laser pulse each time
the electromagnetic field reverses sign. As shown in Fig. 33, the X-ray pulses itself are signif-
icantly shorter (sub-femtosecond) than the period of the original electromagnetic wave. Using

2 X−Ray Pulses
1 Figure 33: Electromagnetic field oscillation in an
ultrashort light pulse. Near the zero crossings
0
−5 0 5 bunches of X-rays are generated.
Time [fs]

extremely short light pulses will produce a single X-ray pulse in the attosecond regime for each of
the incident light pulse.
The wavelength of the emitted light depends on the amount of energy acquired by the electrons
over a half-cycle. Yet, despite the similarity to bremsstrahlung no continuous X-ray spectrum is
generated. Due to the short overall interaction time the excitation and the X-ray generation are not
independent from each other. Thus conservation laws and symmetry relations have to be obeyed,
yielding peaks at odd harmonics of the fundamental frequency.
Several techniques can be used to enhance special regions in the generated X-ray spectrum. For
lower energies, e. g., enhancement due to resonances in the electronic potential scheme is possible.
Of course this doesn’t work at higher energies where the electrons are regarded as quasi-free. And,
albeit not so expressed as in the case of nonlinear crystals, phase matched [17] and quasi phase
58 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

matched arrangements [18] are important enhancement schemes also in the case of harmonic
generation in gases.

References

[1] P. A. Franken, A. E. Hill, C. W. Peters, G. Weinreich. Generation of Optical Harmonics.

Phys. Rev. Lett. 7, 118 (1961).

[2] T. H. Maiman. Stimulated Optical Radiation in Ruby. Nature 187, 493–494 (1960).

[3] J. A. Armstrong, N. Bloembergen, J. Ducuing, P. S. Pershan. Interactions between Light

Waves in a Nonlinear Dielectric. Phys. Rev. 127, 1918–1939 (1962).

[4] David A. Roberts. Simplified Characterization of Uniaxial and Biaxial Nonlinear Optical
Crystals: A Plea for Standardization of Nomenclature and Conventions. IEEE J. Quantum
Electron. QE-28, 2057 (1992).

[5] V. Ya. Shur, E. L. Rumyantsev, E. V. Nikolaeva, E. I. Shishkin, R. G. Batchko, M. M. Fejer,

R.L. Byer. Recent Achievements in Domain Engineering in Lithium Niobate and Lithium
Tantalate. Ferroelectrics 257(191–202) (2001).

[6] M. M. Fejer, G. A. Magel, D. H. Jundt, R. L. Byer. Quasi-Phase-Matched Second Harmonic

Generation Tuning and Tolerances. IEEE J. Quantum Electron. QE-28, 2631 (1992).

[7] Jörg Schmiedmayer. Laser Physics and Photonics. University of Heidelberg, 2002.
http://www.physi.uni-heidelberg.de/ schmiedm/Vorlesung/LasPhys02.

[8] G. D. Boyd, A. Ashkin, J. M. Dziedzic, D. A. Kleinman. Second-Harmonic Generation of

Light with Double Refraction. Phys. Rev. 137, 1305 – 1320 (1965).

[9] Anne L’Huillier. Atoms in Strong Laser Fields. Europhysics News 33, No. 6 (2002).

[10] P. Agostini, F. Fabre, G. Mainfray, G. Petite, N. K. Rahman. Free-Free Transitions Following

Six-Photon Ionization of Xenon Atoms. Phys. Rev. Lett. 42, 1127 (1979).

[11] V. V. Suran, I. P. Zapesochnyi. Sov. Tch. Phys. Lett. 1, 420 (1975).

[12] A. L’Huillier, A. Lompre, G. Mainfray, C. Manus. Multiply charged ions induced by multi-
photon absorption in rare gases at 0.53 µm. Phys. Rev. A 27, 2503 (1983).

[13] A. McPherson, G. Gibson, H. Jara, U. Johann, T. S. Luk, I. A. McIntyre, K. Boyer, C. K.

Rhodes. J. Opt. Soc. Am. B 4, 595 (1987).

[14] M. Ferray, A. L’Huillier, X. F. Li, L. A. Lompré, G. Mainfray, C. Manus. J. Phys. B: At.

Mol. Opt. Phys. 21, L31 (1988).

[15] J. K. Krause, K. J. Schafer, K. C. Kulander. High-order harmonic generation processes from

atoms and ions in the high intensity regime. Phys. Rev. Lett. 68, 3535 (1992).
5 H ARMONIC G ENERATION 59

[16] Ultrafast X-ray Generation via Harmonic Generation. Vrije Universiteit Amsterdam.
http://www.nat.vu.nl/atom/x-ray.html.

[17] Katsumi Midorikawa, Yusuke Tamaki, Jiro Itatani. Phase-matched high-order harmonic
generation with self-guided intense femtosecond laser pulses. RIKEN Review 31, 38 (2000).

[18] A. Paul, R. A. Bartels, R. Tobey, H. Green, S. Weiman, I. P. Christov, M. M. Murmane, H. C.

Kapteyn, S. Backus. Quasi-phase-matched generation of coherent extreme-ultraviolet light.
Nature 421, 51 – 54 (2003).
60 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

6 Measurement of Nonlinear Optical Properties

Nonlinear optical materials are important for many applications in optics. Therefore an intensive
search for new, better materials is still in progress in many research institutes. To characterize these
new materials, several techniques have been developed which are widely applied [1]. Various
properties are of importance. If a material should be usable, e. g., for second-harmonic generation,
it should belong to a non-centrosymmetric point group. Thus a test for this should be possible at
a very early state, the powder technique may be used for this purpose. All other investigation
methods need larger crystals which are more difficult and time-consumptive to fabricate. Larger
crystals in general are also necessary for the investigation of the linear optical properties important
for nonlinear optical applications like transmission range and refractive index.

6.1 Powder Technique

This technique is described in the comprehensive article A Powder Technique for the Evaluation of
Nonlinear Optical Materials by S. K. Kurtz and T. T. Perry in 1968 [2]. Since that time it is widely
used as one of the simplest methods for a rapid classification of new materials. For the application
of the technique the material is only required in powder form (which is easily available in most
cases). Thus it can be applied at a very early state after the first fabrication of a new material,
for instance in a chemists lab. The basic configuration for powder SHG is shown in Fig. 34 (the
figure is taken from the original article, in a present-day setup several parts would be replaced by
up-to-date ones).

Figure 34: Setup for

studying the second-
harmonic generation in
powder samples.

Light from a Q-switched Nd:YAG laser is directed onto the powder sample, the second-harmonic
light is collected by appropriate optics and – after filtering out the fundamental light – detected by
a photomultiplier. In this original setup the photomultiplier signal and a monitor signal from the
fundamental beam are displayed on an oscilloscope.
6 M EASUREMENT OF N ONLINEAR O PTICAL P ROPERTIES 61

In the powder sample the light, fundamental and harmonic, is randomly scattered. This scattering
can be greatly reduced when the powder is immersed in a liquid of similar refractive index. Usu-
ally, however, immersion is regarded as an additional complication, and, what is more important,
it is difficult to find liquids with matching refractive index – especially when working with ma-
terials of high refractive index and/or large birefringence. Thus the usual way is to work without
immersion. The scattering leads to an angular distribution, which is similar to that of a planar
radiator obeying Lambert’s cosine law, with an appreciable amount in backward direction. This
angular dependence is sketched in Fig. 35.

Figure 35: Angular distribution

of second harmonic generated in
a powder sample (picture taken
from [2]). When the powder is
immersed in an index-matching
liquid, a narrow angular distrib-
ution in forward direction shows
up, otherwise a broad angular dis-
tribution in forward and in back-
ward direction is found.

In a practical application it is thus advisable to collect the generated light from all spatial direc-
tions. This can be done by placing the sample within an integrating Ulbricht sphere [3, 4] which
collects a certain amount of light from all directions.
The generated harmonic intensity depends in a characteristic way on the average particle size in
the powder. This size dependence is different for materials which are phase matchable and those
which are not. The two dependences are schematically sketched in Fig. 36.
A detailed theory for these dependences can be found in [2], to understand it in principle, we can
find simpler arguments. Let us assume that we have a powder volume V completely filled with
randomly oriented particles of size r. The number of particles will be in the order of N = V /r3 .
All particles are illuminated by the fundamental laser light, every particle contributing an area
62 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

Second Harmonic Intensity

1
Phase Matchable
0.8

0.6

0.4
Non Phase Matchable
0.2

0
0 1 2 3 4 5 6 7 8 9 10
Average Particle Size <r>/<Lc>

Figure 36: Schematic representation of different particle-size dependences for phase-matchable

and non-phase-matchable materials.

A = r2 . Due to the random orientation, the second harmonic intensities of different particles add
up incoherently. According to Fig. 21, for a non-phase-matchable material the SHG intensity Is
for a single particle of size r will increase quadratically for small sizes Is (r < Lc ) ∝ r2 , then
approaching a constant average value Is (r > Lc ) ∝ L2c . The total SHG intensity I = N · A · Is
then is proportional to r for r < Lc and proportional to L2c /r for r > Lc .
For a phase-matchable material we get the same result for small particle sizes. For large particles
the single-particle intensity still would further increase quadratically with the particle size – but
only for particles properly oriented. The ‘sharpness’ of this condition scales with particle size,
thus the share of properly oriented particles scales with r−1 . Putting all together, we get constant
intensity for large particle sizes in a phase-matchable material.
Using the powder technique, materials can be classified into different categories at a very early
state of the investigations. Thus an early decision about new materials is possible. These categories
include:

Centrosymmetric: No second-harmonic intensity found by the powder technique.

Phase Matchable: Constant second-harmonic intensity at increasing particle sizes.

Non Phase Matchable: Second-harmonic intensity decreasing as a function of the particle size.

The decision about centric symmetry can be found in one measurement without the necessity of
using particle size fractions. A test for phase matching can be made using several particle sizes
which have to be larger than the average coherence length. Comparing different materials – known
and unknown ones – it is also possible to get a rough estimate about the magnitude of the effective
tensor elements of the SHG tensor.
6 M EASUREMENT OF N ONLINEAR O PTICAL P ROPERTIES 63

6.2 Maker Fringe Method

The observation of periodic maxima and minima in the second-harmonic intensity as a plane
parallel slab is rotated about an axes perpendicular to the laser beam was first reported by Maker
et al. [5] for SiO2 in 1962. The geometry for such a measurement is sketched in Fig. 37. A thin

Figure 37: Rotating slab geometry for the measurement of Maker fringes. The plane parallel slab
is rotated around the indicated axis which is perpendicular to the beam direction. The generated
second-harmonic intensity is measured as a function of the rotation angle.

crystal platelet is rotated, thus a variation in the wave vector mismatch ∆k between the harmonic
polarization (forced wave) and free harmonic waves is caused
|∆k| = |k2ω − 2kω | = (4π/λ)|n2ω cos β2 − nω cos β1 | (6.1)
where β1 and β2 are the angles of refraction for the fundamental and harmonic waves, respectively.
As shown in Fig: 38, the wave vector mismatch ∆k remains perpendicular to the crystal faces
even for arbitrary nonnormal incidence of the fundamental beam. This can be derived from simple

∆k
2kω k2ω
Figure 38: Wave vectors of the second-harmonic po-
β1 β2 larization (forced wave) and of the free harmonic
wave and the corresponding mismatch ∆k for a fun-
α damental wave incident at an arbitrary angle α onto a
slab.
L

geometric considerations. From Snellius law we get

sin β1 = sin α/nω and sin β2 = sin α/n2ω . (6.2)
The lengths of the wave vectors are
|2kω | = (4π/λ)nω and |k2ω | = (4π/λ)n2ω (6.3)
where λ is the fundamental wavelength. Their components parallel to the crystal faces are equal
|2kω,|| | = |2kω | sin β1 = (4π/λ) sin α and |k2ω,|| | = |k2ω | sin β2 = (4π/λ) sin α . (6.4)
64 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

Therefore the difference vector ∆k is perpendicular to the crystal faces and can be expressed
according to Eq. 6.1.
The total second-harmonic intensity is found by integration over the slab thickness L (similar as
in section 5.1, Eqs. 5.5–5.8)
sin2 (∆k(α) L/2)
I (2) (α) = C · d2eff (α) · I (1) 2 · . (6.5)
(∆k(α)/2)2
This angular dependence of the second-harmonic intensity calculated for a slab of 1 mm thickness
with the refractive indices 2.00 and 2.04 and for a fundamental wavelength of 1 µm is shown in
Fig. 39.
Harmonic Intensity

−40 −30 −20 −10 0 10 20 30 40

External Angle [deg]

Figure 39: Calculated Maker fringes for a slab geometry: angular dependence of the second-
harmonic intensity for a plane parallel slab rotated about an axes perpendicular to the laser beam.

Fitting the angular dependence given in Eq. 6.5 to a measured fringe pattern yields ∆k(α). Rel-
ative measurements of the various tensor elements of one material deff and extrapolations to the
respective dik are possible by using plates of different orientations and different light polarizations.
The values of one material can be referred to a ‘standard’ by comparing to slabs of this standard
material using the identical geometry. The magnitude of the effective second-harmonic tensor
element relative to that of the standard material can be obtained from the relation [1]
IM (0) η 1/2 Lstd
 
deff c (0)
std
= std std
(6.6)
deff IM (0) η Lc (0)
where IM is the intensity envelope, η the reflection correction, and Lc the coherence length, all
taken at normal incidence (α = 0).
Instead of rotating a plane parallel slab, one can use a wedge shaped crystal to produce Maker
fringes. The geometry is shown in Fig. 40. In such a geometry the orientation of the crystal is
fixed, the wave vector mismatch ∆k thus is kept constant, only the effective length L is varied
according to the lateral shift. The second-harmonic intensity is given in a similar way as in Eq. 6.5
sin2 (∆k L(r)/2)
Z
I (2) (L) = C · d2eff · I (1) 2 (r) · dr , (6.7)
(∆k/2)2
6 M EASUREMENT OF N ONLINEAR O PTICAL P ROPERTIES 65

Figure 40: Wedge geometry for the measurement of Maker fringes. A crystal wedge is moved
perpendicular to the laser beam, the second-harmonic intensity is measured as a function of the
lateral shift.

the integration has to be performed over the laser beam area. Depending on crystal orientation
and light polarizations, deff in general can be expressed by a single element dik . Again, accurate
relative measurements are possible using different orientations and polarizations and comparisons
to standard crystals.
Typical (calculated) intensity dependences as a function of the lateral shift of the wedge are shown
in Fig. 41. Due to the constant wave vector mismatch ∆k the measured dependences in a wedge
measurement are much simpler – strongly sinusoidal with constant amplitude if absorption can be
neglected – and thus easier to evaluate than in a slab measurement.
Harmonic Intensity

0 0.1 0.2 0.3 0.4 0.5 0.6

Lateral Shift [mm]

Figure 41: Calculated Maker fringes for a wedge geometry: second-harmonic intensity as a func-
tion of the lateral shift for three different laser beam sizes (0, 70, 200 µm). For the calculation
refractive indices of 2.00 and 2.04 were assumed, a fundamental wavelength of 1 µm and a wedge
angle of 5◦ .

6.3 Absolute Measurements by Phase-Matched SHG

The methods discussed in the preceding two sections both are not well suited for absolute mea-
surements of d, although Maker fringe measurements in principle could be evaluated in that way.
66 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

Properties of Gaussian beams

For most theoretical considerations in optics,
plane waves are assumed as a solution of the
w0⋅√2
wave equation. Things are kept simple in that w0
way. In all practical systems, however, wave Θ R
fronts can not extend to infinity, we never have 0 z
0
exact plane waves.
To describe, what we colloquially characterize as
a ‘light beam’, the so-called paraxial approxima-
tion can be used. It’s useful for the description of
laser beams as well as e. g. for wave transforma-
Figure 42: Profile of a Gaussian beam near the
tion calculations in conventional optical systems
focus.
like combinations of lenses.
Restricting it to a single frequency and sepa- The beam waist w is defined as the distance from
rating off the time dependence, from the wave the beam axis where the amplitude has decreased
equation the Helmholtz equation is derived to 1/e. In terms of the minimum beam waist it is
given by
(∆ + k 2 )E(r) = 0 . (6.8) p
w(z) = w0 1 + (z/z0 )2 . (6.12)
In the paraxial approximation, it is assumed, that
the wave propagates only in z direction, not in The distance z0 from the minimum beam waist,
the x and y direction where the beam area is twice the minimum area,
is called the confocal parameter or Rayleigh
length
E(r) = Ψ(x, y, z)e−ikz . (6.9) π
z0 = wo2 . (6.13)
λ
Neglecting ∂ 2 Ψ/∂z 2 , as Ψ varies only slowly The curvature radius R of the phasefront of the
with z, we arrive at the paraxial wave equation wave is

R(z) = z 1 + (z0 /z)2 ,

 
∂2Ψ ∂2Ψ ∂Ψ (6.14)
+ − 2ik =0. (6.10)
∂x2 ∂y 2 ∂z
the beam divergence
The further treatment of this equation can be λ
found in textbooks about optics, e. g. in [6]. Θ = w0 /z0 = . (6.15)
πw0
The simplest solution is a circular symmet-
ric Gaussian amplitude distribution. Such a From Eq. 6.14 follows that the phasefronts have
Gaussian beam then can be characterized by pa- their maximum curvature at z0 . The region
rameters which all can be referred to the mini- |z| < z0 is often called near field, that outside
mum beam waist w0 and the wavelength λ (|z| > z0 ) far field.
 2 The minimum achievable beam waist for a
x + y2

w0 Gaussian beam can be derived from Eq. 6.15 (it
Ψ(x, y, z) = A0 exp − 2 (6.11)
w(z) w (z) is limited by Fresnel diffraction)
x2 + y 2
 
z λ F# λ
· exp −ik + i arctan , w0,min = = (6.16)
2R(z) z0 πΘmax 2π

A0 is an amplitude factor. The geometry near the where F# is the F number (aperture) of the opti-
minimum beam waist is sketched in Fig. 42. cal system used.
6 M EASUREMENT OF N ONLINEAR O PTICAL P ROPERTIES 67

To get accurate absolute values, one can apply phase-matched harmonic generation carried out
under a well-defined geometry.
One scenario of a ‘well-defined geometry’ is the application of Gaussian beams as delivered e. g.
by an ideal laser working in TEM00 mode. Some basics of Gaussian beams are summarized in the
box on page 66. As shown there, the spatial behavior of the light amplitude in a Gaussian beam
can be exactly described. In a nonlinear crystal this spatial behavior is modified by the refractive
index, in addition, walk-off effects (see section 5.4) may hamper the generation of harmonic light.
Considering all these geometry influences, Boyd and Kleinman obtained an exact integral expres-
sion for the second harmonic power generated by a focused Gaussian beam. The mathematical
description is found in their rather comprehensive publication [7] or – summarized – in [1]. The
application of their mathematical formalism allows for the absolute determination of effective
SHG tensor elements deff from the measurements of fundamental and second harmonic powers
and the evaluation of the beam and crystal geometries. Many authors have shown in numerous
measurements that an accuracy of approximately 10 % for deff may be achieved. A drawback of
the method is that it delivers the effective d for the special phase-matching configuration which
for many symmetries is an angle-dependent combination of several dik s. To get the individual
elements, the method thus has to be combined with a relative one (Maker fringes).

6.4 Z-Scan Technique

Some nonlinear properties of materials can be measured using an experimental setup where the
material under consideration is moved along the beam axis (z axis) through the focus region of
a focused beam. The properties which can be measured in such a geometry include nonlinear
absorption, also referred to as two-photon absorption, and nonlinear refraction. Measuring these
two quantities, the complex third order susceptibility can be derived.
According to Eq. 4.12 the third order nonlinear polarization for ω = ω + ω − ω can be written as

P(3) (k, ω) = 0 χ(3) (k = k + k − k, ω = ω + ω − ω)E(k, ω)E(k, ω)E(k, ω) . (6.17)

This is a contribution to the (linear) polarization at ω which acts like an intensity-proportional

contribution to the linear susceptibility. Writing absorption and refractive index with constant and
intensity-dependent terms

α = α0 + βI and n = n0 + n2 I , (6.18)

the real and imaginary part of χ(3) can be derived from the intensity-dependent terms

Re χ(3) ∝ n2 and Im χ(3) ∝ β . (6.19)

In Z-scan measurements the light intensity on the sample varies when moving through the focus
of a Gaussian beam (for the properties of Gaussian beams see the box on page 66). Thus the
intensity-dependent parts of absorption and refraction are influenced.
68 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

Figure 43: Z-scan: open aperture geometry, the integrated light intensity is measured as a function
of crystal position. Left: thin sample ( < z0 of the Gaussian beam), right: thick sample ( > z0 ).

Figure 44: Z-scan: closed aperture geometry, the light intensity in the center of the beam is
measured as a function of crystal position.

The typical geometries for Z-scan measurements are sketched in Fig. 43 and 44. The focused
Gaussian beam is propagating in z direction, the crystal is moved through the focus. The inte-
grated intensity will be influenced mainly by the nonlinear absorption, the angular distribution of
the intensity, however, will be affected by both nonlinear absorption and refraction. Thus in an
open aperture geometry the nonlinear absorption can be measured, in a closed aperture geometry
the nonlinear refraction. One has to discriminate whether the sample is thin or thick (compared
to z0 of the Gaussian beams). For both cases comprehensive mathematical descriptions have been
developed [8, 9] which can be used for the evaluation of Z-scan measurements.
The experimental results of typical Z-scan measurements (here on lithium niobate crystals) are
shown in Fig. 45 together with fit curves [10]. From the fit, the authors derive the values for the
real and imaginary parts of χ(3) to be 1.02 × 10−20 m2 V−2 and 2.03 × 10−21 m2 V−2 , respectively.
It should be emphasized that for Z-scan measurements lasers with extremely short high-power
pulses should be used due to two main reasons:

• Values of χ(3) in general are small and the (relative) effects scale with the laser power. High
laser power thus facilitates the measurement distinctly.
• Thermal effects and other slow effects like the photorefractive effect may lead to similar
results as the third order susceptibility. They can be efficiently suppressed when extremely
short pulses are applied.

References

[1] Herbert Rabin, C. L. Tang, editors. Quantum Electronics, Volume I: Nonlinear Optics, Part
A, Chapter 3. Stewart K. Kurtz: Measurement of Nonlinear Optical Susceptibilities, page
6 M EASUREMENT OF N ONLINEAR O PTICAL P ROPERTIES 69

Figure 45: Z-scan measurements on lithium niobate for various laser intensities [10]: (a) 22,
(b) 12, (c) 6 GW/cm2 . Experimental data (circles) and theoretical fits (solid lines). Left: open
aperture geometry – nonlinear absorption, right: closed aperture geometry – nonlinear refraction.
The curves for (b) and (c) are vertically shifted for presentation.

209. Academic Press, Inc., 1975.

[2] S. K. Kurtz, T. T. Perry. A Powder Technique for the Evaluation of Nonlinear Optical Mate-
rials. J. Appl. Phys. 39, 3798 (1968).
[3] R. Ulbricht. Das Kugelphotometer. R. Oldenburg, München und Berlin, 1920.
[4] D. G. Goebel. Generalized Integrating-Sphere Theory. Appl. Opt. 6, 125–128 (1967).
[5] P. D.Maker, R. W. Terhune, M. Nisenoff, C. M. Savage. Effect of Dispersion and Focusing
on the Production of Optical Harmonics. Phys. Rev. Lett. 8, 21 (1962).
[6] Robert D. Guenther. Modern Optics. John Wiley & Sons, 1990.
[7] G. D. Boyd, D. A. Kleinman. Parametric interaction of focused gaussian light beams. J.
Appl. Phys. 39, 3597–3639 (1968).
[8] P. P. Banerjee, R. M. Misra, M. Maghraoui. Theoretical and experimental studies of porpa-
gation of beams through a finite sample of a cubically nonlinear material. J. Opt. Soc. Am.
B 8, 1072–1080 (1991).
[9] J. A. Hermann, R. G. McDuff. Analysis of spatial scanning with thick optically nonlinear
media. J. Opt. Soc. Am. B 10, 2056–2064 (1993).
[10] Heping Li, Feng Zhou, Xuejun Zhang, Wei Ji. Picosecond Z-scan study of bound electronic
Kerr effect in LiNbO3 crystal associated with two-photon absorption. Appl. Phys. B 64,
659–662 (1997).
70 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

7 Non-Collinear Harmonic Generation

Usually nonlinear optical processes are regarded to be collinear which means that all participating
light beams are pointing approximately into the same direction. Such collinear geometries have
the advantage of large interaction lengths, thus optimize the efficiency of the nonlinear interaction
– provided that phase matching or quasi phase matching is obeyed. In collinear geometries the
momentum conservation law is fulfilled in a scalar sense, the lengths ki of all vectors ki add up to
zero.
However, it’s not a must to work with collinear beams, non-collinear interactions are possible as
well. The momentum conservation law then is only fulfilled in a vectorial sense
X X
ki = 0 yet ki 6= 0 . (7.1)

As the interacting beams are inclined to each other, the intersection volume will be small, the
resulting short interaction length will hamper efficiency. Non-collinear geometries are therefore
not suitable for efficient frequency conversion, they are ‘only’ interesting for their physics and –
as we will see – they can be useful for material characterization. Some examples for non-collinear
interactions shall illustrate this.

7.1 Induced Non-Collinear Frequency Doubling

This technique utilizes two fundamental light beams inclined to each other to fulfill the vectorial
phase matching condition
k2 = k1 + k01 . (7.2)
The corresponding geometry is sketched in Fig. 46.

k1
k1’ Figure 46: Momentum diagram for induced non-collinear fre-
Θ Θ’
quency doubling.
k
2

The vectorial phase matching condition of Eq. 7.2 can be referred to a condition for the respective
refractive indices n(ω, k). Using
ω
|k2 | = |k1 | cos Θ + |k01 | cos Θ0 and |k| = np (ω, k) (7.3)
c
(p indicates the light polarization) yields

(ω1 + ω10 )np (ω1 + ω10 , k1 + k01 ) = ω1 nq (ω1 , k1 ) cos Θ + ω10 nr (ω10 , k01 ) cos Θ0 . (7.4)

The two fundamental beams usually are derived from the same laser as schematically sketched
in Fig. 47 which means ω1 = ω10 = ω. Furthermore, a geometry can be chosen where the
7 N ON -C OLLINEAR H ARMONIC G ENERATION 71

two fundamental beams are arranged symmetrically with respect to the index ellipsoid and have
symmetric polarization, which further simplifies Eq. 7.4 to
np (2ω, k2 ) = nq (ω, k1 ) cos Θ . (7.5)

Figure 47: Experimental arrangement for measuring induced non-collinear frequency doubling.
S: beam splitter, O: focussing lens, K: temperature controlled sample holder, moveable in all three
spatial directions, B: aperture for blocking the fundamental beams, PM: photomultiplier.

The angle Θ and the polarizations of the incident beams have to be chosen in an appropriate way
to fulfill Eq. 7.5. Obviously this condition is very sensitive to variations in the refractive indices.
As in more detail shown in Fig. 48, the interaction volume, i. e. the region from which second
harmonic light originates, is limited in all three spatial dimensions. Thus such an experiment can

Figure 48: Induced non-collinear frequency

doubling: detailed beam geometry inside
the sample.

be used to get information just about the volume element under illumination. Moving the sample
in all spatial directions yields a fully three-dimensional topography. The resolution depends on
the beam geometries and on the angle Θ.
The technique may be illustrated by two typical applications concerning the characterization of
optical crystals – composition measurements in lithium niobate and detection of domain borders
in potassium niobate [1].

Composition Measurements in Lithium Niobate Lithium niobate is one of the most important
crystals for many electro-optic and acousto-optic devices. Its chemical formula is LiNbO3 but the
real composition usually deviates from the stoichiometry described by the formula. Crystals of
lithium niobate are commonly grown at the congruently melting composition, i. e. at the compo-
sition where liquid and solid states of equal compositions are in an equilibrium. This composition
is at approximately 48.5 mole% of lithium oxide. Crystals grown at this congruently melting
composition are of excellent optical quality and of good homogeneity. Some of the properties,
however, could be improved in crystals of stoichiometric composition. So for instance the electric
field necessary for periodic poling would be considerably lowered. Various efforts therefore have
been made to achieve material of stoichiometric composition.
72 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

One technique now used by several groups is the so-called vapor transport equilibration (VTE)
where thin plates of lithium niobate are heated up in a stoichiometric mixture of lithium oxide
and niobium oxide. Diffusion then leads to the composition equilibration between crystal and
surrounding oxide powder. To improve and optimize the technique, the success of these treat-
ments has to be carefully checked. Induced non-collinear frequency doubling is one possibility to
monitor the composition inside the crystal after the treatment with a good spatial resolution.
Many of the material properties of lithium niobate depend on the composition, these include the
refractive indices. The ordinary index is practically independent from composition, the extraordi-
nary index shows an expressed dependence which is approximately linear. The two dependences
for various wavelengths are shown in the left part of Fig. 49.

Figure 49: Composition dependence of the refractive indices of lithium niobate for various wave-
lengths (left) [2] and therefrom calculated functional dependence between phase matching tem-
perature for induced non-collinear frequency doubling and composition for several angles Θ
(right) [3]. The calculation is made for a fundamental wavelength of 1064 nm (Nd:YAG laser).

As – like in every material – the refractive indices are temperature dependent, phase matching
conditions can be adjusted using the temperature as a parameter. The two dependences can be
combined, the measured phase matching temperature for a fixed angle Θ can be utilized as a very
sensitive indicator for the crystal composition. This functional dependence, composition versus
phase matching temperature, is shown in the right part of Fig. 49 for several angles Θ. The curves
are calculated using a generalized fit for the refractive indices of lithium niobate as a function of
wavelength, composition, temperature, and doping [3, 4].
From the dependences in Fig. 49 an excellent sensitivity of the method is apparent, at least for
relative measurements. One degree variation in the phase matching temperature corresponds to a
variation of 0.005 mole% in the lithium oxide concentration in the crystal.
A typical measurement on a VTE-treated sample is shown in Fig. 50. The sample had been treated
for a comparably short time, thus the crystal had not reached the final homogeneity. Instead, the
diffusion profiles in two different directions, z and x, with their characteristic form of a comple-
mentary error function (erfc) are observed.
7 N ON -C OLLINEAR H ARMONIC G ENERATION 73
PM−Temperatur [°C]

150
Figure 50: Composition profile
100 in a VTE-treated lithium niobate
50 sample. At the borders (z = 0,
x = 0, x = 4) a stoichiometric
0
composition of 50 mole% lithium
0 oxide is reached whereas near the
0.5 center of the sample the composi-
0
1 1 tion is still the congruently melt-
2
z [mm] 1.5 4
3 x [mm] ing one of the untreated material.

48.5 49 49.5 50
Li−Content [%]

Domain Borders in Potassium Niobate Ferroelectric materials commonly undergo a phase tran-
sition from a high temperature paraelectric to a low temperature ferroelectric phase. Depending
on the symmetries of the high- and the low-temperature phases ferroelectric materials may con-
tain ferroelectric domains in different geometric configurations. Thus materials with a tetrago-
nal or trigonal symmetry both in the high- and in the low-temperature phase can form domains
only in two polarization directions – parallel or antiparallel to the crystallographic c-axis. The
refractive indices are identical for both domain directions. In contrast to this, materials with a
high-temperature cubic and a low-temperature tetragonal phase can form domains with their polar
axis pointing into any of the six directions of the former cubic axes. There are thus three possible
orientations of the index ellipsoid. Materials belonging to the first group include lithium nio-
bate, lithium tantalate and strontium barium niobate. To the second group belong all perovskites
including barium titanate and potassium niobate.
To detect such misoriented domains one can utilize the different phase-matching directions for the
different orientations. Adjusting the two crossed laser beams such that phase matching for one
orientation is achieved, large second-harmonic intensities are measured when inside a properly
oriented domain and practically no intensity outside. The spatial derivative of the intensity field
then yields the borders between adjacent domains of different orientation. Fig. 51 gives an example
for such a measurement.

7.2 Spontaneous Non-Collinear Frequency Doubling

In contrast to induced non-collinear frequency doubling, spontaneous non-collinear frequency

doubling is a type of optical second harmonic generation that uses randomly scattered light to
provide additional fundamental beams in order to accomplish non-collinear phase matching [5].
This scattered light may arise from the crystal itself due to inhomogeneities or impurities or may
be forced by suitable optics (ground glass plate in front of the sample).
74 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

1 Figure 51: Border plane between two

z [mm]

domains in potassium niobate which

0.5 have different orientation.
0.2
0
0.1
0.5 0.4 0.3 x [mm]
0.2 0.1
y [mm] 0 0

The corresponding momentum diagram is shown in Fig. 52. Again the vectorial phase matching
condition described by Eqs. 7.2 – 7.4 has to be fulfilled.

k2
Figure 52: Momentum diagram for spontaneous non-collinear
Θ’
frequency doubling. Out of the infinite number of scattering
Θ k1’
angles (indicated by the gray vectors) only Θ + Θ0 matches.
k1

As light is scattered in all three-dimensional directions, phase matching now can be achieved for
a multitude of angles Θ + Θ0 around the direction of the fundamental beam. This leads to a cone
of second harmonic light. The cone angle Θ depends on the crystallographic direction and the
respective effective refractive indices. To keep it simple, the fundamental beam is directed along
one of the axes of the index ellipsoid yielding a cone of approximately elliptic shape. A typical
experimental arrangement for the measurement is shown in Fig. 53

Figure 53: Setup for measur-

ing spontaneous non-collinear fre-
quency doubling. The input po-
larization can be varied by half-
wave plate λ/2 and polarizer P.
The beam is slightly focused by
lens L1 onto the crystal K which
can be scanned in two directions
by means of stepping motors SM.
The generated light cone is pro-
jected onto the ground glass plate
S yielding an elliptic ring which is
viewed by a CCD camera.
7 N ON -C OLLINEAR H ARMONIC G ENERATION 75

The cone of second-harmonic light is projected onto a ground glass plate yielding a nearly elliptic
ring which is captured by a video system. The fundamental light is removed by an appropriate
optical filter of type BG18. The ring parameters depend very sensitively on the refractive indices
for the fundamental and the second harmonic light at the position of the focused fundamental light
beam. Thus a two-dimensional topographical characterization of crystals is possible when the
sample is moved perpendicular to the fundamental beam direction. The spatial resolution depends
on the fundamental beam geometry, i. e. the focusing of the Gaussian beam (see box on page 66).
The result of such a measurement, where a sample is two-dimensionally scanned, is usually a large
set of ring images. One can show that it is sufficient to measure the length of one of the principal
axes of the ellipses, thus the amount of data can be drastically reduced. An automatic scheme had
to be developed to do this in a reliable way [6]. Fig. 54 shows some typical ring pictures (left) and
the ellipses calculated by the evaluation program (right).

Figure 54: Spontaneous non-collinear frequency doubling: ring pictures from different positions
of a lithium niobate sample. Left: original video images, right: overlaid with the calculated
ellipses.

Again, two examples may illustrate the application of the technique for materials characteriza-
tion, the homogeneity and composition measurement of a pure lithium niobate crystal and the
characterization of so-called growth striations in Mg-doped lithium niobate.

Homogeneity and composition of lithium niobate A lithium niobate crystal grown near the stoi-
chiometric composition had to be characterized. The crystal had been grown along the z-direction.
For the measurements a small sample was cut out of the grown crystal and was two-dimensionally
scanned with the technique along the z- and the x-axis. From the detected ellipses the refractive
indices and therefrom the crystal composition can be derived. The result is plotted in Fig. 55, a
two-dimensional topography of the crystal composition.
A nearly linear variation of the composition in the growth direction of the crystal is clearly de-
tectable. The figure also gives an impression of the sensitivity of the technique, composition
variations down to approximately 0.01 mole% in the lithium oxide content can be detected.

Growth striations in Mg-doped lithium niobateIn crystals sometimes narrow stripes are visible
which indicate some sort of inhomogeneity. Crystal growers call these striations. Several ex-
 76 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

49.5

49.48
Li−Content [%]

49.46 Figure 55: Homogeneity and

composition of a lithium nio-
49.44
0 bate crystal grown with a lithium
49.42
1 oxide content of approximately
49.4 2 49.5 mole%.
2 3
4 x [mm]
6
z [mm] 8 4

49.5 49.49 49.48 49.47 49.46 49.45

planations are possible: conglomeration of impurities, internal stress, composition variations etc.
Fig. 56 shows the topography of such striations in Mg-doped lithium niobate measured with the
spontaneous non-collinear frequency doubling technique. At the striations small deviations in the
refractive indices are detectable which indicate a corresponding slight variation in the composition
of the crystal.

8.6 Figure 56: Growth striations in Mg-

ϕSNCFD [deg]

doped lithium niobate. The striations

8.5
are found to be perpendicular to the
growth direction (z). The slight vari-
ations in the cone angle indicate cor-
c [mm]

8.4
responding variations in the refractive
indices and in the crystal composition.
Pos. ⊥

0.2
8.3
0
3 2.5 2 1.5 1 0.5 0
Pos. || c [mm]

7.3 Non-Collinear Scattering

In a strict definition, non-collinear scattering is not a real non-collinear harmonic generation

process. However, the experimental results are quite similar. It was described by Kawai et al. [7]
who detected it in strontium barium niobate. If a strong infrared laser is directed onto a crystal of
strontium barium niobate perpendicular to the polar axis (c-axis) non-collinear second-harmonic
light propagating in a plane perpendicular to the c-axis is visible. Fig. 57 gives an illustration.
The effect is only detected in unpoled crystals where needle-like microdomains exist. In the
domains second-harmonic light is generated via the tensor elements d31 , d32 , and d33 . No collinear
7 N ON -C OLLINEAR H ARMONIC G ENERATION 77

Figure 57: Non-collinear scattering. Pictures from left to right: (1) SBN crystal on a rotation
stage. (2) Infrared laser directed along the c-axis (visible due to the sensitivity of the video camera
at 1064 nm). (3) Crystal rotated by 90 ◦ – infrared laser directed perpendicular to the c-axis. (4)
Ditto but infrared light now suppressed by a suitable filter.

phase-matching condition can be fulfilled in SBN due to the small birefringence of the material.
Therefore, no intense collinear harmonic light is generated. Instead a part of the harmonic light
is scattered at the domain boundaries, and – as the domains are directed along the c-axis – this
scattering occurs perpendicular to the c-axis.

7.4 Conical harmonic generation

An interesting mechanism for the generation of harmonic light is the use of higher order nonlin-
earities. This mechanism for Conical Harmonic Generation was described and experimentally
verified in 2002 by Moll et al. [8]. The wave vector geometry for second-harmonic generation via
this mechanism is shown in Fig. 58.

k2
k2’ Figure 58: Wave vector diagram for conical second-harmonic
Θ
generation via a 5th order nonlinear interaction.
k1 k1 k1 k1

Five waves (4 × k1 , k02 ) have to interact to produce a second-harmonic wave k2 . As k02 also has
to be generated by the fundamental pump wave k1 , the whole process can be regarded as paramet-
ric amplification of a signal and an idler beam, k2 and k02 , respectively. In the pump mechanism
an appropriate higher-order nonlinear term has to be included. As usual for parametric ampli-
fication, the theoretical description consists of three coupled equations for the three interacting
waves k1 , k2 , k02 . Generally, the two generated waves may be of different frequencies, most effec-
tive amplification is achieved, however, when the frequencies of signal and idler are identical. A
comprehensive treatment is given in the above cited publication.
78 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

The wave vector geometry in Fig. 58 shows that for the generation of the second-order harmonic
a 5th order nonlinear interaction is responsible. This can be generalized: radiation at the mth
order harmonic can be generated through the use of a (2m+1)-order nonlinearity. The tensor of
the corresponding nonlinear susceptibility is of rank (2m+2), i. e., always of even rank. Thus
this process allows for the generation and amplification of both odd- and even-order harmonics in
all materials, even in isotropic ones. Additionally, this process can always be phase matched in
normal-dispersion materials without the use of birefringence. From the wave vector diagram we
can derive
cos Θ = n(ω)/n(2ω) (7.6)
or – for the generation of the mth order harmonic –

cos Θ = n(ω)/n(mω) . (7.7)

Both equations can always be fulfilled for normal dispersion as in this case n(mω) > n(ω).
In isotropic materials these conditions for Θ lead to circular cones of generated harmonic light.
Fig. 59 shows the experimental results for third-harmonic generation in sapphire.

Figure 59: Experimental spectrum of

conical third-harmonic emission from
sapphire and the corresponding photo-
graph of the output ring (inset) for the
case in which the wavelength of the in-
put pulse is centered at 1500 nm. The
spectral width results from the band-
width of the fundamental pulse. The
cone angle is ≈12° and the conver-
sion efficiency is ≈10-6 (taken from
Ref. [8]).

7.5 Domain-Induced Non-Collinear SHG

A new non-collinear mechanism for the generation of second-harmonic light has been recently
found in strontium barium niobate (SBN) [9]. A circular cone of second-harmonic light is gen-
erated when a fundamental beam of intensive laser light is directed along the crystallographic
c-axis. The corresponding ring projected onto a screen is shown in Fig. 60 (it’s the second image
of Fig. 57 but now with the infrared light suppressed).
The nonzero elements of the SHG tensor of strontium barium niobate derived in section 4.7 show
that no second harmonic wave in c-direction can be expected, no collinear SHG is possible for a
fundamental beam along the c-axis. The light polarization in the ring is radial, the polarization
7 N ON -C OLLINEAR H ARMONIC G ENERATION 79

Figure 60: SHG ring in strontium barium niobate. The fundamental

laser beam is directed along the crystallographic c-axis. In this direc-
tion, no SHG light is visible, instead a circular cone of green light is
visible. The image shown corresponds to the second image of Fig. 57,
the infrared light is suppressed by an appropriate filter.

direction points to the center of the ring (Fig. 61). And it is independent from the polarization
of the fundamental beam. Both facts – radial polarization and no influence of the fundamental
beam’s polarization – conform with the fourfold symmetry around the c-axis.

Figure 61: Polarization of the SHG ring in strontium barium niobate. From left to right: (1)
without analyzer, (2) analyzer horizontal, (3) analyzer diagonal, (4) analyzer vertical.

Several authors have demonstrated that micrometer-sized needlelike domains play an important
role for light scattering and for the type of the phase transition in SBN [7, 10, 11, 12]. These do-
mains are in antiparallel order, the ferroelectric polarization is parallel or antiparallel to the crys-
tallographic c-direction. To prove whether these domains also are responsible for the non-collinear
second-harmonic process, a sample was poled by cooling it down from the high-temperature para-
electric phase with an electric field applied in c-direction. After that the ring structure had van-
ished. This is also a strong indication that higher nonlinearities of odd order [8], discussed in the
preceding section, which are insensitive to poling and the corresponding symmetry aspects, do not
contribute to the effect. Having thus proven that antiparallel ferroelectric domains are the basic
cause for this non-collinear SHG effect, model calculations based on antiparallel domains were
carried out to explain the ring structure.
Plane light waves propagating along the c-direction of SBN contain only electric field components
perpendicular to this direction, E1 and E2 . According to the shape of the SHG tensor for SBN,
these field components produce a second order nonlinear polarization P3 . The sign of P3 depends
on the domain orientation, here indicated by arrows:

P3 (⇑) = d31 E1 E1 + d32 E2 E2 and (7.8)

80 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

P3 (⇓) = −d31 E1 E1 − d32 E2 E2 . (7.9)

For simplicity, all oscillatory factors have been omitted from E and P . E may be assumed to
be monochromatic at frequency ω, then P accordingly is monochromatic at 2ω. The induced
second-harmonic polarization P3 acts as a source for dipolar radiation at this frequency 2ω.
The simplest nontrivial arrangement of domains contains just two antiparallel ordered ones. For
the calculation, the domain sizes were assumed to be in the order of the second-harmonic wave-
length. To compute the far-field behavior, the domains were replaced by suitable dipolar point
sources. The angular intensity distribution due to the interference of the respective dipolar radia-
tion fields is schematically sketched in Fig. 62 for the plane defined by the two dipole vectors.

Figure 62: Angular distribution of the second-

harmonic radiation originating from two antipar-
allel domains in SBN. The exciting wave propa-
gates in c-direction from the left side.

No intensity in forward direction, instead a broad angular intensity distribution around two distinct
angles symmetric to the c-direction is found. The dominant angles are determined by the domain
sizes. Due to the oscillation direction of the dipoles, the polarization of the second-harmonic light
is in the plane shown.
Increasing the number of equally sized domains leads to a narrowing of this angular distribution
similar to the diffraction through an optical grating. Yet in real crystals it cannot be expected that
one deals with ideal equally-sized domains. A generalization consequently has to assume a large
number of domains with a random distribution of sizes. A model calculation on an arbitrarily
chosen domain distribution reveals an angular dependence of the generated second-harmonic light
as shown in Fig. 63.

Figure 63: Angular distribution of the

second-harmonic light intensity arising
from a planar array of 200 randomly-
sized antiparallel ordered domains in
SBN. A part of the domain arrange-
ment is sketched on the left side: c-
direction is horizontal, dark domains
are polarized parallel, light ones an-
−40 −20 0 20 40 tiparallel to this direction, the exciting
Internal Angle Θ [deg]
wave propagates in c-direction.

Again, the polarization of the second-harmonic light is in-plane. Varying the random distribution
of the domain widths varies the random fine structure of the intensity distribution; the common
7 N ON -C OLLINEAR H ARMONIC G ENERATION 81

features, however – no intensity in forward direction and a broad angular distribution starting a
approximately 10°– are maintained. Extending the model to an arrangement of needle-like long
domains means that, in addition to the calculated angular distribution of Fig. 63, strong momentum
conservation has to be obeyed, yielding

k2 = k1 + k01 + kg . (7.10)

Here, kg represents any spatial periodicity present in the domain arrangement, k1 = k01 charac-
terizes the fundamental beam in c-direction, k2 one of the harmonic waves. Due to the random
distribution of domain widths, kg shows up a corresponding reciprocal distribution. The direction
of kg , however, is strictly perpendicular to the c-axis according to the extent of the domains in
c-direction. The momentum geometry for the phase-matching condition of Eq. 7.10 is sketched in
Fig. 64.

k2 kg Figure 64: Wave vector diagram for Eq. 7.10. k1 and k01 are
Θ in c-direction, kg perpendicular to it with a distribution as indi-
cated by the dashed line.
k1 k1’

The angle Θ between fundamental and harmonic wave vectors inside the crystal is defined by

k1 + k10 no (ω)
cos Θ = = e . (7.11)
k2 n (Θ, 2ω)
Using the refractive index data for SBN, Eq. 7.11 yields an internal angle Θ of 17.1°, correspond-
ing to an external angle of 44.8°. This is in excellent agreement with the measured angle of
approximately 45°.
The extension of the model to a three-dimensional arrangement of needle-like long domains with
randomly distributed widths is straightforward. Angular intensity distribution and phase-matching
condition of Eq. 7.10 lead to a cone of second-harmonic light with internal cone angle Θ. In-plane
polarization for all radial directions then accounts for the radial polarization experimentally found
in the ring.

Generalization: Recently, it could be shown, that domain-induced non-collinear SHG is also the
effect responsible for non-collinear scattering described in section 7.3. The mechanism is rather
general, the wave vector diagram for a general geometry is depicted in Fig. 65. The fundamental
beam propagates in the crystal at an angle α from the c-axis of the crystal. A cone of second
harmonic light around the c-axis is generated. The cone angle β has to fulfill the condition

2k1 cos α = k2 cos β (7.12)

or – expressed in refractive indices –

n1 (α) cos α = n2 (β) cos β . (7.13)

82 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

kg
Figure 65: Wave vector diagram for domain-
c induced non-collinear SHG in a generalized
k2 k
1 geometry. The fundamental beam is at an ar-
kg bitrary direction enclosing α with the c-axis,
α
vectors kg show a distribution as indicated but
k1 β are strictly perpendicular to the c-axis, a cone
k2
angle β for the generated harmonic light k2 re-
sults.

For α = 0, this results in β = Θ, Θ of Eq. 7.11, for α = 90° we get the effect shown in section 7.3.
Typical cone sections, projected onto a screen behind the crystal, are shown in Fig. 66.

Figure 66: Cone sections for domain-induced non-collinear SHG: Circle for α = 0 (left), line for
α = 90° (right), ellipse or hyperbola for arbitrary α (middle).

References

[1] A. Reichert, K. Betzler. Induced noncolinear frequency doubling: A new characterization

technique for electrooptic crystals. J. Appl. Phys. 79, 2209–2212 (1996).

[2] U. Schlarb, K. Betzler. Influence of the defect structure on the refractive indices of undoped
and Mg–doped lithium niobate. Phys. Rev. B 50, 751 (1994).

[3] U. Schlarb, K. Betzler. A generalized Sellmeier equation for the refractive indices of lithium
niobate. Ferroelectrics 156, 99 (1994).

[4] K. Kasemir, K. Betzler, B. Matzas, B. Tiegel, T. Wahlbrink, M. Wöhlecke, B. Gather, N. Ru-

binina, T. Volk. Influence of Zn/In dodoping on the optical properties of lithium niobate. J.
Appl. Phys. 84, 5191 (1998).
7 N ON -C OLLINEAR H ARMONIC G ENERATION 83

[5] K.-U. Kasemir, K. Betzler. Characterization of photorefractive materials by spontaneous

noncolinear frequency doubling. Appl. Phys. B 68, 763 (1999).

[6] K.-U. Kasemir, K. Betzler. Detecting Ellipses of Limited Eccentricity in Images with High
Noise Levels. Image & Vision Computing Journal 21, 221–227 (2003).

[7] Satoru Kawai, Tomoya Ogawa, Howard S. Lee, Robert C. DeMattei, Robert S. Feigelson.
Second-harmonic generation from needlelike ferroelectric domains in Sr0.6 Ba0.4 Nd2 O6 sin-
gle crystals. Appl. Phys. Lett. 73, 768 (1998).

[8] K. D. Moll, D. Homoelle, Alexander L. Gaeta, Robert W. Boyd. Conical Harmonic Gener-
ation in Isotropic Materials. Phys. Rev. Lett. 88, 153901 (2002).

[9] Arthur R. Tunyagi, Michael Ulex, Klaus Betzler. Non-collinear optical frequency doubling
in Strontium Barium Niobate. Phys. Rev. Lett. 90, 243901 (2003).

[10] Y. G. Wang, W. Kleemann, Th. Woike, R. Pankrath. Atomic force microscopy of domains
and volume holograms in Sr0.61 Ba0.39 Nd2 O6 :Ce3+ . Phys. Rev. B 61, 3333–3336 (2000).

[11] W. Kleemann, P. Licinio, Th. Woike, R. Pankrath. Dynamic light scattering at domains and
nanoclusters in a relaxor ferroelectric. Phys. Rev. Lett. 86, 6014–6017 (2001).

[12] P. Lehnen, W. Kleemann, Th. Woike, R. Pankrath. Ferroelectric nanodomains in the uniaxial
relaxor system Sr0.61−x Ba0.39 Nd2 O6 :Ce3+
x . Phys. Rev. B 64, 224109 (2001).
84 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

8 Continuous wave solid-state laser systems with intra-cavity second

harmonic generation [1, 2, 3, 4]

8.1 Fundamentals

We will at first recall the key aspects of the laser process. The basic principle of amplification
of a light wave transmitting through a laser medium is shown in Fig. 67, where uin and uout
denote the incoming and outgoing photon flux of the light wave with the relation uout >> uin .
The phenomenon of amplification and its efficiency result from light interaction processes with
the laser medium shortly summarized in the following.

Figure 67: Basic Principle of light amplification. uin and uout denote the incoming and outgoing
flux of the light wave with the relation uout >> uin

8.1.1 Absorption

Resonant excitation of electrons from the ground state E1 into an excited atomic state E2 of the
laser medium occurs if the energy of the incoming photon Eph = ~ω reaches the energetic differ-
ence between both states Eph = ∆E = E2 − E1 as sketched in Fig. 68a). Here, N1,2 denote the

Figure 68: a) Energy model of the absorption process. Resonant excitation of electrons from
the ground state E1 into an excited atomic state E2 occurs if the energy of the incoming photon
Eph = ~ω reaches the energetic difference between both states Eph = ∆E = E2 − E1 . b)
Absorption band centered at the resonance frequency ω0 with the full width at half maximum ∆ν.

number of atoms in the energy state E1,2 per cm3 . As a result of the resonant excitation process
8 CW L ASERS WITH INTRA - CAVITY SECOND HARMONIC GENERATION 85

the intensity of the transmitted light wave decreases, i.e., absorption occurs at the resonance fre-
quency ω0 with a finite full width at half maximum of the absorption band ∆ν (Fig. 68b). The
number of absorbed photons is given by:

Za = N1 · uin · B12 · f (ω) (8.1)

with the Einstein (or probability) coefficient B12 and the function f (ω) taking the frequency de-
pendence into account. Thus the number of transmitted photons Zt can be expressed with the total
number of incoming photons Z0 via Zt = Z0 − Za .

8.1.2 Spontaneous emission

Assuming a finite number of atoms in the electronic state E2 , i.e., N2 6= 0, the process of spon-
taneous emission occurs (Fig. 69a). It is a result of the limited lifetime of excited atoms, which
is reciprocally proportional to the bandwidth of the absorption band τ ∼ 1/∆ω. Typical values
are τ ∼ 10−8 s. The transition of atoms E2 → E1 , and thus N2 → N1 , is accompanied by the
emission of a photon with energy ∆E. A characteristic feature of this process is the emission of
photons into all directions of space. The number of spontaneously emitted photons is described
via Zs = N2 · A with the Einstein coefficient A ∼ 1/τ . The fraction of the Einstein coefficients
for absorption and spontaneous emission is expressed by:

A 2 ~ω
= (8.2)
B12 π c3
with c the speed of light in vacuum.

8.1.3 Induced emission

Induced emission occurs if there is a finite number of atoms in the electronic state E2 , i.e., N2 6= 0,
and a resonant photon is present (Fig. 69b). In this case a photon Eph ie = ∆E is emitted. In

contrast to spontaneous emission the induced emission of a photon occurs in the same direction as
the incoming photon. Thus the photon flux of the incoming wave can be amplified:

Zt = Z0 + Zi = Z0 + N2 · Uin · B21 · f (ω) (8.3)

The energetic balance of a photon flux exposed to a laser medium with N1 , N2 6= 0 thus results
to:
Zt = Z0 + Zi − Za = ∆N · uin · B12 · f (ω) (8.4)
with ∆N = N2 − N1 . Obviously three cases can be distinguished:

• N2 < N1 : ∆N < 0, i.e. depletion of the incoming light wave

• N2 = N1 : ∆N = 0, i.e. unaffected transmission of the light wave

86 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

se = ∆E due to relaxation processes of excited

Figure 69: a) Spontaneous emissions of a photon Eph
atoms. b) Induced emission of a photon Ephie = ∆E by an incoming photon. In both cases N 6= 0
2
is required.

• N2 > N1 : ∆N > 0, i.e. amplification of the incoming light wave.

The latter case commonly is denoted as occupation inversion. In the thermal equilibrium the occu-
pation of the excited state depends on the temperature T according to N2 = N1 exp(−∆E/kB T ),
with the Boltzmann constant kB . Note that N2 ≈ 0 at room temperature since kB T << ∆E. For
very high temperatures (T → ∞) N2 = N1 can be reached resulting in an unaffected transmission
of a light wave through the laser medium. Thus an occupation inversion can not be realized in the
thermal equilibrium at any temperature. In order to overcome this problem laser media offering
an energetic 3- or 4-level system are required.

8.1.4 3-level system

The energetic scheme of a 3-level system, e.g. of a ruby laser, is shown in Fig. 70. An occupation

Figure 70: Energetic scheme of a 3-level system.

p
inversion ∆N = N2 − N1 is reached under intense illumination with light of Eph = E3 − E1 , so
8 CW L ASERS WITH INTRA - CAVITY SECOND HARMONIC GENERATION 87

that light of Eph = E2 − E1 can be amplified. Population of N2 occurs via de-excitation of the
optically excited atomic state E3 → E2 . Thus this process is commonly called optical pumping.
However, the population of each state and especially the occupation inversion is very sensitive to
the intensity W of the pump light as shown in Fig 71. Several characteristic population ratios can

Figure 71: Dependence of the ratio ∆N/N0 on the intensity of the pump light for a 3- and 4-level
system.

be distinguished depending on the intensity:

• W = 0 : N2 = 0 ⇒ ∆N/N0 = −1
• W < W0 : N1 > N2 ⇒ ∆N/N0 < 0
• W = W0 : N1 = N2 ⇒ ∆N/N0 = 0
• W > W0 : N2 > N1 ⇒ ∆N/N0 > 0
• W  W0 : N1 ≈ 0 ⇒ ∆N/N0 = 1

It is obvious that the occupation inversion ∆N/N0 > 0 occurs for intensities > W0 . In contrast,
absorption processes dominate the transmission of the light wave for intensities < W0 .

8.1.5 4-level system

The scheme of a 4-level system, e.g. Nd-YAG laser, is shown in Fig. 72. The key feature of the
4-level system is that E1 is empty in the thermal equilibrium, i.e. occupation inversion is present
as soon as N2 6= 0. This feature is connected with a comparable small lifetime of the atomic states
in E1 . As a result there is no threshold behavior of the ratio ∆N/N0 on the intensity as shown in
Fig. 71.
The efficiency of amplification further depends on the interaction length of the light wave in the
laser media by:  
B12 ∆N
Iout = Iin exp ·l (8.5)
c
88 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

Figure 72: energetic scheme of a 4-level system.

It should be noted, that N2 reaches saturation with increasing intensity of the amplified light wave
and that there is a non-linear dependence of the amplified intensity on the pump intensity as well
as on the interaction length. The gain Γ = Iout /Iin is introduced as measure for the amplification.

8.1.6 Optical resonator

An enhancement of the gain can be reached by using an optical resonator consisting of two mirrors
M as shown schematically in Fig. 73. The incoming light wave is focused by the lenses L into the
laser medium in order to enhance the intensity of the incoming fundamental wave. In dependent

Figure 73: Optical resonator by two mirrors M with the laser medium. The incoming light wave
is focused to enhance the incoming intensity of the fundamental wave.

on the reflectivity of the mirrors M the light wave passes by 1/(1 − R) times through the laser
medium, e.g. with a reflectivity of R = 0.95 an enhancement by a factor of 20 is reached by the
use of the optical cavity.

8.1.7 Pump processes

• Optical pumping. Absorption of (laser) light in the laser medium. Typically found in solid
state and liquid laser systems.
8 CW L ASERS WITH INTRA - CAVITY SECOND HARMONIC GENERATION 89

• Electrical pumping. Gas recharging in gas- and semiconductor lasers

• Chemical pumping. A + B → AB∗ (AB*: excited molecule) or dissociative: AB + hν →
A + B∗ (B*: excited atom)

Fig. 74 displays three common configurations for optical pumping using lamps: a) helix-configuration,
b) elliptic cavity and c) close coupling. For an efficient optical pumping the spectrum of the pump

Figure 74: Optical pumping with lamps a) helix-configuration, b) eliptic cavity and c) close cou-
pling.

source (lamp or laser) should be matched to the absorption spectrum of the laser medium. As an
example Fig. 75a shows the emission spectrum of a Kr-high pressure lamp and 75b the absorption
spectra of the laser media Nd:YAG and Nd:Glass. Absorption bands of the Nd-center occur in the
near-infrared region at about 800 nm and show a broad absorption band when embedded in glass.
Here, the exposure to light of the Kr-high pressure lamp will ensure efficient optical pumping,
whereas light of a semiconductor laser with λ = 808 nm is preferable in Nd:YAG.

Figure 75: a) Emission spectrum of a Kr-high pressure lamp, b) Absorption spectra of the laser
media Nd:YAG and Nd:Glass.

8.2 Cavity design

8.2.1 Optical resonator

In the following we will focus on a 4-level laser system of a Nd:YAG laser medium optically
pumped using a semiconductor laser. Such systems are widely used and designated as diode-
pumped solid-state laser. A typical scheme illustrating a corresponding cavity design is shown
90 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

in Fig. 76. The divergent light of a Ga-Al-As-semiconductor laser (λ = 808 nm)is focused via a
lens into the Nd:YAG laser rod. As a remarkable feature the optical cavity is realized by dielectric
mirrors coated onto the entrance surfaces of the laser rod. A difference in the reflectivity of 99.9 %
and 99.8 % ensures high and low reflector properties such that the emission of laser light occurs
into a preferred direction. According to the energetic scheme of the Nd:YAG 4-level system light
of wavelength λ = 1064 nm is emitted. Typical system specifications are a pump power of 1 -
2 W and infrared light of several 100 mWs. It is noteworthy that this cavity design enforces high

Figure 76: Schematic setup of a diode-pumped Nd:YAG laser cavity. The divergent light of a
Ga-Al-As-semiconductor laser (λ = 808 nm) is focused via a lens into the Nd:YAG laser rod. The
optical cavity is realized by dielectric mirrors coated onto the entrance surfaces of the laser rod
with different reflectivity.

demands to the polishing of the laser rod and to the parallelism of the two entrance surfaces to
each other. Other possibilities for a compact cavity design are the prism and spherical resonator
(confocal as well as concentric) as shown in Fig. 77a and 77b. Open resonators are of advan-
tage to get linearly polarized light. E.g. in Fig. 77c the entrance faces of the laser rod are cut
corresponding to Brewsters law. Internal reflections are suppressed in the in-line configuration
by dielectrically coated surfaces (77d). Further, it is possible to influence the laser light by e.g.
diaphragms, modulators, filters, optical switches, etc. .

Figure 77:

8.2.2 Laser medium

In addition to the optical cavity great demands are required from the Nd:YAG laser-rod itself.
Beyond the most important are:
8 CW L ASERS WITH INTRA - CAVITY SECOND HARMONIC GENERATION 91

• high optical quality: no striations, high optical homogeneity in the refractive index and
absorption coefficient, perfect surfaces
• high optical damage threshold: e.g. cw-laser light up to 1 kW IR at a diameter of 100 µm.
• high conversion efficiency: Nd:YAG e.g. 1-2 %
• high heat flow in order to avoid thermal lens effects
• good preparation and growth conditions in order to get high quality and to reduce costs

8.2.3 Losses

One of the key aspects in the cavity design is the balance between the light amplification Γ and its
losses L. For an efficient laser process the condition Γ > l has to be fulfilled with the threshold
condition Γ − L = 0. Losses are distinguished from 1) the laser rod:

• scattering in the volume or on the surface of the laser rod

• absorption in the volume of the laser rod
• reflection losses at the laser rod entrance faces
• beam distortion due to refraction or diffraction processes at refractive index inhomogeneities

and b) the laser cavity:

• reflection losses and scattering at the mirrors

• absorption losses in the surrounding medium
• coupled-out intensity
• filters, switches, modulators, diaphragms.

8.2.4 Dimensions of the laser rod

The dimensions of the laser rod, i.e. the length l and the diameter d = 2 · r with l  r, are related
via the Fresnel number:
n · r2
F = . (8.6)
λ·l
In order to reduce losses by diffraction the condition F  1 has to be fulfilled. Typical values are
5 < l < 20.0 mm. On the other hand the volume of the rod is decisive for the efficiency of optical
pumping, which is described by the Schawlow-Townes relation for a 4-level system:
P0 V
P = · , (8.7)
P(Γ−L=0) B12 · τ21 · τc
92 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

whereby τc denotes the lifetime of the photons in the laser cavity and P0 the pump power. Typical
values of ζ = P0 /P(Γ−L=0) are ∼ 1000 for Nd:YAG and ∼ 30 for ruby (3-level system).

8.2.5 Estimation of the cavity parameter τc

The measure τc is strongly dependent on the cavity losses and of importance a) to determine
laser losses in order to optimize the cavity design and b) to determine the optimum pump power.
However, τc can not be measured inside the laser cavity. A widely used experimental procedure is
the optical pumping of the laser process with a single light pulse and the subsequent detection of
the kinetics of the out-coupled intensity. The value τc is then determined from the periodicity and
the damping of the retrieved signal as described in the following.
Optical pumping with pulsed light leads to a temporal development of the number of atoms N2 in
the energy level E2 of the Nd:YAG 4-level system and thus of the number of photons Q within the
optical cavity. The laser rod contributes via
dN2 B12 N Q N
=P − − (8.8)
dt V τ21
with τ21 the characteristic lifetime of the spontaneous emission N2 → N1 . The second and third
terms of equation (8.8) account for induced and spontaneous emission, respectively. The temporal
development of the number of photons in the cavity follows
dQ B12 N Q N Q
= + − (8.9)
dt V M τ21 τc
and is enlarged by induced and spontaneous emission (1st and 2nd terms) and is minimized by
the restricted lifetime of photons. The measure M accounts for photons which participate in the
eigenmode of the optical cavity. The equation system is solved with the linear approximation:
V
N = N0 + ; N0 =
B12 τc
Q = Q0 + η; Q0 = M − P τc (8.10)
where  and η are small fluctuations of N0 and Q0 . Here, the power P inside the laser cavity and
the pump power are connected by P = ζ · P0 = ζ · N0 /τ21 . Solution of eq. (8.8) and (8.9) yields:
)   r
η ζt sin ζ −1
∼ exp − (8.11)
τ 2τ21 cos τ21 τc
which represents a harmonic oscillation of period
τc τ21
T 2 = 4π 2 (8.12)
ζ −1
and a damping constant
2τ21
τd = (8.13)
ζ
In the approximation ζ ≈ 1 we get : T 2 = 2π 2 τc τd , so that τc can be determined by the periodicity
T and the damping constant τd of the detected laser intensity.
8 CW L ASERS WITH INTRA - CAVITY SECOND HARMONIC GENERATION 93

8.2.6 Reduction of unwanted Eigenmodes

The suppression of unwanted longitudinal Eigenmodes is related with a cavity of high mechanical
stability. This can be realized using a temperature controlled cavity where all optical elements
including the laser rod are stabilized thermally. Further materials with extreme low extension
coefficients are commonly used (super invar). Unwanted transversal Eigenmodes are suppressed
by introducing diaphragms inside the optical cavity. A birefringence filter, i.e. a combination of
polarizer and retarder wave-plate, is commonly used to get linear polarized laser light with an
extremely small bandwidth. Fig. 78 shows the setup of an optical resonator with a temperature
controlled base plate and laser rod, a birefringence filter BF and a diaphragm D. Note the specific

Figure 78: Setup of an optical resonator with a temperature controlled base plate and laser rod, a
birefringence filter BF and a diaphragm D.

demands for the dielectric coatings of the laser rod (low reflection coating for λ = 1064 nm and λ
= 808 nm) and for the transmission of the high reflector (high transmission for λ = 808 nm, high
reflection for λ = 1064 nm). Typical specifications of such laser systems are a single pass power
of 20-40 mW of infrared light (λ = 1064 nm) with a pump power of P (λ = 808 nm) = 2 W and
dimensions of the laser rod of 10 mm length and 3 × 3 mm1 surface area. An optimum cavity
design leads to an intra-cavity power of 20 - 50 W and of ≈ 500 mW extra-cavity.

8.2.7 Cavity design with intra-cavity second harmonic generation

The next step is the design of a Nd:YAG laser system with intra-cavity second harmonic generation
to get intense continuous-wave laser light of wavelength λ = 532 nm. The demands for the design
of an optical cavity with intra-cavity second harmonic generation (SHG) are

• Two independent adjustable beam waists, one localized in the laser rod and one in the non-
linear crystal for SHG. The dimensions of the beam waist in the laser rod has to be adapted
for the beam waist of the laser light for optical pumping. The beam waist within the non-
linear crystal should be optimized for a high intensity under consideration of the crystal
length.

• A high mechanical stability of the optical cavity over a long timescale.

• Linear polarized laser light.

Further the redesign of the Nd:Yag optical cavity should account for the following aspects
94 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

• losses due to the non-linear crystal, especially losses due to SHG

• dielectric coatings for the non-linear crystal (λ = 1064 nm and λ = 532 nm)

• transmission of the low reflector (high transmission at λ = 532 nm)

• refractive index of the non-linear crystal influences the beam waist intra-cavity.

With respect to these demands and aspects it should be stressed that intra-cavity second harmonic
generation is inevitably necessary to get intense continuous wave laser light. The power of the
frequency doubled beam is ∼ I10642 ic
and I1064 ec , where ic and ec denote intra- and extra-
>> I1064
cavity, respectively. In contrast, SHG with pulsed laser light is commonly realized in an extra-
cavity configuration.
The disadvantage of intra-cavity SHG is two-folded: a) a more complicated design of the optical
cavity is enforced and an exchange of the non-linear crystal is impossible, e.g. for purposes of
optimization, b) the demands to the non-linear crystal are enormous especially due to the extremely
high power of the fundamental wave (high risk for optically induced mechanical damage). Fig. 79

Figure 79: principle setup for an optical cavity with intra-cavity SHG.

shows the principle setup for an optical cavity with intra-cavity SHG. All laser properties are
restricted for the generation of infrared light at λ = 1064 nm, i.e., the cavity does not amplify light
of λ = 532 nm. The emission of the frequency doubled laser light occurs in both directions, but
is blocked by the polarizer of the birefringence filter (the orientation of the electric field vector
for type I and type II phase matching are different to the electric field vector of the fundamental
wave). The intensity of the visible light is comparable small in such systems, e.g., with a pump
power of 2 W and an intra-cavity power of 10 - 50 W a laser beam with ≈ 150 mW at λ = 532 nm
is generated in the output.

8.2.8 Losses by the non-linear crystal

As already mentioned intra-cavity SHG represents an additional loss and thus the demand for a
large SHG coefficient is questionable. The condition for the threshold of the laser process with
2
intra-cavity SHG now follows the connection: Γ − L − (K · P1064 ) with P532 = K · P1064 and the
non-linear coupling coefficient

K = KIR · li · kIR · h(σ, ζ) · 107 (8.14)

8 CW L ASERS WITH INTRA - CAVITY SECOND HARMONIC GENERATION 95

Here, kIR = 2πnIR /λIR denotes the wave vector of the infrared laser beam, li the interaction
length of the fundamental and harmonic waves and KIR is a material specific constant, e.g. KIR =
128π 2 ω 2 /c3 n2IR nV IS ·d32 for Ba2 NaNb5 O15 . The function h(σ, ζ) is given by the theory of Boyd
and Kleinmann and takes diffraction, double refraction and absorption processes into account.
Here σ = 1/2b∆K is connected to the phase matching parameter ∆K and ζ = li /b to the
confocal parameter b = ω02 /kIR with the beam waist ω0 . The dependencies of PSHG on the
coupling coefficient and of the coupling coefficient on the beam waist are shown in Fig. 80.

Figure 80: PSHG as a function of the coupling coefficient K and of K on the beam waist ω0 .

8.2.9 Selection of the non-linear crystal

The selection of an adequate non-linear crystal is restricted by

• a large SHG coefficient

• the refractive index and the dispersion

• the optical transmission range

• the phase matching properties

• the optical damage threshold

• the optically induced mechanical damage threshold

• the optical homogeneity of refractive index and absorption coefficient

• the hardness, chemical stability.

Some of the commonly used non-linear crystals are given in tab. 6.

KTiOPO4 is widely used for intra-cavity second harmonic generation of cw-laser light. An im-
portant feature is its pronounced birefringence, which is used in combination with a polarizer as
birefringence filter.
96 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

Transparency Damage threshold FOM

range (nm) (GW/cm2 )
β-BaB2 O4 198-3300 10 15
Ba2 NaNb5 O15 0.001
KH2 PO4 200-1500 0.5 1
LiB3 O5 2 1
LiNbO3 0.02
LiIO3 300-5500 0.05 50
KTiOPO4 350-4500 1 215
KNbO3 410-5000 0.35 1755
CsD2 AsO4 1660-2700 0.5 1.7
(NH)2 CO 210-1400 1.5 10.6
LAP 220-1950 10 40
m-NA 500-2000 0.2 60
MgO-LiNbO3 400-5000 0.05 105
POM 414-2000 2 350
MAP 472-2000 3 1600
COANP 480-2000 4690
DAN 430-2000 5090
PPLiNbO3 400-5000 0.05 2460

Table 6: Properties of non-linear crystals. FOM is determined by (d2 /n3 )(EL/λ)∆θ2 . LAP:
L-arginine phosphate monohydrate, m-NA: meta nitroaniline, POM: 3-methyl-4-nitropyridine
N-Oxide, MAP: methyl (2,4-diinitrophenyl) aminopropanoate, COANP: 2N-cyclooctylamino-5-
nitropyridine

References

[1] O. Svelto. Principles of Lasers. Plenum Press, New York und London, 1998.

[2] A. E. Siegmann. Lasers. Unversity Science Books, Mill Valley, California, 1986.

[3] W. Koechner. Solid State Engineering. Springer Verlag, New York, 1998.

[4] A. Yariv. Quantum Electronics. John Wiley and Sons, New York, 1967.
A M ATRICES FOR SYMMETRY OPERATIONS 97

A Matrices for symmetry operations

selection of matrices for point group operations in orthogonal systems

at origin [0 0 0]
   
1 0 0 −1 0 0
1(E)  0 1 0  1̄(i)  0 −1 0 
   

0 0 1 0 0 −1
along [1 0 0]
   
1 0 0 −1 0 0
2(C2 )  0 −1 0  2̄ = m  0 1 0 
   

0 0 −1 0 0 1

   
1 0 0 −1 0 0
4(C4 )  0 0 1  4̄ = (S43 )  0 0 −1 
   

0 −1 0 0 1 0

42 (C42 ) = 2(C2 ) 4̄2 (S42 ) = 2(C2 )

   
1 0 0 −1 0 0
43 (C43 )  0 0 −1  4̄3 = (S4 )  0 0 1 
   

0 1 0 0 −1 0

along [0 1 0]
   
−1 0 0 1 0 0
2(C2 )  0 1 0  2̄ = m  0 −1 0 
   

0 0 −1 0 0 1

   
0 0 −1 0 0 1
4(C4 )  0 1 0  4̄ = (S43 )  0 −1 0 
   

1 0 0 −1 0 0

42 (C42 ) = 2(C2 ) 4̄2 (S42 ) = 2(C2 )

   
0 0 1 0 0 −1
43 (C43 )  0 1 0  4̄3 = (S4 )  0 −1 0 
   

−1 0 0 1 0 0
98 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

along [0 0 1]
   
−1 0 0 1 0 0
2(C2 ) 0 −1 0  2̄ = m 0 1 0 
   
 
0 0 1 0 0 −1
   
0 −1 0 0 1 0
4(C4 ) 1 0 0  4̄ = (S43 ) −1 0 0 
   
 
0 0 1 0 0 −1

42 (C42 ) = 2(C2 ) 4̄2 (S42 ) = 2(C2 )

   
0 1 0 0 −1 0
43 (C43 )  −1 0 0  4̄3 = (S4 )  1 0 0 
   

0 0 1 0 0 −1

along [1 1 0]
   
0 1 0 0 −1 0
2(C2 )  1 0 0  2̄ = m  −1 0 0 
   

0 0 −1 0 0 1

along [1 0 1]
   
0 0 1 0 0 −1
2(C2 )  0 −1 0  2̄ = m  0 1 0 
   

1 0 0 −1 0 0

along [0 1 1]
   
−1 0 0 1 0 0
2(C2 )  0 0 1  2̄ = m  0 0 −1 
   

0 1 0 0 −1 0

along [1 1̄ 0]
   
0 −1 0 0 1 0
2(C2 )  −1 0 0  2̄ = m  1 0 0 
   

0 0 −1 0 0 1
A M ATRICES FOR SYMMETRY OPERATIONS 99

along [1̄ 0 1]
   
0 0 −1 0 0 1
2(C2 )  0 −1 0  2̄ = m  0 1 0 
   

−1 0 0 1 0 0

along [0 1 1̄]
   
−1 0 0 1 0 0
2(C2 )  0 0 −1  2̄ = m  0 0 1 
   

0 −1 0 0 1 0

along [1 1 1]
   
0 0 1 0 0 −1
3(C3 )  1 0 0  3̄(S65 )  −1 0 0 
   

0 1 0 0 −1 0

   
0 1 0 0 −1 0
32 (C32 )  0 0 1  3̄5 (S6 )  0 0 −1 
   

1 0 0 −1 0 0

3̄2 (S64 ) = 32 (C32 )

3̄3 (S63 ) = 1̄(i)
3̄4 (S62 ) = 3(C3 )

along [1̄ 1 1]
   
0 −1 0 0 1 0
3(C3 )  0 0 1  3̄(S65 )  0 0 −1 
   

−1 0 0 1 0 0

   
0 0 −1 0 0 1
32 (C32 )  −1 0 0  3̄5 (S6 )  1 0 0 
   

0 1 0 0 −1 0

3̄2 (S64 ) = 32 (C32 )

3̄3 (S63 ) = 1̄(i)
3̄4 (S62 ) = 3(C3 )
100 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

along [1 1̄ 1]
   
0 −1 0 0 1 0
3(C3 )  0 0 −1  3̄(S65 )  0 0 1 
   

1 0 0 −1 0 0

   
0 0 1 0 0 −1
32 (C32 )  −1 0 0  3̄5 (S6 )  1 0 0 
   

0 −1 0 0 1 0

3̄2 (S64 ) = 32 (C32 )

3̄3 (S63 ) = 1̄(i)
3̄4 (S62 ) = 3(C3 )

along [1 1 1̄]
   
0 1 0 0 −1 0
3(C3 )  0 0 −1  3̄(S65 )  0 0 1 
   

−1 0 0 1 0 0

   
0 0 −1 0 0 1
32 (C32 )  1 0 0  3̄5 (S6 )  −1 0 0 
   

0 −1 0 0 1 0

3̄2 (S64 ) = 32 (C32 )

3̄3 (S63 ) = 1̄(i)
3̄4 (S62 ) = 3(C3 )

B A tiny group theory primer

Definition 1: Group
A group G is a set of elements together with a binary composition called a product such that

(i) the product of any two elements in the group is defined and is a member of the group: if
A, B ∈ G then A · B ∈ G

(ii) the product is associative: A · (B · C) = (A · B) · C for all A, B, C ∈ G,

(iii) there exists a unique identity E in the group: E · A = A · E = A for all A ∈ G, and
B A TINY GROUP THEORY PRIMER 101

(iv) every element has a unique inverse element: given A ∈ G there exists a unique element A−1
such that A · A−1 = A−1 · A = E.

Definition 2: Order of a group

The number |g| of elements in a group, G, is called the order of the group.
Definition 3: Order of an element
The order of an element A ∈ G is the least positive integer s such that AS = E.
The full information of a group is given by its multiplication table that means a quadratic table of
all products of two elements with all elements in the first row and the first column. Usually not all
|g| elements are needed to construct the multiplication table. There exist relations between several
elements. For example some elements are simply formed by operations like those in definition
3. In general, if every element of G is expressible as a product of a smaller number of distinct
elements, we call those elements generating elements (generators) of G. The choice of generators
is not unique!
Definition 4: Abelian group
If G is a group and G1 · G2 = G2 · G1 für alle G1 , G2 ∈ G then it is called an Abelian group.
Definition 5: Cyclic group
If all elements of G can be expressed by products of one distinct element, then it is called a cyclic
group. All cyclic groups are Abelian groups.
Definition 6: Conjugate elements
Two elements G1 , G2 ∈ G are said to be conjugate if there exists an element G ∈ G such that
G2 = G · G1 · G−1 .
Definition 7: Subgroup
A subset H of a group G that is itself a group is called a subgroup of G
Definition 8: Invariant subgroup
If H is a subgroup of G such that , for all G ∈ G and for all H ∈ H, G · H · G−1 ∈ H, then H is
said to be an invariant subgroup of G.
Definition 9: Direct product of groups
Let G be a group with subgroups H and K such that

(i) if H ∈ H, K ∈ K then H · K = K · H

(ii) all G ∈ G can be expressed in the form G = H · K

(iii) and if the intersection H ∩ K consists only of the identity, then G is called the direct product
of H and K and we write G = H × K = K × H. This product is sometimes called the outer
direct product.

Definition 10: Matrix representation

Let G be a finite group of order |g| with elements g ∈ G, and let a square matrix D(g) be associated
102 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

with each group element g. If the matrices satisfy

D(gi ) D(Gj ) = D(Gk )

for the corresponding relation of the group elements gi gj = gk , then the set of matrices is called a
representation. The dimension of the representation is equal to the dimension of the matrix.
Definition 11: Direct sum of representation
With two representations D(1) and D(2) of a group G, a larger representation, the direct sum, is
constructed according to " #
D(1) (g) 0
D(g) =
0 D(2) (g)
If no equivalence transformation exists which allows to construct a block-diagonal structure with
elements of smaller dimension, then D(g) is called an irreducible representation , otherwise a
reducible one .
Definition 12: Trace and character
The trace of a matrix D is the sum of its diagonal elements, often written tr D. The character of a
matrix group D is the function χ defined on all elements D(g) such that χ(D(g)) = tr (D(g)).

C Some completions for point groups

The Table 7 of the appendix C contains 32 point groups with their notation (Schönflies and In-
ternational). In addition the classes of symmetry elements and a suggestion for generators are
given.
The figures 81 and 82 show objects which reflect the symmetry of each point group.
C S OME COMPLETIONS FOR POINT GROUPS 103

Int. Int. Schönflies symmetry- generating pace-

short full operation elements groups

triclinic
1 1 C1 E E 1
1̄ 1̄ S2 (Ci ) Ei i 2

monoclinic
2 2 C2 E C2 C2 3-5
m m C1h (Cs ) E σh σh 6-9
2
2/m m C2h E C2 i σh i C2 10 - 15

orthorhombic
222 222 D2 (V ) E C2 C20 C200 C2 C20 16 - 24
mm2 mm2 C2v E C2 σv σv0 C2 σv 25 - 46
mmm 2 2 2
mmm D2h (Vh ) E C2 C20 C200 i σh σv0 σv00 i C2 C20 47 - 74

tetragonal
4 4 C4 E 2C4 C2 C4 75 - 80
4̄ 4̄ S4 E 2S4 C2 S43 81 - 82
4
4/m m C4h E 2C4 C2 i 2S4 σh i C4 83 - 88
422 422 D4 E 2C4 C2 2C20 2C200 C2 C4 89 - 98
4mm 4mm C4v E 2C4 C2 2σv 2σd C4 σv 99 - 110
4̄2m 4̄2m D2d (Vd ) E C2 2C20 2σd 2S4 C2 S43 111- 122
4/mmm 4 2 2
mmm D4h E 2C4 C2 2C20 2C200 C20 C4 123 - 142
i 2S4 σh 2σv 2σd
trigonal (rhombohedral)
3 3 C3 E 2C3 C3 143 - 146
3̄ 3̄ S6 (C3i ) E 2C3 i 2S6 i C3 147 - 148
32 32 D3 E 2C3 3C20 C20 C3 149 - 155
3m 3m C3v E 2C3 3σv σv C3 156 - 161
3̄m 2
3̄ m D3d E 2C3 3C20 i 2S6 3σd i C20 C3 162 - 167

hexagonal
6 6 C6 E 2C6 2C3 C2 C2 C3 168 - 173
6̄ 6̄ C3h E 2C3 σh 2S3 C3 σh 174
6
6/m m C6h E 2C6 2C3 C2 i C2 C3 175 - 176
i 2S3 2S6 σh
622 622 D6 E 2C6 2C3 C2 3C20 3C200 C2 C20 C3 177 - 182
6mm 6mm C6v E 2C6 2C3 C2 3σv 3σd C2 σv C3 183 - 186
6̄m2 6̄m2 D3h E 2C3 3C20 σh 2S3 3σv C2 σh C3 187 - 190
6/mmm 6 2 2
mmm D6h E 2C6 2C3 C2 3C20 3C200
i 2S3 2S6 σh 3σv 3σd i C20 C2 C3 191 - 194
cubic
23 23 T E 8C3 3C2 C2 C3 195 - 199
2
m3̄ m 3̄ Th E 8C3 3C2 i 8S4 3σh i C2 C3 200 - 206
432 432 O E 8C3 3C2 6C20 6C4 C3 C4 207 - 214
4̄3m 4̄3m Td E 8C3 3C2 6σd 6S4 C3 S43 215 - 220
m3̄m 4 2
m 3̄ m Oh E 8C3 3C2 6C20 6C4
i 8S4 3σh 6σd 6S4 i C3 C4 221 - 230

Table 7: Notation and symmetry elements of the 32 point groups

104 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

Figure 81: Part 1 of the objects of the 32 point groups

C S OME COMPLETIONS FOR POINT GROUPS 105

Figure 82: Part 2 of the objects of the 32 point groups

 106 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

D Some completions for space groups

In this part of appendix D we collect some useful information on notations and conventions.

crystal system direction of symmetry element

primary secondary tertiary
triclinic none
monoclinic [010]
orthorhombic [100] [010] [001]
tetragonal [001] [100] / [010] [110]
hexagonal / trigonal [001] [100] / [010] [120] / [11̄0]
cubic [100] / [010] / [001] [111] [110]

Table 8: Conventions for directions of symmetry elements in the seven crystal systems

We give some more information on conventions and some examples.

Cubic - The secondary symmetry symbol will always be either 3 or 3̄ (i.e. Ia3, Pm3m, Fd3m)

Tetragonal - The primary symmetry symbol will always be either 4, 4̄, 41 , 42 or 43 (i.e. P41 21 2, 14/m,
P4/mcc)

Hexagonal - The primary symmetry symbol will always be a 6, 6̄, 61 , 62 , 63 , 64 or 65 (i.e. P6mm,
P63 /mcm)

Trigonal - The primary symmetry symbol will always be a 3, 3̄, 31 or 32 (i.e P31m, R3, R3c, P312)

Orthorhombic - All three symbols following the lattice descriptor will be either mirror planes, glide planes,
2-fold rotation or screw axes (i.e. Pnma, Cmc21 , Pnc2)

Monoclinic - The lattice descriptor will be followed by either a single mirror plane, glide plane, 2-fold
rotation or screw axis or an axis/plane symbol (i.e. Cc, P2, P21 /n)

Triclinic - The lattice descriptor will be followed by either a 1 or a 1̄.

T EXTBOOKS ON N ONLINEAR O PTICS 107

Textbooks on Nonlinear Optics

G. C. Baldwin. An Introduction to Nonlinear Optics. Plenum Press, New York, 1969.

P. N. Butcher and D. Cotter. The Elements of Nonlinear Optics. Cambridge University Press,
Cambridge, 1991.

N. Bloembergen. Nonlinear Optics. Benjamin, New York, 1965.

R. W. Boyd. Nonlinear Optics. Academic Press, Boston, 1992.

P. Meystre and M. Sargent III. Elements of Quantum Optics. Springer, Berlin/Heidelberg, 1991.

Herbert Rabin and C. L. Tang, editors. Quantum Electronics, Volume I: Nonlinear Optics. Acad-
emic Press, Inc., 1975.

Y. R. Shen. The Principles of Nonlinear Optics. John Wiley & Sons, Inc., 1984.

A. Yariv. Quantum Electronics. John Wiley and Sons, New York, 1967.

Frits Zernike and John E. Midwinter. Applied Nonlinear Optics. John Wiley & Sons, Inc., 1973.
108 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

Index
3-level system, 86 diamond, 3
4-level system, 87 diode-pumped solid-state laser, 89
direct sum, 102
absorption, 84 dispersion, 49
amplification, 84 Domain-Induced Non-Collinear SHG, 78
anharmonic oscillator, 40 domains, 79
attosecond pulses, 57 ferroelectric, 72
axis
screw, 19 effective d, 51
Einstein coefficient, 85
Bass, 1 electrical pumping, 89
beam waist, 66 EUV, 55
biaxial, 48 extreme ultraviolet, 55
birefringence, 48, 49
Boyd, 67 far field, 66
Brillouin zone, 13 ferroelectric domains, 72
figure of merit, 96
calcite, 52 FOM, 96
cavity design, 89 Franken, 1
cavity parameter, 92
centrosymmetric, 62 GaAlAs, 90
character, 102 gain, 88
chemical pumping, 89 gases
classification technique, 60 harmonic generation, 55
closed aperture, 68 Gaussian beam, 66
coherence length, 46 beam waist, 66
confocal parameter, 66 confocal parameter, 66
conical harmonic generation, 77 far field, 66
conjugate elements, 101 near field, 66
contraction, 44 Rayleigh length, 66
crystal structure, 3 glide
crystal system, 7 axial, 20
cubic, 7 diagonal, 20
hexagonal, 7 diamond, 20
monoclinic, 6 plane, 19
orthorhombic, 7 group
tetragonal, 7 Abelian, 101
triclinic, 6 centrosymmetric, 16
CsCl, 2 cyclic, 15, 101
dihedral, 15
d-tensor, 44 generator, 101
damage threshold, 96 group, 15
I NDEX 109

holohedral, 17 non-collinear harmonic generation, 70

madnetic, 36 non-collinear scattering, 76
order of, 101 nonlinear absorption, 67
space, 17 nonlinear crystals, 96
subgroup, 101 nonlinear refraction, 67
notation
harmonic generation Herman-Mauguin, 20
non-collinear, 70 International, 4
Helmholtz equation, 66 Schönflies, 4
HHG, 55 Wyckoff, 24
higher harmonics, 55
holohedry, 16 occupation inversion, 86
open aperture, 68
index ellipsoid, 47 operation
symmetry, 48 point symmetry, 3
indices operator
contraction, 44 active, 3
induced emission, 85 passive, 3
induced non-collinear frequency doubling, 70 optic axes, 48
integrating sphere, 61 optical pumping, 87, 88
International System, 4 optical resonator, 88
International Tables, 22 oscillator
intra-cavity second harmonic generation, 93 anharmonic, 40
Kleinman, 43, 67 P-Lattices, 7
KTiOPO4 , 95 paraxial approximation, 66
Kurtz, 60 particle size, 61
periodically poled, 50
L’Huillier, 56
phase matchable, 62
lattice, 2
phase matching, 39, 45, 47, 49
Bracais, 17
non critical, 54
centering, 8
type I, 50
Laue pattern, 16
type II, 50
lithium niobate, 71, 75
phase mismatch, 46
losses, 91
point
Maker fringes, 63 lattice, 2
Maxwell equations, 37, 53 point group
microdomains, 76 isogonal, 18
Moll, 77 polarization
multiplication table, 101 acoustic, 39
electronic, 39
Nd-YAG laser, 87 ionic, 39
near field, 66 linear, 37
non critical phase matching, 54 mechanisms, 39
non phase matchable, 62 nonlinear, 38, 40
110 U NIVERSITY OF O SNABR ÜCK IG RADUATE C OLLEGE 695

orientational, 39 Spontaneous Non-Collinear Frequency Dou-

populational, 40 bling, 73
redistribution, 40 Strontium Barium Niobate, 43
poling strontium barium niobate, 76, 78
periodic, 50 structure
potassium niobate, 72 unfilled tungsten-bronze, 26
powder technique, 60 susceptibility
product linear, 37
direct, 101 nonlinear, 38
outer direct, 101 symmetry
index ellipsoid, 48
quasi phase matching, 50 Kleinman, 43
effective d, 51 permutational, 43
higher order, 52 structural, 42
Rayleigh length, 66 tensor, 30
representation axial, 30
block-diagonal, 102 c-tensor, 36
direct sum, 102 i-tensor, 36
irreducible, 102 polar, 30
matrix, 102 rank, 30
reducible, 102 topography, 71, 75
resonant excitation, 84 transparency range, 96
rotaion type I phase matching, 50
order of, 4 type II phase matching, 50
rotation
improper, 5 Ulbricht sphere, 61
order of, 4 uniaxial, 48
proper, 4 unit cell, 5
pure, 4 primitive, 5
screw rotation, 19
vapor transport equilibration, 72
ruby laser, 86
vector
SBN, 43, 76, 78 primitive translation, 5
Schawlow-Townes relation, 91 VTE, 72
Schönflies System, 4
walk-off, 52
second-harmonic generation, 45
walk-off angle, 54
Seitz-Operator, 17
wedge geometry, 65
SHG tensor, 44
Wigner-Seitz cell, 12
slab geometry, 63
space group X-ray generation, 57
nonsymmorphic, 18
notation, 20 Z-scan, 67
symmorphic, 18 closed aperture, 68
spontaneous emission, 85 open aperture, 68