Crystal Defects and Crystalline Interfaces

W. Bollmann

Crystal Defects
and Crystalline Interfaces

With 158 Figures

and a Set of Moire-Models

Springer-Verlag Berlin Heidelberg GmbH

w. BOLLMANN, Dr. SC. nat., Dipl. Phys., Battelle Institute, Advanced
Studies Center, Geneva, and Privatdozent, Eidgen6ssische Technische
Hochschule, Ziirich, Switzerland

ISBN 978-3-642-49175-7 ISBN 978-3-642-49173-3 (eBook)

DOI 10.1007/978-3-642-49173-3

Illustration on the dust cover: Dislocations and grain boundaries in stainless steel, taken with the
1500 kV electron microscope at Toulouse, France, by G. Dupouy and F. Perrier.
Tbis work is subiect to copyright. AII rights are reserved, whether the whole or part of tbe material is con-
cemed, specifically those of translation, reprinting, re~use of illustrations, broadcasting, reproduction by
photocopying machine or similar means, and storage in data banks.
Under § 54 of the German Copyright Law where copies are roade for otber tban private use, a fee is payable
to the publisher, the amount of the fee to be determined by agreement with the publisher.
<1:> by Springer-Verlag Berlin Heidelberg 1970. Library of Congress Catalog Card Number 77-124069.
Originally published by Springer-Verlag Berlin Heidelberg in 1970
Softcover reprint of the hardcover 1st edition 1970
The use of general descriptive names, trade names, trade marks, etc. in this publications, even if the fonner
are not especially identified, is not be taken as a sign that such names, as understood by the Trade Marks
and Merchandise Marks Act, may accordingly be used freely by anyone. Printed by Universitătsdruckerei
H. Stiirtz AG, Wiirzburg. Title No. 1700.
To my Mother and my Wife
 Preface
It is nonnal for the preface to explain the motivation behind the
writing of the book. Since many good books dealing with the general
theory of crystal defects already exist, a new book has to be especially
justified, and here its main justification lies in its treatment of crystal-
line interfaces. About 1961, the work of the author, essentially based on
the fundamental work of Professor F. C. Frank, started to branch away
from the main flow of thought in this field and eventually led to a
general geometrical theory which is presented as a whole for the first
time in this book. Although nearly all that is presented has already
been published in different journals and symposia, it might be difficult
for the reader to follow that literature, as a new terminology and new
methods of analysis had to be developed.
Special emphasis is given to discussion and many diagrams are
included in order that a clear view of the basic concepts be obtained.
Intennediate summaries try to bring out the main points of the chapters.
Instead of specific exercises, general suggestions for them are given. The
part up to chapter 9 is considered more or less as introductory, so that
the book can be studied without specific knowledge of crystals and
crystal defects. The presentation of that part developed out of lectures
given by the author at the Swiss Federal Institute of Technology (ETH)
in Zurich. As metallurgists and mineralogists are often not familiar with
linear algebra, which is the mathematical basis of the interface theory,
the procedures involved in this technique are collected in the Appendix.
Many of the ideas contained in the book originated from observations
made on moire models and the subsequent fonnulation of these observa-
tions in mathematical terms. In order that the reader may participate
in this experience, a set of such models is presented with the book.
Now the work is finished, it is my great pleasure to express my
gratitude first to the Battelle Institute (BI) which by its financial sup-
port made this work possible, and in particular to Dr. B. D. Thomas,
fonner president of the Battelle Memorial Institute, Dr. H. Thiemann,
General Director of the Battelle Laboratories, Geneva and Dr. F. I. Mil-
ford, Director of Physical Sciences of BI, for their moral support of this
work. Much inspiration came from discussions with my friend and fonner
colleague Prof. D. G. Brandon, and I would like to express my special
VIII Preface

thanks to him. I should also like to thank Mr. P. Fontaine, my col-

laborator for the last fourteen years, who prepared many of the dia-
grams appearing in this book, including the precision drawings of the
moire models, Dr. and Mrs. E. Anderson for their contribution to the
translation of the original German manuscript and Mrs. R. Anderson
for the typing of this manuscript. Finally, I should like to thank the
many others, not mentioned by name here, without whose contributions
the book would not have become what it is, especially the publisher and
his collaborators.

Geneva, July 1970 w. Bollmann

 Contents
Chapter 1. Introduction . . . . . . . . . . 1
1.1. The Significance of Crystal Defects . 1
1.2. The Nature of Crystal Defects . . . 1
Chapter 2. General Aspects of the Structure of Crystals 3
2.1. Outline . . . . . . . . . . . . . . . 3
2.2. Mathematical Description of the Crystal 5
2.3. Classification of Crystal Structures 11
Chapter 3. Some Special Crystal Structures 13
3.1. Outline. . . . . . . . . . . . 13
3.2. Close-Packed Structures . . . . 13
3.3. The Face-Centred Cubic Structure 17
3.4. The NaCl-Structure. . . . . . . 19
3.5. The Diamond Structure. . . . . 20
3.6. The Body-Centred Cubic Structure. 22
3.7. The Close-Packed Hexagonal Structure 25
3.8. The Graphite Structure. . . . . . . 27
3.9. Irregularities in the Stacking Sequence 28
3.10. Energy of Stacking Faults . . . . . 33
3·11. Fibrous Structures. . . . . . . . . 34
Chapters 2-3. Summary and Discussion. 35
Chapter 4. Point Defects . . . . . . . . . . . . 37
4.1. General Considerations of the Field of Poi!1t Defects 37
Chapter 5. The Individual Dislocation: Geometrical Basis. . 41
5.1. Outline . . . . . . . . . . . . . . . . . . . . 41
5.2. Definition of the Dislocation Line and the Burgers
Vector . . . . . . . . . . . 41
5.3. Relative Orientation of b and I 45
5.4. Motion of Dislocations . . . . 47
5.5. Dislocations and Moire-Figures . 49
Chapter 6. Interaction Between a Dislocation and its Sur-
roundings . . . . . . . . . . . 51
6.1. Outline . . . . . . . . . . . . 51
6.2. Representation of the Stress Field 51
x Contents

6.3. Forces Acting on a Dislocation 53

6.4. Energy of a Dislocation. . . . 55
6.5. The Line Tension . . . . . . . 58
6.6. The Strain Field of a Dislocation. 61
6.7. The Stress Field of a Dislocation . 63
Chapters 5-6. Summary and Discussion . 67
Chapter 7. The Interaction of Dislocations . . . . 70
7.1. Outline . . . . . . . . . . . . . . . 70
7.2. Interaction Between two Parallel Screw Dislocations 70
7·3· Interaction Between two Parallel Edge Dislocations 71
7.4. Dislocation Reactions. . . . . . . . . . . . . . 72
7.5. Dislocation Reactions in Special Crystal Structures. 75
Chapter 8. Partial Dislocations . . . . . . . . . . . . . . 78
8.1. Outline . . . . . . . . . . . . . . . . . . . . 78
8.2. Partial Dislocations in the Face-Centred Cubic Struc-
ture . . . . . . . . . . . . . . . . . . . . . 78
8.3. The Sequence of Partial Dislocations . . . . . . . 81
8.4. Transformation of the Face-Centred Cubic - into the
Close-Packed Hexagonal Structure. . . . . . . . 87
Chapter 9. Dualistic Representation of Dislocation Reactions 98
9.1. Introduction. . . . . . . . . . . . . 98
9.2. The Significance of the b-Net . . . . . . . . . . 101
9·3· The Prediction of Dislocation Reactions. . . . . . 102
9.4. Dislocation Reactions Involving Partial Dislocations 107
9·5. Thompson's Notation of the Burgers Vectors . . . . 109
Chapters 7-9. Summary and Discussion. . . . . . 110
Chapter 10. Short Description of the Relations Between Dis-
locations and Point Defects, and Dislocation Dy-
namics . . . . . . . . . . . 113
10.1. Dislocations and Point Defects . . . . . . " 113
10.2. Dislocation Dynamics. . . . . . . . . . . . 117
Chapter 11. Dislocation Networks - Subgrain Boundaries 118
11.1. Introduction. . . . . . . . . . . . 118
11.2. Frank's Formula. . . . . . . . . . 119
11. 3. Classification of Subgrain Boundaries . 124
11.4. "Special" Subgrain Boundaries 125
11.5. Foreign Dislocations . . . . . . . 129
11.6. "General" Subgrain Boundaries . . 130
11.7. The Energy of a Subgrain Boundary 131
11.8. Some Considerations of Recrystallization 133
 Contents XI

11.9. Limiting Cases . . . . . . . . . . . . . . 138

11.10. Dualistic Representation in Three Dimensions. 139
Chapter 11. Summary and Discussion . . . . 141
Chapter 12. General Geometrical Theory of Crystalline Inter-
faces . . . . . . . . . . 143
12.1. Introduction . . . . . . . 143
12.2. The Coincidence-Site Lattice 143
12.3. The O-Lattice. . . . . . . 148
12.4. The O-Lattice and Frank's Formula. 153
12.5. Theb-Lattice. . . . . . . . . . . 154
12.6. Various Aspects of the O-Points. . . 155
12.7. The Actual Formulation of the Transformation A . 158
12.8. Choice Between Various Possible Transformations A 160
12.9. Solutions of the Basic Equation. . . . . . . . . 169
12.10. The Behaviour of the O-Lattice upon Translation of
Lattice 2. . . . . . . . . . . . . . . . . . . 175
12.11. Translation of the O-Lattice for the Case of
Rank (T) < 3. . . . . . . . . . . . . 177
12.12. Subdivision of the Crystal Space into Cells 182
Chapter 12. Summary and Discussion . 183
Chapter 13. Applications of the O-Lattice Theory. . 186
13·1. Outline . . . . . . . . . . . . . . . 186
13·2. Example of an Optimum Phase Boundary 188
13·3· The Periodicity of the Pattern of Lattice Points. 196
13.4· Displacement of the Pattern . . . . . . . . . 206
13·5. Dislocation Networks in General Crystalline Inter-
faces . . . . . . . . . . . . . . . . . . . . 209
Chapter 14. Completion of the Linear O-Lattice Theory and
Extension to Non-Linear Problems . . . . . . 215
14.1. Outline . . . . . . . . . . . . . . . . . . . 215
14.2. Dislocations and Burgers Vectors in Crystalline
Interfaces . . . . . . . . . . . 215
14.3. Non-Linear Problems . . . . . . 221
Chapters 13-14. Summary and Discussion 226
Appendix • 229
Ai Matrix Calculation Procedures . . . . . 229
A2 Moire-Models . . . . . . . . . . . . . 243
A3 The Direct Observation of Crystal Defects 245
References . . . . . . . 247
Sources of the Illustrations 250
Subject Index . . . . . . 251
 1. Introduction
1.1. The Significance of Crystal Defects
On opening a book about crystal defects the reader may wonder why
this field should be investigated at all. It is well known that most of the
solid materials exist in crystalline form, whether this fact is directly
evident as in minerals or whether it becomes apparent only after pre-
paration (polishing, etching) as in metals. Crystals are distinguished by
their periodic, highly symmetric structures which possess a strong
aesthetic appeal. The word" defect", however, indicates a disturbance,
an imperfection, which, at first sight, appears to be neither aesthetic
nor attractive.
The final aim of research is to attain an understanding of the objective,
in this case of solid matter. Here it was seen quite early that the as-
sumption of a perfect crystal led to results differing greatly from the
measured data, especially with respect to flow stress, where the cal-
culated results were higher than the measured data by a factor 100.
Such observations led to the assumption that defects were already
present in crystals. An understanding of the" real crystal" (in contrast
to the" ideal crystal ") is unthinkable without the knowledge of crystal
defects. Crystal defects were investigated theoretically long before they
could be made directly visible. We do not enter the history of the research
on crystal defects as this is given in the books of Nabarro (1967) and
Hirth and Lothe (1968).

1.2. The Nature of Crystal Defects

In order to become acquainted with the nature of crystal defects we
first have to consider their environment, i.e. the crystal itself (Chap-
ters 2-3).
By neglecting lattice oscillations, the crystal defects can be divided
into the following groups, according to their dimensional nature:
a) Point defects: Vacancies, interstitial atoms, impurity atoms, and
combinations of these;
b) line defects: i.e. dislocations;
c) two-dimensional defects: stacking faults, grain boundaries, etc.;
d) three-dimensional defects: precipitates, inclusions, holes, etc.
I Bollmann, Crystal Defects
2 1. Introduction

The central element of our study is the line defect or the dislocation.
The theory of point defects is essentially independent of the theory of
line defects, and the methods of investigation of these defects also differ.
After a brief discussion of point defects (Section 4.1) the book deals
with the individual dislocation (5-6), the interactions between two dis-
locations (7-9), the interactions between dislocations and point defects
(10) and the fonnation of dislocation networks (11). The second part
of the book treats in detail the general geometrical theory of crystalline
interfaces (12-14) which may be either boundaries between crystals of
the same nature (grain boundaries) or boundaries between crystals of
different types (phase boundaries). The applications show that this
theory, which is based on purely geometrical concepts, already leads to
significant physical infonnation, and we believe that it will make a
sound foundation for a wide field of physical research into intercrystalline
boundaries.
The summaries try to provide a general picture of a group of chapters
and, in this manner, to discuss the most important points. Occasionally
it might be advantageous to read the summary before studying the
detailed text.
As the subject matter for this book has originated from the direct
electron microscope observation of crystal defects, emphasis is placed
more on the static than on the dynamic aspects. The problems of dis-
location dynamics, i.e. the theory of plastic deformation, is mentioned
only briefly (10.2).
 2. General Aspects of the Structure of Crystals

2.1. Outline
Solid state matter can be considered theoretically from different
view-points depending on the range of problems to be investigated. For
example, the elastic behaviour is generally discussed in terms of the
continuum aspect. Other properties can be explained on the basis of the
atomic arrangement. Our main interest concerns problems related to
the plastic behaviour of materials and, for dealing with those, we need a
certain knowledge of the atomic structure.
Concerning the atomic order two extreme states are distinguished,
the amorphous state, which is mainly disordered (e.g. glass) and the
essentially ordered crystalline state. In reality a certain local order
prevails even in the amorphous state. The characteristic feature of a
material in the crystalline state-i.e. of a "crystal" -is the periodicity
of its structure. A crystal consists of identical elements (building blocks)
which are joined together additively.
We may distinguish one-, two- and three-dimensional crystals.
Examples of one-dimensional crystals are chain molecules such as nylon
or silicate fibers. These molecules can, of course, assemble to form three-
dimensional arrangements; however, the bonds along the individual
fibers are much stronger than the inter-fiber bonds with the result that,
in solution nylon separates into single molecular chains (Fig. 2.1/1).
An example of a two-dimensional crystal is graphite (see Section3.8).
Within the layers the bonds are covalent and therefore very strong, while
the interlayer bonds are of the weak Van der Waals type. Other materials,
which cleave easily, such as mica, can also be understood as being made
up of stacked layers of two-dimensional crystals (Fig. 2.1/2). This concept
of the two-dimensional crystal is even useful for the understanding of
certain three-dimensional crystals (Section 3.2).
The word "crystal" usually means a three-dimensional crystal
already distinct by its external shape (e.g. the tungstane crystals of
Fig. 2.1/3). The external shape is-as is well known-a consequence of
the internal structure and most substances are crystalline without this
fact being evident from the exterior.
We have to distinguish clearly between the "model" of a crystal, its
abstract image, i.e. the ideal crystal, and the real substance, i.e. the
t*
4 2. General Aspects of the Structure of Crystals

Fig . 2.1 / 1. Disentangled nylon fibres as an example of a composite of one-

dim ensional crystals

F ig. 2.1/2. Graphite in a Ni-C eutectic as an example of a layer structure

composed of two-dimensional crystals. (Scanning electron micrograph;
M. Grages)
 2.2. Mathematical Description of the Crystal 5

real crystal, with all its imperfections and variations. The ideal crystal
represents only certain restricted aspects of the real material.
Before going into a discussion of certain specific crystal structures we
shall sketch some principles of crystallography. It might be thought that
the understanding of crystal defects would require a comprehensive
knowledge of the perfect crystal. However, in reality only a very limited
acquaintance with it is necessary.

Fig. 2.1/3. Tungsten produced by the gas phase decomposition of W5 2 , as an

example of three-dimensional crystals. (Scanning electron micrograph;
M. Grages)

2.2. Mathematical Description of the Crystal

As already mentioned in Section 2.1, the decisive feature of a (three-
dimensional) crystal is the periodic arrangement of the atoms (in three
dimensions). If one were to imagine part of the atomic pattern cut out,
then this cut out part could be moved by parallel displacement (i.e. by
translation) in such a manner that it would coincide at the new place
with the pattern already existing there. Suitable selection of the cut out
part could lead to a situation where the whole pattern could be built
up without gaps left by translation of the chosen part. In general, the
smallest possible part in the form of a parallelepiped with which this
6 2. General Aspects of the Structure of Crystals

construction can be carried out is called elementary cell or unit cell.

Sometimes a larger elementary cell is chosen in order to bring certain
symmetries to light (Fig. 2.2/1) .

Fig. 2.2/1. Choice of different elementary cells; pattern designed by

M. C. Escher

The crystal framework obtained by joining unit cells is called a

translation lattice *. The elementary cell of the three-dimensional transla-
tion lattice is given by three linearly independant translation vectors ~,
t2 and t3. Starting from a chosen origin every position x inside the
* In crystallography the translation lattice which is a mathematical
concept is simply called "lattice". However, as we deal here with crystal
defects we would not be allowed to use this word as soon as a defect were
present. Therefore we use the word "lattice" in a wider sense to designate a
more or less regular point arrangement and, if necessary, we give special
emphasis to the translation lattice.
 2.2. Mathematical Description of the Crystal 7

crystal is determined by:

(2.2-1)
The meaning of the upper and lower indices is explained in the Ap-
pendix A 1. Vectors and tensors are indicated by small- and capital
boldface letters respectively, and the coordinates by normal letters.
The translation vectors ti in their turn can be characterized by the
so-called lattice constants, i.e. the lengths of the three translation vectors
a (= It1 1). b (= It2 1) and c (= Ital) and the angles between them rx (= <)::,t 2 , ta),
f3 (= <)::, t a , t 1) and y ( = <)::, tv t2)'
For quantitative studies we have to place the crystal into a coordinate
system. Of all the possible coordinate systems two types are of particular
interest: The first type is the orthogonal system with three mutually
perpendicular unit vectors of equal length (e.g. 1 A). The second type
is the crystal coordinate system which is determined by three linearly
independent translation vectors (if possible the smallest translation
vectors of the lattice).
The orthogonal coordinates are more convenient for plotting the
results. With crystal coordinates a new coordinate system would have
to be traced for each individual case. The crystal coordinate system
corresponds better to the physical situation, as all the lattice points are
given by integer coordinates. In practice it is often necessary to change
from one coordinate system to the other.
In the orthogonal coordinate system the vector x shall be represented
by:
(2.2-2)
where, due to the orthogonality of the unit vectors:

(2.2-3)

The transition from the orthogonal to the crystal coordinate system

is formulated as follows: At first we express the basic vectors of the
crystal system ti by the orthogonal unit vectors u i (without decomposing
them into coordinates):
(2.2-4)
e
* We use the Greek letters'; and in order to emphasize that the same
vector is represented in different coordinate systems. The letters x and x'
stand for coordinates of different vectors in one single coordinate system.
a
According to Einstein's notation .;i ui means ~ !;i ui' The symbol : = means:
equal by definition. i=1
8 2. General Aspects of the Structure of Crystals

The orthogonal coordinates ~i are then expressed by the crystal co-

ordinates ~'i by means of the same tensor or matrix 8.

(2.2-5)

The following example may serve as an illustration: We assume that tl

coincides with 1lt, i.e. with the x-axis, that t2 is the xy-face diagonal of
the unit cube and t3 the cube diagonal (Fig. 2.2/2), i.e.

tl =u1 ·1 +u2 . 0 +u3 · 0

t2 = 1lt . 1 + u 2 . 1 + u 3 . 0 (2.2-6)
t3=1lt·1+U2·1+U3·1.

Fig. 2.2/2. Transition t o a new elementary cell [Eq. (2.2-6)]

In matrix notation [Eq. {2.2-6)J is written:

(2.2-7)

The (vertical) column vectors of the 8-matrix are therefore the new
basic vectors expressed in the coordinate system of the old ones. Thus,
the transformation of the coordinates of an arbitrary vector is given by:

~i= (~1)
~2 =
(1 1 1) ({~')1)
0 1 1 (f)2 =S1{~')k. (2.2-8)
~3 0 0 1 {$')3

For example the vector ;r with the crystal coordinates (f)i= [0,1, OJ
is given in orthogonal coordinates ~i = [1, 1, 0]. Accordingly, the in-
verse matrix 8-1 describes the transition from the orthogonal to the
 2.2. Mathematical Description of the Crystal 9

(1-11-10)
crystal coordinates.

8-1 = 0 (2.2-9)
o 0 1
(.;')i = (S-I)~ e (2.2-10)

(1 -1 0) (ee
i.e.

(.;'(.;')2
)1) 1
)
= 0 1 -1 2 • (2.2-11 )
(.;')3 0 0 1 e
3
All the transitions from one coordinate system to another or from one
vector system to another, described by matrices with constant terms as
above are called "linear transformations". We distinguish between
coordinate transformations and vector transformations. The formulation
of the transformation matrix 8 for a general crystal with given lattice
constants is indicated in the Appendix Ai Eq. (A 1-26).
If we need to know the length of a vector or the angle between two
vectors while those are expressed in an arbitrary coordinate system, this
information can be acquired in two ways: The first consists in trans-
forming the vectors into the orthogonal system where the required data
are obtained by means of the usual scalar products:
(2.2-12)
and
(xy)
cos (-9: x , y) = Ixllyl (2.2-13)
(where x and yare expressed in orthogonal coordinates).
The more direct method, however, makes use of the metric tensor G,
which depends on the coordinate system
G = 8 T 8. (2.2-14)
(8 T= transposed matrix of 8. See Appendix A 1.)
In the general case with x' and y' in the arbitrary coordinates, this
leads to
(2.2-15)
and
" (x'TGy')
cos ( -9: x y ) = ~1TYT . (2.2-16)
The metric tensor is always symmetrical. For a cubic crystal coordinate
system with the lattice constant a it becomes:

G= ( 0
a2 0 0)
a2 0 . (2.2-17)
o 0 a2
10 2. General Aspects of the Structure of Crystals

For the example (2.2-7), Gis

G~G ~ D (2.2-18)

Throughout this book we shall indicate vectors (translation vectors,

Burgers vectors etc.) by their dimensionless coordinates. Their dimensions
(length, angle, etc.) will be given by the metric tensor of the corresponding
coordinate system.
It is necessary, on occasion, to choose new basis vectors in a crystal
coordinate system such that the volume of the elementary cell remains
unchanged. This is done by a so-called "unimodular transformation" U.
The fact that the volume stays unchanged is expressed by

det(U) = ±1. (2.2-19)

The group of the linear transformations contains as a subgroup the

unimodular transformations and those contain the symmetry operations,
which leave the size and shape of the elementary cell unchanged (Sec-
tion 2.3). The transformation S of the example (2.2-7) is a unimodular
transformation, yet not a symmetry operation.
In the general theory of crystalline interfaces two adjacent crystals
of the same structure may need to be described by different elementary
cells. For this reason the relations between different possible elementary
cells are emphasized here.
If we choose in a crystal coordinate system an elementary cell as
unit cell *, then all the integer coordinates indicate discrete points of the
translation lattice. We name these points "lattice points" A lattice point
is the origin of the corresponding unit cell. Non-integer coordinates
indicate points within a unit cell (which mayor may not be atomic
positions).

Fig. 2.2/3. Two dimensional example of Wigner-Seitz cells

-----
'" In this connection we understand as elementary cell the smallest part
of the crystal pattern (of atomic positions) in the shape of a parallelepiped
through which the complete pattern of the crystal can be obtained by trans-
lation. The unit cell, however, is described by the units of the coordinates.
 2.3. Classification of Crystal Structures 11

Up to now we have assumed that the elementary cell is of the shape

of a parallelepiped. However, other shapes are conceivable. Of particular
interest is the' Wigner-Seitz cell' which is delimited by planes perpendic-
ularly bisecting the straight line segments between two nearest neigh-
bour lattice points. In this manner every lattice point is placed at the
centre of a polyhedron (Fig. 2.2/3). We shall have to deal with Wigner-
Seitz cells in connection with dislocation networks (Section 12.12).

2.3. Classification of Crystal Structures

We restrict ourselves in this Section to a few remarks. More detailed
descriptions are to be found in all standard textbooks on crystallo-
graphy.
All types of crystals are divided into seven crystal systems according
to the shape of the elementary cell, whereas this latter is chosen such as
to show the highest symmetry of the system.

Crystal system Lattice constants

Triclinic a =1= b =1= c, Q( =1= f1 =1= Y

Monoclinic a =1= b =1= c, Q( = f1 = 90°, Y =1= 90°
Orthorhombic (Rhombic) a =1= b =1= c, Q( = f1 = Y = 90°
Tetragonal a = b =1= c, Q( = f1 = Y = 90°
Cubic a = b = c, Q( = f1 = Y = 90°
Rhombohedral (Trigonal) a = b = c, Q( = f1 = Y =1= 90°
Hexagonal a=b=l=c, Q(=f1=900, y=120°

The cubic system is a limiting case of both the tetragonal and the
rhombohedral systems. The rhombohedral system itself can be inter-
preted as a special case of the hexagonal system.
All possible translation lattices are divided into 14 Bravais lattices.
The first group of seven is identical with the seven crystal systems while
the second group is derived from the first by adding face centres, base
centres and body centres.
The internal order is specified by symmetry operations. Those are
transformations which recreate the same point arrangement as the
initial one. Translations by translation vectors are specific symmetry
operations. A special class are the point symmetry operations which leave
at least one point (i.e. the origin) fixed and which are represented by
homogenious linear transformations (see Appendix A 1). Point symmetry
operations are rotations by 360 0 (i-fold rotation axis), 180 0 (2-fold),
120 0 (3-fold), 90 0 (4-fold) and 60 0 (6-fold), inversion at a point, reflection
at plane and all the possible combinations of these operations. In this
way 32 different point groups are distinguished.
12 2. General Aspects of the Structure of Crystals

230 space groups result from the combinations of the point symmetry
operations with translations. Space symmetry elements are screw axes
and glide reflection planes. Those operations are represented by non-
homogenious linear transformations (Appendix Ai).
In the next chapter we shall discuss in greater detail several special
crystal structures.
 3. Some Special Crystal Structures
3.1. Outline
In this chapter we shall discuss several common crystal structures,
sQme of which are of prime importance in metallurgy-such as the face-
centred cubic structure (austenitic steels, aluminium, nickel, copper,
silver, gold, etc.), the close packed hexagonal structure (zinc, cobalt and
others) and the body-centred cubic structure (ferritic steels, chromium,
molybdenum, tungsten, etc.). The diamond structure prevails in the
field of semiconductors, while (the hexagonal) graphite is important as
a moderator material in reactor technology and as a precipitate in
cast-iron.
Moreover, we shall consider different types of imperfections related
to the various structures.

3.2. Close-Packed Structures

The basic element of the so-called close packed structures is a planar
layer of equal spheres in the densest possible arrangement. Every sphere
represents an atom (Fig. 3.2/1), and as such is surrounded in the plane
by six nearest neighbours. This kind of a layer of spheres can be inter-
preted as a two-dimensional hexagonal crystal. Various stacking ar-
rangements of these layers lead to different three-dimensional structures.
We imagine a close packed layer to be given and the centres of the
spheres to be projected on the basal plane. This results in a triangular
(planar) point lattice. The set of these points defines the position of this
layer which we term as position A. If a second layer is placed on top of
the first one, two different stable positions are possible and these we call
Band C. The spheres of the second layer fit into the hollows formed
between the spheres of the first layer. The fact that the number of
hollows between the spheres is twice the number of spheres accounts
for the two different possible position Band C of the second layer. As
long as we assume that the atoms are held together mainly by isotropic
central forces, the position A is excluded for the second layer since the
spheres would then be in labile positions (on top of one another).
With only two layers given the positions Band C are physically fully
equivalent. They are transferred into one another by a rotation of 60°
around a vertical axis through an A-point. Therefore we may choose
14 3. Some Special Crystal Structures

arbitrarily for the second layer the position B or C, in other words, we

name the chosen position B or C.

o position A
+ {JositionB
~ position [
Fig. 3.2/1. A-, B- and C-positions in the close-packed arrangement of spheres

The position of the second layer shall now be called B. The third
layer can again be placed in two positions, namely C and A. The physical
situation differs, depending on the choice of that position. If we choose
the position C, a sequence ABC is obtained. Relative to an atom in
the B-position (in the second layer) all the direct neighbour atoms form
a centrally symmetrical arrangement (Fig. 3.2/2a). However, if we
chose position A for the third layer the sequence A BA is realized. In
this case the nearest neighbours of a B-atom are arranged III mIrror

c A

as bB

A A
Fig. 3.2/2a and b. Nearest neighbours of an atom in the B-position, a) in the
ABC-sequence, b) in the ABA-sequence
 3.2. Close-Packed Structures 15

symmetry (Fig.3.2/2b). For every subsequent layer a choice can be

made between two of the three positions: A, Band C. These three
positions are thus the only layer positions in close packed structures.
Of all the possible sequences of the positions A, Band C (whereby a
sequence of the type AA is excluded} two are particularly important,
namely ... ABCABC ... , the face-centred cubic structure and
... ABABAB ... , the close packed hexagonal structure. However,
sequences of the type AA have been found in non close packed layer
structures.
We introduce as an abbreviation (following Jagodzinski) the letter
"K" ("cubic") for a sequence of the type ABC (CBA, BAC, etc.). In
an analogous manner we denote by "H" (" hexagonal") a sequence of
the type ABA (BAB, CBC, etc.). In this way we can describe the face-
centred cubic structure by a series of K-sequences and the close packed
hexagonal structure by H-sequences. For completeness we denote by
"L" (labile) the (unstable) sequence of the type AAB (ABB, etc.).
The signs "\7" for the sequence AB (BC, CA) and" S' for AC (CB, BAJ
introduced by F. C. Frank, are often used in the literature. With this nota-
tion the K-sequence is expressed by \7\7 or /':,./':,. and the H-sequence by
/':,.\7 or \7 /':,.. We shall only use the K- and H-notation, which describes a
situation that is invariant with respect to orthogonal transformations, while
the /':,.-sequence is inverted when the layer stack is rotated by 1800 around
a horizontal axis.

- C
B
K A
/ / / /
B
K C V V /
K B V / /
A
K A
/ / / l(
K C
/ V V
K B
V V /
K A
/ V V
/
C V" r-....V /
V" V /
K
K B
" V V
/ "" V
b
A
"
ABCABCABCAB

Fig. 3.2/3a and b. The face-centred cubic (f.c.c.) structure, a) projections.

b) stacking sequence
16 3. Some Special Crystal Structures

From a vertical cut placed through the D-D-line in Fig. 3.2/1 result
diagrams which are convenient for representing the stacking sequence
such as Fig. 3.2/3 for the face-centred cubic and Fig. 3.2/4 for the close
packed hexagonal structure. The K, Hand L-sequences are indicated
according to Fig. 3.2/ 5.

:=

,
-- A A
H B
-- ~ -- 8 H A
H B
k=:::: ..::;:: - A
<=.:: ' l,I .:::::=- H A
H B
H A
H B
H A
H B

BCABCA B C A B
a b

Fig. 3.2/4a and b. The close-packed hexagonal (c .p.h.) structure, a) projections

b) stacking sequence

c
K A

H 8

c
c
A
Fig. 3.2/ 5. K-, H- and L-sequences
 3.3. The Face-Centred Cubic Structure 17

3.3. The Face-Centred Cubic Structure

The elementary cell of the face-centred cubic structure (abbreviation:
fcc-structure) is a cube with atoms at the comers and at the centres of
the cube faces. It becomes evident that the fcc-structure corresponds
to the ABC-stacking of close packed layers (Section 3.2) when the cube
is set on one of its comers such that a cube-diagonal is vertical
(Fig. 3.2/3).
Provided that we place the coordinate axes into the cube edges and
choose the length of the cube edge as unit length, then the unit cell
contains four atoms in the positions [OOOJ, [1/2, 1/2, OJ, [1/2,0, 1/2J and
[0,1/2, 1/2J * (Fig. 3.3/1). From the centrally symmetrical arrangement
of the neighbour atoms around a given atom it follows that the same

[1111 [111]

[ 0 111 1111-+-__

[1/1 01/2]-+--!---'.:-+-III/

[000] [liz liz 0]

Fig. 3.3/1. Face-centred cubic elementary cell with atomic positions ei==ui

structure can be built by ABC-stacking in the four <111) directions

([111 J, [111 J, [111 J, [n 1J **). The smallest translation vectors are of the
type (1/2) <110), i.e. of half a surface diagonal of the elementary cube.
The next larger translation vectors are of the <100)-type. They may be
composed of the smaller in the following way:
(1/2) [11 OJ +(1/2) [110J=[100J. (3·3-1 )
* Followinga general convention we denote by the kind of brackets:
special directions or vectors or positions by [... J, the type of direction by
<... >, special planes by (... ), (where the coordinates are the reciprocal
segments cut on the coordinate axes or the coordinates of the normal vector
in the reciprocal lattice), and the type of the planes by {... }.
** According to the general convention l' = - 1.
2 Bollmann, Crystal Defects
18 3. Some Special Crystal Structures

Vectors of the <11 0 )-type are even larger translation vectors; they
result from:
(1/2) [11 0] + (1/2) [11 0] = [11 0] (3·3-2)
or
[100] +[010] = [110]. (3·3-3)
The fcc-structure can also be understood as a rhombohedral structure
in which the (cubic) vectors [1/2, 0, 1/2], [1/2,1/2,0] and [0, 1/2, 1/2] are
taken as unit vectors of the rhombohedral system (Fig. 3.3/2). Expressed

III

111
Fig. 3.3/2. Rhombohedral elementary cell of the f.c.c. structure

by the cubic unit vector u. the rhombohedral unit vector u~ becomes

(Section 2.2):
(3·3-4)
i.e.
1/2 1/2 0 )
(
(tAt usUs) 0 1/2 1/2 = (u~ u; u~). (3·3-5)
1/2 o 1/2
The rhombohedral coordinates E" of a vector are then expressed in terms
of the cubic E' as follows:
(3·3-6)
i.e.

(fa~::) = (!
-1
!-!) (~:) .
1 1 E3
(3·3-7)

The volume of the rhombohedral unit cell is one quarter of the volume
of the cubic unit cell (det(S) = 1/4). If the value of the cubic lattice
constant is a, the metric tensor of the rhombohedral coordinate system
is given by:
1/2 1/4 1/4)
G=as ( 1/4 1/2 1/4 . (3·3-8)
1/4 1/4 1/2
 3.4. The NaCI-Structure 19

The rhombohedral unit cell contains only one atom, in contrast to

the four atoms of the fcc-cell. The cubic symmetry is not immediately
evident in the rhombohedral structure. However, four equivalent
rhombohedral elementary cells can be defined in the direction of the four
<111 )-axes. On the other hand, it becomes clear from the rhombo-
hedral elementary cell, that the fcc-structure can be understood as a
pure translation lattice.

3.4. The NaCl-Structure

As is well known, rocksalt is composed of Na+- and Cl--ions. If there
were no difference between these ions the structure would be simple

cr

a. cr

(1!z)(A+81

C
('Iz)(C+A I

(1J2)(8+CI

b C
Fig. 3.4/1 a and b. Elementary cell of the NaCI structure a) parallel to
individual ion layers, b) projection in the [111]-direction
2*
20 3. Some Special Crystal Structures

cubic. However, as they differ, we pay attention, at the moment only,

to one, e.g. the Cl--ions. For these ions alone the structure is face-
centred cubic. The Cl--ions are located in the complete crystal in the
standard positions: [000], [1/2, 1/2,0], [1/2,0,1/2] and [0, 1/2, 1/2]
(Section 3.3), while the positions oftheNa+-ions are: [1/2,0,0], [0,1/2,0],
[0,0, 1/2] and [1/2, 1/2, 1/2] (Fig. 3.4/1). The complete unit cell thus
contains 8 atoms.
The Cl--ions are arranged in a series of ABC-sequences (Section 3.2).
Between every two Cl--Iayers an Na+-Iayer is located in the positions
(1/2) (A +B), (1/2) (B+C) and (1/2) (C +A), where A, B and C re-
present positions of nearest-neighbour atoms in the adjacent layers.
The NaCl-structure can thus be understood as two interpenetrating
fcc-structures, translated by [1/2,0,0] with respect to one another.
The N aCI-structure can also be interpreted as a rhombohedral
structure with a unit cell containing a Cl--ion at [000] and a Na+-ion at
[1/2, 1/2, 1/2], i.e. at the centre of the unit cell (Fig. 3.4/2).

Fig. 3.4/2. Rhombohedral elementary cell of the NaCl structure

3.5. The Diamond Structure

The diamond structure may be understood as composed of a basic
face-centred cubic framework with atoms in the standard positions
[000], [1/2, 1/2,0], [1/2,0, 1/2] and [0,1/2,1/2] and a further group of
atoms in the positions [1/4, 1/4, 1/4], [3/4,1/4, 1/4], [3/4, 1/4, 3/4] and
[3/4,3/4,3/4] (Fig. 3.5/1). The atoms in the two groups may be chemi-
cally identical as in diamond (carbon), germanium and silicon or chemi-
cally different as in the III-V-compositions such as ZnS, GaAs, etc.
As in the NaCI structure the diamond unit cell contains also 8 atoms
arranged in two interpenetrating fcc-lattices except that here the trans-
lation between the two lattices is [1/4, 1/4, 1/4].
The diamond structure can also be described as rhombohedral with
a unit cell containing two atoms in the positions [000] and [1/4, 1/4, 1/4]
 3.5. The Diamond Structure 21

([3/4, 1/4, 1/4J respectively, depending on the choice of the unit vectors).
The translation vectors of the diamond structure are therefore those of
the fcc-lattice.

r---r-t::=t:~L_-I-- [1/4 .l/., l/41

-r''t--- -I-- P/•. lf4 .3/41
[ 0.lf2.'hl - -+-+
[lh,O.1/21 - - + - -----+
[lA .1/4 .v.l --f--~

[0.0.01
[lh .l/z.O)
Fig. 3.5/1. Elementary cell of the diamond structure

Instead of a close packed structure of spheres, the diamond structure

can be interpreted as a loose stacking of tetrahedra with atoms at the
corners and at the centre of the tetrahedron (Fig.3.5/2). Considering
layers of single atoms, their stacking sequence can be written:
... AA'BB'CC'AA' ... . Here AA'BB'CC' becomes a K-sequence and
AA'BB'AA' an H-sequence.

A
Fig. 3.5/2. The diamond structure represented as the stacking of tetrahedra
22 3. Some Special Crystal Structures

3.6. The Body-Centred Cubic Structure

The body-centred cubic structure (abbreviation: bcc-structure) is
certainly one of the most important in metallurgy, since iron is crystallized
in this structure at room temperature, as are other metals such as the
alkali metals, V, Nb, Ta, Cr, Mo, W, etc.
The body-centred cubic unit cell contains two atoms in the positions
[000] and [1/2, 1/2, 1/2] (Fig.3.6/1). This structure is a translation

[1/2,1/2,1/21---jf---!---I;.

[0.0.01

Fig. 3.6/1. The body-centred cubic elementary cell

lattice (Section 2.2) for which the following translation vectors, expressed
in cubic coordinates, can be chosen (Fig. 3.6/2):
~= [1/2 1/2 1/2]
t2 = [-1/2 1/2 1/2] (3·6-1)
ia=[O 0 1].

Fig. 3.6/2. Monoatomic unit cell of the b.c.c. structure [Eq. (3.6-1)]
 3.6. The Body-Centred Cubic Structure 23

From this follows the matrix 8 [compare Eq. (2.2-4)]:

1/2 -1/2
8 = ( 1/2 1/2
1/2 1/2 ~) (3·6-2)

which transforms the cubic into the oblique angled unit vectors. The
matrix 8-1

8-1 = (-~o ~ ~)
-1 1
(3·6-3)

transfomls the cubic coordinates of a given vector into its oblique angled
ones (2.2-10). The metric tensor of this new coordinate system is:

3/4 1/4 1/2)

G=a 2 ( 1/4 3/4 1/2 . (3.6-4)
1/2 1/2 1

The new unit cell contains one atom and therefore its volume is
half that of the cubic unit cell (det (8) = 1/2). Its faces are {11 O} planes.
The shortest translation vectors of the bcc-structure are of the (1 /2) <111)
type and the next longer ones of the <100)-type,

(1/2) [111]+(1/2} [111]=[001] (3·6-5)

and the next of the <110 )-type

(1/2) [111] + (1/2) [111] = [110]. (3·6-6)

Another primitive elementary cell can be defined by the following
translation vectors (Fig. 3.6/3):

t~ = [1/2 -1/2 1/2]

t~= [1/2 1/2 1/2] (3·6-7)
t~= [-1/2 1/2 1/2]
with 8:
1/2 1/2 -1/2)
8= ( -1/2 1/2 1/2 (3·6-8)
1/2 1/2 1/2
and 8-1 :

8-1 = ( ~ -~ ~) . (3·6-9)
-1 0 1
24 3. Some Special Crystal Structures

Here two faces of the unit cell are {iii }-planes and one is a {11 0}-plane.
The metric tensor is:
3/4 1/4 -1/4)
G=a 2 ( 1/4 3/4 1/4. (3·6-10)
-1/4 1/4 3/4

o
--- x

Fig. 3.6/3. Monoatomic unit cell of the b.c.c. structure [Eq. (3.6-7)J

The bee-lattice can also be described by a face-centred tetragonal

elementary cell (Fig. 3.6/4) by the new unit vectors:
t~ = [1 1 0]
t; = [-1 1 0] (3·6-11)
t~= [0 0 1] .

Fig. 3.6/4. Face-centred tetragonal elementary cell of the b.c.c. structure

[Eq. (3.6-11) J
 3.7. The Close-Packed Hexagonal Structure 25

With the matrix S:

1 -1
S= ( 1 1 (3·6-12)
o 0
and S-l:
1/2 1/2 0)
S-l= ( -1/2 1/2 0 (3.6-13)
o 0 1
and G:

(2 0 0)
G=a 2 0 2 0 . (3·6-14)
001

The tetragonal unit cell is obviously twice as large as the bcc cell, since
it contains four atoms.
The above description demonstrates the relationship between the
body-centred and the face-centred cubic lattices. One lattice can, in
principle, originate from the other by pure deformation. This type of
phase change is called the Bain transformation after the American
metallurgist Edgar C. Bain.

3.7. The Close-Packed Hexagonal Structure

The close-packed hexagonal structure (cph-structure) is formed by
an ... ABAB ... -stacking sequence (i.e. by a series of H-sequences) of
close-packed layers of spheres (as already indicated in Section 3.2). Here
the atoms are arranged mirror-symmetrically with respect to any close-
packed layer-plane. The elementary cell can be defined by the three
basic vectors a, band c (Fig. 3.7/1), where a and b lie in the basal plane,
are of equal length (smallest distance between spheres) and enclose an
angle of 120°, and c is perpendicular to a and b with a length equal to
twice the interlayer spacing. The elementary cell thus defined contains
two atoms, at the positions [000] and [1/3,2/3, 1/2]. In the ideal cph-
structure the ratio cIa is given by: cIa = (2/3)VO = 1.630. This becomes
obvious when the positions of atoms are given in cubic coordinates, i.e.
1(2/3) [111]1/(1/2) [11 0]1. In real crystals the values can deviate con-
siderably from the idealratio.
The close-packed hexagonal structure is not a translation lattice,
i.e. not all the atomic positions can be obtained by composing basic
vectors. If we assume the vector [1/3,2/3, 1/2] pointing from the origin
to the atom inside the elementary cell to be a translation vector, then
another atom should be found at the position [-1/3, -2/3, -1/2],
26 3. Some Special Crystal Structures

which is not the case. Therefore the translation lattice of the cph-
structure is determined by a, b and c.
The metric tensor of the hexagonal structure is

(3.7-1)

t--t-t--t--t-[ll.l. 213.1J21

Fig. 3.7/1 a and b. The close-packed hexagonal structure; close-packed layers

from a) side, and b) top

In the ideal cph-structure, c2 is given by:

c2 = (8/3) a2 • (3·7-2)
In order to describe the hexagonal structure, a 4-coordinate-representa-
tions is often applied, in which the four coordinates refer to four unit vectors,
three of which, lying in the basal plane, are rotated by 1200 with respect to
 3.8. The Graphite Structure 27

each other around the c-axis, and therefore are linearly dependant. This
involves the additional condition that the sum of the first three coordinates
must be zero. The unit vector [100] in this representation becomes:

tl = (1/3) [iHO]. (3.7-3)

All the possible vectors of the same kind are obtained by permutation and
change of sign of the first three coordinates; hence, the type of the vector is
immediately evident.
We shall not use this representation, because it is removed from the
general method. We shall use the methods of linear algebra which can be
applied to any crystal system, and are convenient, in particular, in computer
calculations. Furthermore, it might be necessary to choose for one given
structure elementary cells of other systems (e.g. a rhombohedral instead of
a hexagonal one), in which case a notation, exclusively valid for the hexa-
gonal system, proves to be more of an obstacle than an advantage.

3.8. The Graphite Structure

The graphite structure, i.e. the structure of the carbon phase stable
under normal conditions, is a very pronounced layer structure. The
individual layers are formed from condensed benzene rings with the
bonds between the atoms being homeopolar (Fig. 3.8/1). Between the
layers the bonds are of the relatively weak Van der Waals type which
accounts for the easy cleavage of these crystals.

_____-1---jf-::::---t--.----~~-7 1/2

- - {11J.Z13,1121
~~---[0.0.1f21
->-+-~

[0.0.0] [21.!.1/3.01
Fig. 3.8/1. The graphite structure
28 3. Some Special Crystal Structures

The natural graphite crystals are hexagonal in the-ABAB-stack-

ing. The elementary cell contains four atoms in the positions: [000],
[2/3,1/3,0], [0,0,1/2] and [1/3,2/3, 1/2J. The metric tensor of the graphite
structure is the same as that given in (3.7-1). The lattice constants are
a=b=2.46A, c=6.69 A.

3.9. Irregularities in the Stacking Sequence

In Sections 3.3 to 3.8, a number of structures which can be built up
from the uniform stacking of atomic layers were discussed. In this
section we shall investigate several kinds of " stacking faults" (i.e. planar
faults) which originate from irregularities in the stacking sequence.
First we shall consider the face-centred cubic structure resulting from
a series of K-sequences. If the crystal contains one single H-sequence in
the interior, this means-as shown in Fig. 3.9/1-that it is twinned with

A
KB \ r\ \ r\
KC 1\ \ \
K A \ 1\ \
KB \ \ \ 1\
H C 1\ \ \
K B V / /
KA / / / V
K C V V /
KB / / V
A
l;' / / V
ABCABCABCAB

Fig. 3.9/1. Growth stacking fault (twin boundary)

the middle layer of the H-sequence being the twin boundary. This kind
of stacking fault is generally called a growth stacking fault, because it
occurs principally during crystal growth or recrystallisation. A series of
two H-sequences is a so-called deformation stacking fault (Fig. 3.9/2). On
speaking of stacking faults, deformation stacking faults are generally
meant. The configuration of this type of fault can be produced by plastic
deformation, i.e. by shifting one part of the crystal with respect to the
other.
A translation vector for shifting a layer into an equivalent neigh-
bouring position is of the (1/2)(110)-type, and for shifting it into a
stacking fault position it is of the (1/6) (112)-type. Thus, a lattice
 3.9. Irregularities in the Stacking Sequence 29

translation in the (111)-plane can be split into two steps, e.g.

(1/2) [110] = (1/6) [211] + (1/6) [121]

C
K 8 V / /
KA / / / /
KC I V I
K 8
I / V
H A
/ / / V
H8 1\ \ 1\ 1\
K A
I / V /
KC V / V
K 8 / V /
A
/ / V /
A8CA8CA8CA8
Fig. 3.9/2. Deformation stacking fault

where the first step leads from the perfect lattice position into the
stacking fault position and the second step from this stacking fault
position into the new perfect lattice position. The sequence of these
steps cannot be chosen arbitrarily (see Section 8.2).
An important source of stacking faults is the condensation of vacancies
or interstitial atoms. The condensation of vacancies corresponds to
removing one layer followed by closure of the gap. We consider the
sequence
t
... ABCA(B)CABC .. .
... KKKKKKK ... .
If the B-Iayer is removed we obtain
... ABCA/CABC .. .
... KKHHKK .. .
i.e. the configuration of the deformation stacking fault (Fig. 3.9/2).
Is, however, a layer added at the indicated location
t
... ABCABCABC .. .
... KKKKKKK .. .
then it can occupy neither the A- nor the B-position because in these
cases the unstable AA- or BB-sequence would result. Thus, the ad-
30 3. Some Special Crystal Structures

ditionallayer can only be placed in the C-position (Fig. 3.9/3) .

... ABCACBCABC .. .
... KKHKHKKK .. .
According to F. C. Frank two kinds of stacking faults, called "intrinsic"
and" extrinsic" faults are distinguished. At an intrinsic fault the crystal
is undisturbed up to the boundary surface on approaching the disturbed
area from either side [e.g. twin boundary or condensed vacancies (Figs.
3.9/1 and 2)J. An extrinsic fault, however, contains an intermediate
layer which differs from the crystal lattice on both sides of the fault
(condensed interstitials, Fig. 3.9/3).

C
K B / / / I
K A V / V ~
K C / V V
K B / I I
11 A I I I /
K B \ \ \ \
11 C \ \ \
K B I I I
K A / I I V
C V V /
ABCABCABCAB

Fig. 3.9/3. Stacking sequence of condensed interstitial atoms

Stacking faults occur also in the hexagonal sequence. Condensed

vacancies, e.g.
t
... ABAB(A)BABAB ...
... HHH H HHH ...
produce, (Fig. 3.9/4a):
... ABAB/BABAB .. .
... HHLLHHH .. .
As mentioned above, this sequence is generally unstable, in particular
if the atoms of the crystal are bonded by central forces. It can be
stabilized by shifting one part of the crystal to one side or the other.
For example:
... ABA BCBCBCB .. .
... HHKHHHHH .. .
 3.9. Irregularities in the Stacking Sequence 31

Here the stacking fault is intrinsic (Fig.3.9/4b). For condensed inter-

stitials the situation is:
t
... ABABABAB .. .
... HHHHHH .. .

,..
A
\
r..
A
r\ r\ 1\ r\ N \J \J
(V V vr
H B HB
HA I I HA V I I
H B ~ ~ 'J \ HB ~ N \ r\
HA V I I V HA V I / V
1. B
r\ \ \ ~ H B
1\ \ r\ 1\
1. B I I I II
~ If. A
f--

HA
I I V K C
I I V
H B
H A
~
(/
r\
JV
r\
V
r\
V
HA
HC
j - r\
V
r\
V
B
\ ~ ~
~ ~ A r\ r\ r\ )-
ABCABCABCAB ABCABCABCAB
b

Fig. 3.9/4a and b. Stacking sequence of condensed vacancies in the c.p.h.

structure, a) unstable sequence, b) stabilized sequence

Here a C-Iayer is, inserted (Fig. 3.9/5 a. Is the drawing right?):

... ABABCABAB .. .
... HHKKKHH .. .
Thus, three consecutive K-sequences appear, which considerably disturb
the hexagonal order. This arrangement can also be changed to a less
disturbed by shifting one part of the crystal (Fig. 3.9/5 b):
... ABABCABABAB .. .
... ABABABOBOBO .. .
... HHHHKHHHH .. .

Now, only one K-sequence remains.

Considerations of this kind can be applied to the stacking sequences
of other "layer structures" as well.
Also in the NaCI-structure certain kinds of stacking faults are
possible on the {111}-planes, due to the opposite polarity (and hence
attraction) of alternate layers. However, a shift by a (1/2) 1 OO)-type <
32 3. Some Special Crystal Structures

A A
H D
r\ ~ ~ \l H D
H A
V V V V HA
H D
~ ~ ~ 1\ !. D
i. A
[7 lJ V V H C
If. C
V V V H B
If. B
) J J H C
H C - ~ )- I- ~ \ H B
J J )
H D ~I-
If C
H C
~ \ \ H B
B
J J~r'--
V C
A DCA DCA DCA D ABCABCABCAB
a. b

Fig. 3.9/Sa and b. Stacking sequence of condensed interstitial atoms in the

c.p .h. structure, a) high energy sequence, b) sequence with lower energy

A 8
c' A'

C
8'

8~~~~~~~~~~=3~~~==~ 8
A' A'

A A

Fig. 3.9/6. Stacking faults in the diamond structure. (Identical positions in

the front- and top views are marked by the same striated pattern)
 3.10. Energy of Stacking Faults 33

vector in a {1 OO}-plane brings atoms of the same polarity (repelling each

other) together; this accounts for the {1 00}-cleavage facets of rock salt.
As the diamond structure can be considered as being built by the
stacking of layers of tetrahedra, stacking faults analogous to those
observed in the fcc-structure can also occur here. An H-sequence in the
diamond structure can also be understood as being produced by a. 60°
rotation of the middle layer of a K-sequence around a vertical axis
(Fig. 3.9/6).
In the graphite structure all the possible stacking faults of the
hexagonal structure can occur, due to the weak interlayer bonds. Even
the unstable AA-stacking can be found, e.g. in small areas of condensed
vacancies (diameter up to several hundred A). In this case the strong
bonds within the layers prevent a lateral shift of the type illustrated
in Fig. 3.9/4a.
A special kind of faulted layer structure is the turbostratic arrange-
ment. In this case layers or layer packets, rotated by arbitrary angles
with respect to one another, are stacked. Artificial graphite, used, for
instance, as electrodes or as nuclear reactor moderators, is almost always
found in this arrangement.

3.10. Energy of Stacking Faults

The fact that both face-centred cubic (e.g. AI, Au, Ag, Ni, austenitic
steel, etc.) and close-packed hexagonal crystals (e.g. Co, Be, Mg, Cd)
are found in nature shows that there must be a difference between the
energies of the H- and the K-sequences. Otherwise both sequences
would occur in a statistical distribution.
We now envisage an A-layer with a B-layer on top of it. For the
third layer two possibilities exist: the A- or the C-position. Now we add
the third layer. If we assume that the material crystallizes in the fcc-
structure, this third layer must be attracted more strongly into the C-
than into the A-position. The free energy of the crystal in the ABC =
K-sequence is here necessarily lower than in the ABA =H-sequence.
This energy has to be increased on shifting the third layer from the C-
into the A-position (Fig. 3.10/1). The difference in free energy is obviously
connected with the fact that, at the transition from aK-to anH-sequence,
the arrangement of atoms changes from central symmetry to mirror
symmetry. Although the bonding of close-packed structures is mainly
through central forces, a small amount of bonding is directional. The
energy difference LIE between the K- and the H-sequence is a measure
of the so-called "stacking fault energy" 'Y, with its value normally
given in erg/cms. More exactly, 'Y in the fcc-structure indicates the
3 Bollmann Crystal Defects
34 3. Some Special Crystal Structures

energy per unit area of a deformation stacking fault, i.e. of a series of

two H -sequences.
To a first approximation, the energies of different kinds of stacking
faults can be estimated as being proportional to the number of foreign
sequences (e.g. H-sequences in the fcc-structure). Thus, the energy of a
twin boundary is approximately half that of a deformation stacking

ABCABCABCABC
Fig. 3.10/1. Difference in the potential energy between the C- and the
A -position (schematic representation)

fault. The energy of a layer of interstitial atoms in the hexagonal

structure, as shown in Fig. 3.9/5, can thus be considerably lowered by
shifting the layer.
These considerations show the advantage of the K- and H-notation.
The values of stacking fault energies in the fcc-structure vary between
practically zero (stainless steel) and f'::j200 erg/cm" (AI). Numerical data
are given in Hirth and Lothe (1968), p. 764 and Weertman (1964), p. 98.

3.11. Fibrous Structures

The case of graphite (Section 3.8) shows that the chemical bond can
be different in different directions, thus producing strongly anisotropic
materials. In layer structures the bond is stronger within the layers than
between them. It is also possible that the bond in one direction is stronger
than in all the others, thus creating primarily chain molecules with more
or less loose side-bonds. Such arrangements are found in silicates and
polymers (Section 2.1).
Several equivalent side bonds per unit length may be present,
such that the chains can be linked to one another in different (relative)
 Chapters 2-3. Summary and Discussion 35

positions. An example for this is Tobermorite (Fig. 3.11/1), which plays

an important part in the setting of cement. Other possibilities arise from
the folding of chain molecules (polyethylene).

Fig. 3.11/1. Tobermorite as an example of a fibrous structure. (This drawing

is based on the front cover of Scientific American, April 1964)

Chapters 2-3. Summary and Discussion

In describing the crystal we discussed principally its periodicity, i.e.
the configurations produced by pure translations. Irregularities in the
periodicity were also analysed. The point groups and space groups
representing the symmetries of crystals, which are of prime importance
in crystallography, were only briefly considered. These latter groups
playa secondary part in the understanding of crystal defects, and, in
particular, of crystalline interfaces; here elementary cells have often to
be chosen, which do not show the highest symmetry of the crystal
structure.
3'
36 3. Some Special Crystal Structures

Some important structures were discussed, for example:

3.3. the face-centred cubic (fcc) structure,
3.4. the rock salt (NaCl) structure,
3.5. the diamond structure,
3.6. the body-centred cubic (bcc) structure,
3.7. the close-packed hexagonal (cph) structure,
3.8. the graphite structure, and a few indications were given on,
3.9. fibrous structures.
These considerations brought out the close relationship that exists
between Nos. 3.3, 3.4 and 3.5 on the one hand and 3.7 and 3.8 on the
other. Both groups of structures are closely related and are therefore
subject to the same types of possible stacking faults. There exist also
relations and transitions between Nos. 3.3 and 3.6. Phase transforma-
tions occur in nature between Nos. 3.3 and 3.7 (cobalt) as well as between
Nos. 3.3 and 3.6 (y-iron - at-iron). An additional type of disorder is
found in graphite: the turbostratic stacking. Other examples of ir-
regularities appear in fibrous structures.
Suggestions for Problems
As this chapter was aimed at the understanding of the translation
properties of crystals, the transition between different elementary cells
can be practiced. It is recommended that the metric tensor be employed.
For this purpose the absolute length of vectors and the angle between
vectors given by their coordinates in a crystal coordinate system may
be calculated. The methods are summarised in Appendix A 1. A further
series of possible problems may concern the analysis of stacking faults.
 4. Point Defects

4.1. General Considerations of the Field of Point Defects

Since the field of point defects is rather different from the field of
dislocations we will only give a few short indications about the former
here. Details can be found in the books listed in the references.
Point defects are localised in a crystal in such a way that one or
several elementary cells, containing the point defect, can be virtually
replaced by perfect elementary cells, resulting in an undisturbed crystal.
The elementary point defects are vacancies and interstitial atoms,
and also impurity atoms. From these elementary defects, groups can
be formed, such as double vacancies, a vacancy combined with an
impurity, etc.
Point defects, in general, are mobile configurations having a certain
energy associated with them. Apart from local changes of atomic positions
the atoms can remain in place when the configuration moves. This
characteristic is most evident in the case of a vacancy. When a vacancy
moves, it simply exchanges places with a neighbouring atom (Fig.4.1/1).

1
- ti8.
O.
3
fill 2. 8fj3
4

.2.
2.
3

1
4

1
4

·2
3
•

Fig. 4.1/1. Movement of a vacancy

For interstitial atoms, two possibilities exist: the single interstitial atom
can either traverse the crystal on its own, or it can be repeatedly replaced
on its way by other atoms, in analogy with a relay. The first case is
possible, for example, in strongly anisotropic crystals such as graphite,
where an atom is highly mobile between the layers but strongly bonded
within them. In the face-centred cubic structure interstitial atoms, and
especially impurity atoms, can be located in octahedral-(1/2 [111]) -or
tetrahedral-(1/4 [111]) holes, eventually leading to distortion of the
surrounding lattice. On the other hand, there are arrangements such
38 4. Point Defects

as dumbbells (Fig. 4.1/2), in which the interstitial atoms cannot be

located separately, and which must obviously move along the dumbbell
axis as a configuration and not as individual atoms (Fig. 4.1/3). Linear
arrangements are also possible with interstitial atoms "spread" over
the length of several elementary cells. These are called "crowdions".
Impurity atoms in low concentration normally form individual point
defects.

c
Fig. 4.1/2a--c. Dumbbell configuration of interstitial atoms a) and b) in the
f.c.c. structure, c) in the b.c.c. structure

Point defects in homogeneous material can be produced spontaneously

under conditions of thermal equilibrium, or by irradiation or by mechan-
ical deformation. They can be frozen in on quenching from a state of
thermal equilibrium; in this manner the material becomes supersaturated.
If the material is heated so that these defects just become mobile, they
will precipitate, i.e. form clusters (Fig. 10.1/1). In heterogeneous systems,
point defects can be produced by alloying (sometimes involuntarily) or
by the diffusion of gases into the material.
The equilibrium concentration of single point defects as a function
of temperature is given by
EB
n ----;<T
c= N =e (4.1-1)
 4.1. General Considerations of the Field of Point Defects 39

where
c= concentration of point defects,
n= number of point defects per N atoms,
EB = formation energy = work necessary to form the defect,
k= Boltzmann constant = 1.37.10-23 Watt· sec. degree-I,
T= absolute temperature.

Fig. 4.1/3. Movement of a dumbbell configuration

In the Arrhenius plot, In c=f(1/T), Eq. (4.1-1) represents a straight

line, the inclination of which is determined by EB •
The theory of the formation energy for point defects is important
as it helps to identify the defect from an analysis of the thermal be-
haviour of the material (or the electrical resistance). A few theoretical
formation energies for vacancies are:
Cu~0.8-1 eV,
Ag~0.6-0.geV,

Au ~ 0.6-0.8 eV .
For the alkali metals, the energies vary from 0.55 eV for Li to 0.26 eV
for Cs.
The formation energies for interstitial atoms are much higher:
3-6 eV. This means that in a state of thermal equilibrium only vacancies
40 4. Point Defects

are present. Interstitial atoms, therefore, have to be produced by other

means, e.g. by irradiation.
The formation energies of multiple point defects are, in general,
smaller than the sum of the single energies; this accounts for the cluster-
ing of these defects.
If a point defect moves from one stable position to the next, it has
to overcome an energy barrier. The height of this barrier thus represents
the activation energy for the motion. While vacancies are formed more
easily than interstitial atoms, they are less mobile. On the other hand,
a double vacancy is more mobile than a single vacancy. During heat
treatment of a distribution of vacancies and interstitial atoms produced
by irradiation, the interstitial atoms will be the first to cluster, followed
by the vacancies at a higher temperature.
Here we close our brief survey of point defects. Section 10.1 deals
with the relations between point defects and dislocations.
 5. The Individual Dislocation: Geometrical Basis

5.1. Outline
Dislocations are line defects in an otherwise perfect crystal. In this
chapter the basic geometrical concepts shall be explained, such as the
path of the line, the Burgers vector (strength of the dislocation), the
relative orientation of these two, possible movements of dislocations and
their relations to Moire patterns. These latter are a convenient means for
understanding many of the basic dislocation phenomena.

5.2. Definition of the Dislocation Line

and the Burgers Vector
As a thought experiment one can imagine a dislocation being produced
by making a cut in a perfect crystal. Next, the two sides are shifted with
respect to each other by a translation vector of the lattice. Afterwards
either missing material is added or surplus material is removed. The
shifted surface of the cut may grow together in such a way that the cut
is no longer discernible. Inside the crystal, at the edge of the former cut
surface, remains a line of strongly disturbed material. This is the" dis-
location line" (Figs. 5.2/1 and 5.2/2). The dislocation line is surrounded
by perfect, though elastically deformed material. The cut is called a
"Volterra-cut". It does not matter where this cut is made in the
crystal, as long as it is limited by the given dislocation line, and as long
as the displacement of the two surfaces of the cut remains the same.
Dislocations of this kind can only occur in a periodic substrate. In
nature crystals are periodic substrates. In a continuum one could
imagine a wedge-shaped Volterra cut with the displacement being
variable along the line.
Dislocation lines are characterised by two components:
1. the path of the line in the crystal and
2. the so-called "Burgers vector", a pseudo-vector, which is a
measure of the strength of the dislocation.
The material in the immediate surroundings of the dislocation line is
strongly disturbed. This region is called the "dislocation core" and the
material is denoted "bad material". Further away from the line the
42 5. The Individual Dislocation: Geometrical Basis

original crystalline arrangement persists, though it is elastically strained.

The elastically deformed material outside the dislocation core is denoted
"good material".
In order to define the Burgers vector, the ilisturbed crystal is com-
pared with a perfect reference crystal. An arbitrary line sense, i.e. a

/'1/J?' /'
/~/

/ 1/
,.....-
I

-;7
-r-
/'
/
a b

/ .#
.I
-t.~/

I
/
:!
I.!.

V
/ y Fig. 5.2/1. Generation of an edge
dislocation by a Volterra cut
I
c

/ /
J
L /
/1
1
1
1
1
1
1
L
1
//
~ V
b
a
Fig. 5.2/2. Generation of a screw dislocation by a Volterra cut
 5.2. Definition of the Dislocation Line and the Burgers Vector 43

direction, is attributed to the dislocation line. Now, a closed circuit in

the sense of a right-handed screw is made around the line; it has to be
situated entirely in "good" material. This closed circuit is then imaged
step by step onto the reference crystal where it remains open. The vector
pointing from the end to the beginning of the circuit in the reference
crystal is the Burgers vector (Fig. 5.2/3 a).

c R

S S ,ll
F G9 F

\ \
I I
a

1l. ' S S
F @ F

b
Fig. 5.2/3a and b. The Burgers loops as an imaging relation between the
distorted crystal C and the perfect reference lattice R. a) The loop is closed
in the distorted crystal, FS/RH-convention. b) The loop is closed in the
reference crystal

An equivalent procedure consists in imaging a circuit, closed in the

reference crystal, onto the disturbed crystal, where it is now open. In
order to maintain the sign the Burgers vector has here to point from the
beginning to the end of the circuit (Fig. 5.2/3 b). Often the second method
is more convenient for the visual representation of a dislocation. One
has to follow a path in the disturbed crystal which would be closed in the
perfect crystal (as many steps to the right as to the left, as many up as
well as down).

Remark
In reality, we are dealing here with three different coordinate systems, i.e.
1. the Euclidean coordinate system of the surroundings,
2. the distorted system of the crystal, determined by the real positions of
the atoms, and
3. the system of the perfect reference crystal.
44 5. The Individual Dislocation: Geometrical Basis

The equivalence of the coordinate steps of systems 2 and 3 is defined by

the method of imaging the circuits from one crystal onto the other. The
Burgers vector can therefore be considered equally well as a coordinate dif-
ference in either system 2 or 3. Thus, even though the vectors connecting
neighbouring atomic positions change locally in magnitude and direction
with respect to system 1, due to crystal distortion, they remain constant
coordinate differences, independant of location, in system 2 or 3. System 3
merely provides visual evidence of this fact. Hence, the Burgers vector has
to be understood as a coordinate difference, whether the coordinate system
is distorted or not.
The main points of the definition of the Burgers vector may be
summarised as follows:
1. The line sense has to be arbitrarily chosen.
2. A circuit in the sense of a right-handed screw has to be made
around the dislocation in such a way that
a) it is closed in the disturbed crystal, or
b) it is closed in the reference crystal.
3. a) Then, in the reference lattice, the Burgers vector points from
the end to the beginning, and
b) in the disturbed crystal, from the beginning to the end of the
circuit.
4. The Burgers vector is given in crystal coordinates. (The unit
vectors are the edges of the elementary cell.)
This definition of the sign of the Burgers vector is given in the
literature as the FS/RH-sign (Finish-Start/Right Hand) in the reference
lattice (Bilby et al., 1955). The opposite sign is also used (connection
from the end to the beginning of the circuit in the disturbed crystal)
(Read, 1953). It is important for the study of the geometry of dislocation
arrangements to be aware of the sign definition applied in each particular
case. Throughout this book, the FS/RH sign definition shall be used.
An enumeration of the essential properties of the Burgers vector,
b follows:
a) b is independant of the path of the circuit around the dislocation
line and of its starting point, as long as the dislocation is enclosed by the
circuit made in the sense of a right hand screw.
b) b is independant of the position of the circuit with respect to the
dislocation as long as the circuit is moved along the same dislocation, i.e.
b is invariant along the dislocation.
A dislocation cannot start or end in the interior of a crystal, as this
would mean that b would change from a finite vector to zero on moving
the Burgers circuit along the dislocation. The dislocation either starts
or ends at a surface, or forms a closed loop.
 5.3. Relative Orientation of band 1 45

c) The sign of b has significance only when considered together with

the line sense. I is a vector in the dislocation line indicating the chosen
line sense and the local line orientation. If I and b change their signs
simultaneously, they still represent the same unchanged geometrical
situation in the crystal. This is due to the fact that, from the definition
of b, the Burgers circuit has to be made around the line in the sense of
a right-handed screw. If the line sense changes, the sense of the circuit
also changes, and with it the sign of the vector connecting the beginning
to the end of the circuit in the disturbed crystal. Every dislocation is
therefore characterized by two equivalent descriptions: (b, I) and
( - b, -I). On the other hand, (b, -I) and (- b, I) respecitvely describe
a different dislocation. A geometrical function, F(b, I) of dislocations
has to be formulated in such a way that
1. F(b,I)=F(-b, -I). and (5.2-1)
2. F(b, I) =l=F(b, -I) or F( - b, I) , respectively. (5.2-2)
The following distinction has to be made between a vector and a
pseudo-vector: A vector is fixed in space, independant of the coordinate
system; e.g. if the coordinate axes (unit vectors) are inverted u. -+ -u.
(i = 1,2,3), the coordinates also change their sign~' -+ -~'; a pseudo-
vector, however, changes sense along with the coordinate system so
that the values of the coordinates remain unchanged, i.e. ~i -+ ~i. b is
therefore a pseudo-vector with I fixing a coordinate axis.
We have also to distinguish clearly between the two terms "line
sense" and "line orientation". A straight line can, for example, lie in a
North-South orientation, and can point in the" sense" of the arrow
towards either North or South. In every orientation, therefore, there exist
two possibilities for the line sense.
d) b has to be a translation vector of the lattice because only then
can the crystal be "perfect" although elastically distorted (good material)
outside the dislocation line. If b were not a translation vector, the crystal
would be disturbed over the whole surface of the Volterra cut. This will
become evident in Section 8.1 which deals with partial dislocations.

5.3. Relative Orientation of b and I

The Burgers vector is constant in length as well as in orientation
and sense along a dislocation line. The dislocation line, however, can
arbitrarily change its direction. Thus, the angle between b and I can
take any value, but the following two limiting cases are important:
1. b 1-1 is termed "edge orientation", and
2. b III the "screw orientation".
46 5. The Individual Dislocation: Geometrical Basis

Straight dislocations in these limiting orientations are often called

edge or screw dislocations. In the edge orientation an additional plane
appears to be built into the crystal (Fig. 5.2/1). The term" screw orienta-
tion" expresses the fact that the original parallel crystal planes are now
joined as a helical surface (Fig. 5.2/2). Fig. 5.3/1 shows a dislocation
changing from edge to screw orientation.
A general dislocation can be understood as being formed of a screw
and an edge component. The Burgers vectors of these components are:
a) screw component bscrew = b . cos (at) , (5·3-1)
b} edge component bedge = b . sin (at), (5·3-2)
where at is the angle between land b. (at = 0 = screw orientation, at =
11:/2 = edge orientation.)

Fig. 5.3/1. Change of a dislocation from the edge- (1) to the screw-orienta-
tion (2)(b is drawa here in the SF/RH convention!)

By the sign definition (FS/RH), used in this book, a right-handed

screw is represented by b it l, and a left-handed screw by b it l. The
character of a screw component can be determined in the following way:

7'
sgn(l.b)=,
+
right-handed screw
(5·3-3)
" - left-handed screw

Using the FS/RH convention, the right handed screw becomes positive.
This harmonises better than the negative sign with our predominantly
right-handed world.
It might be of interest to determine from a given land b the position
of the additional plane of the edge component in a general dislocation.
Instead of going through the definition of the Burgers vector one may
write
p= [lx bJ. (5·3-4)
 5.4. Motion of Dislocations 47

As the crystal is virtually divided into two parts by a plane which is

determined by I and b, the vector p points into the part containing the
additional layer. The two geometrical statements (5.3-3) and (5.3-4)
fulfill the invariance condition: F(l, b)=F( -I, -b) [see (5.2-1)].

5.4. Motion of Dislocations

One may draw an analogy between a dislocation and a water wave:
A wave can pass along while the water particles move in circles around
one spot. A dislocation moves in a similar fashion, whereby individual
atoms are only slightly displaced. The motion of dislocations is the actual
process of plastic deformation. A dislocation line can, in principle, move
arbitrarily (except along its own extension). However, two limiting
cases must again be distinguished:
1. Glide, the motion in the glide plane which is determined by the
path of the dislocation line and by the direction of the Burgers vector.
This motion involves no mass transport, i.e. no diffusion. The atoms
above and below the glide plane are displaced by b with respect to each
other (Fig. 5.4/1). Since, in the screw orientation, the directions of b

b'J I I I I
1 11 11
I -;
I --\ I ~
I \ ! I "\ ! / I \ / /
Fig. 5.4/1. Motion of a dislocation by glide

and I coincide, a screw dislocation can, in principle, glide on any plane

passing through the dislocation line.
2. Climb. In pure climb the dislocation moves perpendicularly to the
glide plane. This motion is only possible if
a) the orientation of the dislocation has an edge component, and
b) if the temperature or the stress are sufficiently high so that dif-
fusion becomes possible. Fig. 5.4/2 shows that climb involves the build
up or removal of the additional plane.
At low temperature glide is the only mechanism responsbile for the
plastic deformation, whereas climb enters into part at high temperature
processes (recrystallisation, creep).
If a dislocation moves it divides the crystal virtually into two parts,
one of which is displaced with respect to the other. The amount and
48 5. The Individual Dislocation: Geometrical Basis

--
--
'\. o~ ~
1"'- <""""" <==
0

" <=
,,,0 tl°
=*"
l/ ~
~
/ /
/ ,/

\
I \ I \
\ I I I I
Fig. 5.4/2. Motion of a dislocation by climb

direction of this displacement can be determined by the following rule,

which we term the" movement rule". First the following vector product
is formed:
[lxv] =m (5.4-1)
where
l is the direction of the line, and
v is the direction of the motion (v is reminiscent of the velocity vec-
tor).
The vector m, as defined by Eq. (5.4-1), thus is normal to the plane
swept over by the dislocation and marks that part which is displaced
+
by b while the other part remains fixed. This rule holds for any motion
of dislocations in either edge or screw orientation, for glide as well as
for climb. It fulfills the invariance condition (5.2-1). Fig. 5.4/3 shows two
examples; one for glide and the other for climb.

a b
Fig. 5.4/3 a and b. Movement rule. a) Glide, b) climb

The movement rule offers the possibility of defining the Burgers

vector without taking into account the distortion of the crystal. One
only has to compare the states of the crystal before and after the passage
 5.5. Dislocations and Moire-Figures 49

of the dislocation. The crystal is perfect and undistorted in both cases.

The relative displacement of the two parts of the crystal between these
states related to the vectors I, v and m, according to Eq. (5.4-1), is the
Burgers vector.

5.5. Dislocations and Moire-Figures

Let us assume two model crystals consisting of homogenious layers,
vertical to the plane of the paper (Fig. 5.5/h); the crystal 1 lies behind
that plane and the crystal 2 in front of it. Initially, there shall be no

A 8 A 8 A

b
Fig. 5.5/1. a) Moire pattern, b) corresponding screw dislocations

interaction between them. There are areas, in which the crystals match
perfectly, that is where the lines cross (A), and areas of particularly bad
matching, i.e. where the two sets of lines are separated by half the line
spacing (B). This situation can be represented by the moire model
No.4. We obtain a series of light and dark bands. The light bands are
areas of good fit and the dark bands of bad fit.
4 Bollmann Crystal Defects
50 5. The Individual Dislocation: Geometrical Basis

Now we assume the two crystals to grow together. The regions of

good fit (A) will extend by relaxation and those of bad fit (B) will contract.
Since the crystal prefers order to disorder, the bands of bad fit will be
contracted as far as possible; these bands are the dislocations (Fig. S.S/1 b).
The dislocations thus lie in the middle of the dark moire bands. In this
way, dislocations (screw dislocations in our example) can be represented
by means of rigid moire models.
Edge dislocations can be represented by two series of lines in the same
orientation, but with slightly different spacings. This situation can also
be shown with model No.4, only here the two sheets have to be held
at a distance so that the far one appears, in perspective, to be shortened.
This region corresponds to that part of the crystal containing the ad-
ditional plane.
The models can be used to illustrate the glide mechanism and to
check the movement rule. We can, for example, assume that thefront
part ofthe crystal in Fig. S.S/1 is displaced byb (upwards) and determine
to which side the dislocation will move, i.e. in which direction the vector v
will point. From (S.4-1), by cyclic permutation of the terms, we obtain:
v= [mxl] . (S.S-1)
If m points forward and l upwards, v will point to the left, i.e. if the
front part (m) is displaced in direction b upwards, the dislocation will
move to the left.
A twist of the Moire models in the opposite direction represents left-
handed screw dislocations. Models No.4, taken as edge dislocation, can
be used for analogous considerations; however, the signs for I and b
have first to be fixed correctly by the Burgers circuit of by (S.3-3).
We shall see that the relations between dislocations and Moire
figures can be extended much further, and that the latter provide a
useful means for studying dislocation arrangements.
 6. Interaction between a Dislocation
and its Surroundings
6.1. Outline
Up to now the situation and behaviour of a dislocation have been
considered essentially independently of the surroundings. Now, these
surroundings appear in the fonn of an external stress field acting on the
dislocation. The forces which can move the dislocation originate from
this stress field. In this chapter a representation of the stress field will
first be given and afterwards the forces on a dislocation will be discussed.
Dislocations have associated with them a certain amount of energy,
i.e. the work required for their fonnation. A curved dislocation line can
lower its energy by stretching, hence it is subject to a line tension.
Furthennore, the dislocation itself produces strain and stress fields
which, in turn, act on other dislocations, as will be shown in the fol-
lowing chapter.

6.2. Representation of the Stress Field

We introduce here the necessary basic notions and refer to the litera-
ture for the further development of the theory of elasticity, e.g. Hirth
and Lothe (1968).
Different types of forces can act on a body, or specifically on a
crystal: forces originating at the external surface, for example at the
grips in a tensile specimen, or forces spread over the whole crystal, such
as gravitation or magnetic forces (volume forces). Internal forces can
also operate, e.g. forces due to inhomogeneous thermal expansion. We
consider a static crystal, i.e. a crystal in which all the forces acting are
distributed in such a way that their sum, as well as the sum of the angular
momenta, is zero.
In order to determine the forces acting inside the crystal at a given
point we may imagine an infinitesimal cube cut out of the surroundings
of the point under investigation. Next we assume that forces are homo-
geneously distributed over the cube faces such that the cube is placed
into the same state of stress as that which existed inside the crystal.
Since the volume forces decrease with the third power of the linear
dimensions they can be neglected. The force per unit surface area is
called stress. 111 is the stress on that cube face on which the positive x-axis
52 6. Interaction between a Dislocation and its Surroundings

is perpendicular and pointing outwards (accordingly, tT2 and tT3 for the
positive y- and z-axis). tT-1 is the stress on the opposite side of the cube.
Each of these stresses is split into three components corresponding to
the three directions of the coordinate axes: e.g. (tTl = O"u, O"u, 0"13)
(Fig. 6.2/1).
z

x
Fig. 6.2/1. Stress components on the surface of an "infinitesimal" cube

From the condition that the cube should be immobile the following
relation holds for the equilibrium of forces:

(i=1···3,k=1···3) (6.2-1)
and, from the equilibrium of the angular momenta, we have
(6.2-2)
This leads to 6 independent components for the stresses, forming a
symmetrical tensor:

(6.2-3)

The diagonal elements, O"ii' if they are positive, stand for tensile stresses,
otherwise for compressive stresses. Components with mixed indices O"ik are
shear stresses. Occasionally, the shear stresses are indicated by T. This
tensor defines all the stresses at one point (except for the case where
internal torques have to be taken into account (see Hirth and Lothe,
1968, p.31). Obviously all six components of the stress tensor may
depend on the location within the crystal. Elasticity theory deals with
all the possible stress functions and the relations between stress and
strain. We shall apply the results obtained by elasticity theory; for
their derivation we refer again to Hirth and Lothe (1968).
 6.3. Forces Acting on a Dislocation 53

6.3. Forces Acting on a Dislocation

In order to analyse the effect of "forces" on a dislocation we have to
remember that a dislocation is moving through a crystal as a configura-
tion which is not linked to specific particles and therefore does not present
a fixed point for the forces to act on.
If, on the other hand, we admit that the dislocation configuration
possesses a certain physical individuality, we can also assume that its
motion is caused by "forces". In this case we can derive the "force"
from the definition of the "work" (= force X distance moved), i.e.
force = work per unit distance moved.
Fig. 6.3/1 shows a crystal traversed by an edge dislocation, with a
shear stress, i, acting on the surface of the crystal. If the dislocation
moves through the whole crystal, and thus virtually divides it into two
parts, one part is displaced with respect to the other by b [see the
movement rule Eq. (5.4-1)J. The work done by the total force on the
crystal is then given by:

~--------a--------~

Fig. 6.3/1. Work done on a crystal by a dislocation

Work = Shear stress (force per unit surface) X surface X distance Ibl
W=i(l. a) . b. (6.3- 1)

This work can also be considered as the work done by a jorce on the
dislocation:
Work = force per unit length of the dislocation
X length of the dislocation
X distance moved by the dislocation

W= j.l.a. (6.3- 2)

From the comparison of these two equations it follows that the force
per unit length of the dislocation is given by:

j=i· b. (6·3-3)
54 6. Interaction between a Dislocation and its Surroundings

This force, I, is thus proportional to the shear stress 't and the absolute
value of the Burgers vector, b.
In general, the differential of the force, df, on a dislocation is given
by the formula of Peach and Koehler * :
(6·3-4)
with
b = Burgers vector,
dl = vector element in the direction of the dislocation line,
fib = stress on the plane normal to the Burgers vector, and on which b

points outwards.
The formula indicates that df is always normal to dl, i.e. the force
always acts normal to the dislocation line. It has no tangential component
in the line. The dislocation line cannot move within itself; forces in the
direction of the line are therefore meaningless. Furthermore, df is also
normal to fib' Fig. 6.3/2 shows an example of a force with a component
in the glide plane. In contrast to the example of Fig. 6.3/1 the vertical
component of the shear stress has to be taken into account. However,
due to the symmetry of the stress tensor both components are equal.

Fig. 6.3/2. Peach and Koehler formula

The Peach and Koehler formula also applies to climb but, as shown
by Weertman (1964), the conditions have to be defined more precisely.
Under hydrostatic pressure, P, the dislocations do not climb, even if the
temperature is sufficiently high to provide the necessary mobility of the
vacancies. It is therefore meaningful to replace the general stress, a.,. by
a." + P with P = -1/3 (au +a22 +(33 ) = hydrostatic pressure. The
external stress, defined by the Peach and Koehler formula, may be
balanced by other forces, such as "friction" or "chemical" forces, the
latter being produced, for example, by the change in the equilibrium
concentration of vacancies on movement of the dislocation.
* Privately communicated in this form by Professor J. S. Koehler.
 6.4. Energy of a Dislocation 55

6.4. Energy of a Dislocation

Work is required for the creation of a dislocation in a crystal. This
work is stored inside the crystal as positive potential energy. The energy
of a dislocation, naturally, depends on its length but, for dislocations of
equal length it is further determined by their screw or edge character
as well as by their curvature. In the following we calculate the energy
of a length L of an infinitely long straight dislocation in a screw orienta-
tion.
Fig. 6.4/1 shows an unrolled cylindrical element of a screw dislocation.
This element is deformed in shear from a rectangle into a parallelogram.
The angle r is given by:
b
r= 2nr (6.4-1)
with
r = shear strain,
b = length of the Burgers vector = displacement,
r = radius of the cylinder.

Fig. 6.4/1 a and b. Geometrical situation for the calculation of the energy of a
screw dislocation. a) Volume element inside the cylinder, b) evolution of the
volume element
56 6. Interaction between a Dislocation and its Surroundings

The required shear stress, T is defined by

T=Gy (6.4-2)
G = shear modulus (for iron ~ 8 . 1011 dynes/cm 2 ~ 8,000 kg/mm 2) •

By substituting the relation for the shear strain, y we obtain

Gb
i = 2nr • (6.4-3 )

The elastic energy per unit volume, u is

U=iTY· (6.4-4)
This is equal to the work done during the elastic deformation of a volume
element. Substitution of the above values for T and y gives:
1 Gb 2
U=Sn 2 r 2 ' (6.4-5)
The volume element is
dV=L·2nr·dr. (6.4-6)
Thus the energy content of this cylindrical volume element becomes:

dU=u. dV= Gb 2 L .~. (6.4-7)

4n r
In order to calculate the total energy of the dislocation we have to split
it into two parts:
1. the core, i.e. the "bad" material, in which the deformation is so
large that the laws of elasticity are no longer valid. We assume the
radius of this core to be b, i.e. the Burgers vector. This corresponds very
roughly to one atomic spacing,
2. the elastically deformed" good" material.
The total energy follows as:

"
U=U
c
+ Gb4nL
2
J~
r
=TL +
c
Gb 2 L In
4n
(~)
b' (6.4-8)
b
where ~ is the core energy.

Discussion
1. U oc L. This holds, in principle, for long straight parts of the
dislocation. We can therefore write: U =r· L, with r being the energy
per unit length.
2. r contains the term r;" the energy of the dislocation core per
unit length. We obtain a rough estimate of this energy by considering
the dislocation core, with its highly disturbed atomic arrangement, as a
 6.4. Energy of a Dislocation 57

row of atoms in the liquid state. In this caSe 1;, equals the number of
atoms per unit length times the heat of fusion per atom. A closer estima-
tion is obtained from:
(6.4-9)
This leads to
Gb (1'
r= ----;;n In f +2 ) .
2
(6.4-10)

3. rincreases with In (1'o/b) , i.e. if 1'0 tends to infinity, r also becomes

infinitely large, but slower than 1'0. This means that in an infinitely large
crystal the value of the energy per unit length of a dislocation also be-
comes infinite. However, as no infinitely large crystals exist, 1'0 has to
be arbitrarily fixed. For example, 1'0 can be the size of the crystal, e.g.
1 cm, or the size of a crystal grain, e.g. 10 (.L, or in highly deformed
material the distance to the next dislocation. If we put 1'0 = 70 b ~ 170 A,
rbecomes
(6.4-11)
and for 1'0 = 4 . 104 b ~ 10 (.L:
(6.4-12)
4. roc shear modulus G. This means that in a "hard" crystal (hard
with respect to elastic deformation) the energy of the dislocation is
r
higher than in a soft one. This is obvious, since is the work necessary
to produce a unit length of dislocation.
5. rocb 2 can be explained by the fact that the elastic force is pro-
portional to the displacement, i.e. ocb, and that the energy (work =
force X distance moved) is thus proportional to b2 ; this proportionality
is useful in the estimation of dislocation interactions. On the other hand,
we can also see immediately that dislocations with the smallest possible
Burgers vector will be preferentially formed, since for two distant
dislocations the energy is oc b2 + b2 = 2 b2, whereas for united dislocations
it is oc; {2b)2=4b 2 •
6. Up to now we have only dealt with screw dislocations. The energy
of an edge dislocation is higher:
r edge= rscrew
1 _II (6.4-13)

= Poisson's ratio = - (A wlw)/{Alll) R:l1/3,

'/I

(A wlw) = relative change of width (perpendicular to the traction),

(Alll) = relative change of length (direction of traction),
w=width (diameter of rod),
l=length.
58 6. Interaction between a Dislocation and its Surroundings

Since" is ~ 1/3 in usual material, we have

I'edge = 3/2 F.crew . (6.4-14)

The energy of a dislocation in edge orientation is thus about 50% higher

than that of a dislocation in screw orientation.
7. The energy of a dislocation of arbitrary character can be calculated
by summing the energies of the screw and edge components. If ex is the
angle between b and I, the screw component of the Burgers vector is
b . cos ex and the edge component b . sin ex. The energy per unit length is
then:
sin2 oc )
r=F.o (coszex+--
1-'11 (6.4-15)
with
(6.4-16)

As an example, we may estimate the energy stored by dislocations

in 1 cm3 of heavily deformed iron (G = 8.2· 103 kg/mms, b = 2.5 A). We
assume a dislocation density, D of 1012 dislocations per cms. The energy
is given by:
Gb 2
energy=D-
2- ~2.4kg ·m. (6.4-17)

If this energy were transformed spontaneously into heat the cubic

centimeter of iron would warm up by about 7 °e.
Some numerical indications:
The shear modulus, G can be expressed in terms of the elastic modulus,
E and the Poisson's ratio,,, by:
E E
G= 2(1 +'11) ~ 2.6' (6.4-18)

Metal E (dynes/em2) G (dynes/em2) a(A) bmin(A)

Al 7' 1011 2.6' 1011 4.04 2.86

Ni 20' 1011 7.5' 1011 3.52 2.52
eu 12 . 1011 4.5' 1011 3.60 2.54
Au 8· 1011 3.0' 1011 4.07 2.88
Fe 20' 1011 8.2' 1011 2.86 2.47

1011 dynes/em ~ 1,000 kg/mm2 (gravity g = 9.81 m/see l ~ 10 m/sec2).

6.5. The Line Tension

Since the energy of a dislocation is proportional to its length, the
dislocation line will tend to adjust itself, analogous to a rubber band, in
 6.5. The Line Tension 59

such a way that its length-and therefore also its energy-becomes as

small as possible. Hence a force, the so-called line tension, T, is acting
along the line.
dU Gb 2
T= dL =r=-2~· (6.5-1)

The extension from Eq. (6.4-9) to (6.4-10) and (6.5-1), respectively, is

a rough estimation. The energy U of a dislocation depends on the
character of the dislocation, i.e. edge-or screw orientation, as well as
on its curvature and surroundings. A detailed discussion can be found
in Hirth and Lothe (1968), p. 150.
A special problem was treated by Brown (1964) (see also Hirth and
Lothe, 1968, p. 151). If a screw dislocation bows out it becomes longer
and develops edge components. The bowing out and the formation of
edge components both contribute to an increase in energy. On the other
hand, an edge dislocation also becomes longer on bowing out, thus in-
creasing its energy; however, as in this case screw components with
lower specific energy are formed the increase in energy due to the
bowing out is partly compensated by the decrease on the formation
of screw components. A screw dislocation is therefore more resistant to
bowing out than an edge dislocation.
As an example, we consider a circular dislocation, with radius r, in
its glide plane. We assume the dislocation to be of uniform character,
and thus disregard the edge- and screw orientation. If no forces were
acting on the dislocation it would contract and disappear. According to
the Peach and Koehler formula (6.3-4) a shear stress 1" acts as a force,
1=1" . b, on a unit length of the dislocation. 1is normal to the dislocation
line. We must now determine what shear stress 0 would maintain in
equilibrium the dislocation loop of radius r. We assume that the radius
increases from r to r +dr, with the force ob providing the work neces-
sary for this increase. This work can be understood as the work done
by the line tension in increasing the length of the dislocation (Fig. 6.5/1),
i.e. :
Work = force per unit length X length X distance moved
= line tension X increase in length of the line
Gb 2
dW = 0 b . 271: r . d r = -2
- . 271: d r (6.5.-2)
and thus
Gb
1"=-.
2r
(6.5-3)

The shear stress 0, necessary to maintain a circular dislocation with

radius r, is proportional to 1/r, i.e. the larger the loop, the smaller the
required shear stress. The radius r can also be considered as a local
60 6. Interaction between a Dislocation and its Surroundings

radius of curvature, so that Eq. (6.5-3) can also be applied to dislocation

segments.
The deformation of a metal is at first essentially elastic. Only when
the tension exceeds a certain limit, the yield point, does plastic de-
formation start. Some values of yield points are 15-85 kg/mm 2 for
polycristalline carbon steels and about 1 kg/mm2 for pure metal single

a b
Fig. 6.5/1 a and b. Extension of a dislocation loop under the action of a
shear stress T. a) Work done by the force I on the dislocation, b) work of the
line tension T (not T as erroneoulsy indicated in the drawing)

crystals. On the basis of the above considerations we shall try to get

an understanding of these values.
We assume that a straight dislocation is moved forward by a shear
stress. In its path it is locked at two obstacles, e.g. precipitates or other
dislocations, spaced by the distance L. We now calculate the shear
stress, T necessary to force the dislocation through the gap between the
obstacles.
At first the dislocation is straight (r= 00). Then it bows out (r de-
creases as T increases). As long as the curvature decreases with in-
creasing T, the arrangement is stable. The radius of curvature reaches
a minimum at r = L/2. With further increase of T the situation becomes
unstable, as r starts to increase again. The loop expands spontaneously.
Foriron (G = 8 . 103 kg/mm2, b = 2.8 A and L = 1 !L) Tmax is:
Gb
Tmax =L R:i 2.2 kg/mm l •

With these considerations we already fall within an order of magnitude

of the yield strength (Jy for single crystals.
If only one single dislocation loop were expanding, the material
would glide by the amount of one Burgers vector. This could not be
detected macroscopically. In order to enable glide over longer distances
 6.6. The Strain Field of a Dislocation 61

many dislocations of the same kind must be produced. This is done by

the Frank-Read mechanism.
We assume that a part of a dislocation is anchored at two points
[e.g. by precipitates or in a dislocation network (Section 11.2)]. Under
increasing shear stress it becomes curved to the shape of a semicircle.
(We neglect here again the difference between screw- and edge orientation
as well as the crystal anisotropy.) From now on the dislocation can
extend spontaneously under constant shear stress, as shown in Fig. 6.5/2,

7-

Fig. 6.5/2. Frank-Read source. (The numbering refers to the time sequence)

until finally a closed loop and a separate dislocation between the obstacles
are formed. With the latter the initial configuration is reproduced.
Further dislocation loops can thus be formed by the same mechanism,
causing a glide displacement of b. The closing of the dislocation loop, i.e.
the transition from No.6 to No.7 in Fig. 6.5/2 is a dislocation reaction.
Those will be discussed in Section 7.4.

6.6. The Strain Field of a Dislocation

Starting from the stationary position Xo in a perfect crystal the
strain is described as a deviation from the initial position by an ad-
ditional vector u (not to be confused with the unit vector). Thus, the
real position of the atoms is given by:
(6.6-1)
Every dislocation is surrounded by a strain field which is described by
the space function of the displacement vector u with the components
(u, v, w).
62 6. Interaction between a Dislocation and its Surroundings

We indicate here the results for a dislocation in screw- and one in

edge orientation (Read, 1963, p. 115). The dislocation line shall point
in the positive z-direction and the material is assumed to be elastically
isotropic.
Strain Field of a Screw Dislocation (right-handed screw)
in Cartesian coordinates (:v, y, z):
u(x, y, z) =0
v(x, y, z) =0 (6.6-2)
b y
w(x, y,z)=-2
n
arctan-;
x
in cylindrical coordinates (1', 6, z):
u(r, 0, z) =v(r, 0, z) =0
b (6.6-3)
w(r, 0, z) = 2:ii0.

It is obvious that the displacement, which transforms the xy-plane into

a screw plane, points in the z-direction (Fig. 6.4/1).
Strain Field of an Edge Dislocation
in Cartesian coordinates (:v, y, z):
(The dislocation line is assumed to lie in the z-axis and the extra
half-plane in the positive y-axis.)

u(x,y,z)= 2bn [arctan(~)+ 2 (/-v) XBX':yB]

v(x, y, z) = 8n(;~v) [(1-2V) .In(xll+y2) + ;:~~:] (6.6-4)
w(x, y, z) =0
v = Poisson's ratio (Section 6.4).
Fig. 6.6/1 represents a calculated example of this strain field (v = 1/3);

°
in cylindrical coordinates (1',6, z):
(With the angle starting on the positive x-axis.)

u(r, 0, z) = :n [0 + =i0~~)]
b [ 1 -2v cOS(20) ] (6.6-5)
v(r, 0, z} = - 2:ii2(1 -v) In(r) + 4(1 -v)
w(r, 0, z) =0.
Eq. (6.6-4) shows that for y > 0 the crystal is stretched in the direction
of the x-axis and for y < 0 it is compressed. Furthermore, the displace-

° °
ment by the Volterra cut can be seen in Eq. (6.6-5) from the difference
in u between = 0 and = 2~.
 6.7. The Stress Field of a Dislocation 63

The Burgers circuit can be expressed formally by a closed line integral.

The Burgers vector thus becomes
b= ~ ~: ds (6.6-6)

with the sence of path of the integral being linked to the line sense of
the dislocation, forming a right-handed screw.

I'" I'" I'" '"

'" '" I'"

'" '" '"

'" '" '"
<:J <:J
'" '" <:J

'" '" '" '" '" x

'" '" '" '"
v v
'"
v

'" '" '"

Fig. 6.6/1. Strain field of an edge dislocation calculated from Eqs. (6.6-1)
and (6.6-4) (v = 1/3)

For the calculation of the integral in the above mentioned examples

of strain fields, the representation in cylindrical coordinates is chosen
with () variable and rand z constant. b (bl bs ba) is obtained for the
screw dislocation as:
bl =b 2 =0
2,.
(6.6-7)
b3=~fd()=b
2n
o
and for the edge dislocation as:
b= (b, 0, 0). (6.6-8)

6.7. The Stress Field of a Dislocation

A stress field (1 is necessarily coupled with the strain field of the dis-
location u. Whereas the strain field is a vector field, the stress field is,
64 6. Interaction between a Dislocation and its Surroundings

as indicated in Section 6.2, a tensor field (symmetrical tensor). We

consider the same dislocations as in Section 6.6. Since the stress tensor
can be described by the force components on the surfaces of a cube, the
orientation of this cube has to be fixed. In the case of Cartesian co-
ordinates we choose its edges parallel to the three main axes (Fig. 6.7/1 a);
for the representation in cylindrical coordinates, however, the edges are
set parallel to the axes r, () and z (Fig. 6.7/1 b). It might be useful to
describe the components by cylindrical coordinates, even in the Cartesian
orientation of the cube.
y

------
x
IZ
I
I
I
I
Z x

y
r

Z x
b

Fig. 6.7/1 a and b. Coordinates of the strain field. a) Orthogonal coordinates,

b) cylindrical coordinates

a) Stress Field of the Right-Handed Screw Dislocation

in Cartesian coordinates (cube edges parallel to x, y, z):
The stress tensor is here:

(6.7-1)
 6.7. The Stress Field of a Dislocation 65

with

(6.7-2)

G=shear modulus (Section 6.4).

If the cube orientation is given in the direction x, y, z, while the compo-
nents are indicated in cylindrical coordinates the stress components
become:
Gb . ()
0'13=0'".= - -2-
nr
s1n
Gb (6.7-3)
0'2S=O'YZ= 2nrcos().
b) Stress Field of the Right-Handed Screw dislocation
in cylindrical coordinates (cube edges parallel to r, (), z):
Stress tensor:

H
0

G (6.7-4)
0
0'32
with
Gb
0'23=0'0z= 2nr • (6.7-5)
A shear stress thus acts on the plane normal to the ()-direction (i.e. on
the plane crossing the dislocation line) in the z-direction (or on the z-
plane in the ()-direction, respectively) (Fig. 6.7/2). It is proportional to
y

Fig. 6.7/2. Strain field of a screw dislocation

5 Bollmann, Crystal Defects
66 6. Interaction between a Dislocation and its Surroundings

the shear modulus G, and to the Burgers vector b and decreases pro-
portional to 1/r with increasing distance from the dislocation.
c) Stress Field of the Edge Dislocation (according to Section 6.6)
in Cartesian coordinates (cube edges parallel to x, y, z):
Stress tensor:

(6.7-6)

Y(XB_yB)
To b (Xl+ yl)1
X(XI_yB) (6.7-7)
Tob (XB+yl)1
0'33 =0'.. ='11(0'"" +O'yy)
y(2X2+yl)
= -To b'll (XI+y2)B
TO =G/2:n;(1-'II),
G = shear modulus,
'II = Poisson's ratio Rj 1/3.
Cube parallel to x, y, z, but components in cylindrical coordinates:

0'11=0'",,=- T~b sin 0(2 + cos 20)

0'22 =0'"" = T~ b sin 0 cos 20
(6.7-8)
TO b O O
0'12=0'",,= -r-cos cos2
To
0'33=O'U=--r- 2 '11s1O.
b . 0

d) Stress Field of the Edge Dislocation

in cylindrical coordinates (cube edges parallel to r, 0, z):
To b . 0
0'11 =0'22 = O'rr =0'88 = - - r - sm

0'12=0',6 =+ -r-cos
To b 0 (6.7-9)
TO
=---2'11s1O
b . 0.
0'33 = 0',.
r

The stress distribution of the edge dislocation in Cartesian coordinates

is given in Fig. 6.7/3.
 Chapters 5-6. Summary and Discussion 67

The stress distribution of an arbitrary dislocation with the line

lying in the z-axis, results from the superposition of the screw and the
edge components. The screw component is
bscrew = b . cos ex (6.7-10)
and the edge component
bedge = b . sin ex (6.7-11)

Fig. 6.7/3. Strain field of an edge dislocation

with ex being the angle between the Burgers vector and the dislocation
line.
The derivation of the stress- and strain field is given in Hirth and
Lothe (1968), p. 71.

Chapters 5-6. Summary and Discussion

In Chapters 5 and 6 the concept of the single dislocation, together
with the basic definitions were introduced. First we looked at the
" skeleton" of the dislocation (line, Burgers vector in 5.2, screw- and
edge orientation in 5.3). Next, the dislocation was affected by its sur-
roundings (motion, 5.4; forces on the dislocation, 6.3), and finally it
became active (energy, 6.4; line tension, 6.5) and, in its tum, affected
the surroundings (strain- and stress fields, 6.6 and 6.7).
The following points should be emphasized:
a) Isotropic Elastic Medium. Up till now the crystalline character of
matter has hardly been stressed. Only in the definition of the Burgers
5·
68 6. Interaction between a Dislocation and its Surroundings

vector was it mentioned that the Burgers vector has to be a translation

vector of the crystal lattice. All the other considerations are also valid
for isotropic elastic materials which contain a dislocation introduced by
a Volterra cut. Therefore, the basis is pure isotropic elasticity theory.
Some modifications are, of course, possible and occasionally even neces-
sary, and these can be made by the application of the elasticity theory
of anisotropic media.
b) Pseudo-Vector Character of the Burgers Vector. The Burgers vector
is invariant along a dislocation in magnitude, orientation, and sign,
independent of the curves in the dislocation line, as long as the line
sense is uniformly attributed to the line. If the line sense is inverted, the
sign of the Burgers vector is also inverted. Its magnitude and orientation,
however, remain unchanged. The sign of the Burgers vector can therefore
only be given together with the line sense. The coupling between the
sign of the Burgers vector and the line sense has to be taken into account
in the correct formulation of geometrical statements, such as, for example,
the movement rule [Eq. (5A-1}J and the Peach and Koehler formula
[Eq. (6.3-4)]. Only then is a consistent formulation possible.
This invariance is guaranteed only if the Burgers vector is con-
sidered as an "abstract" pseudo-vector, i.e. as a coordinate difference,
as demonstrated by Frank by transcription on to the perfect reference
lattice. Only then is the Burgers vector independent of the local crystal
orientation and strain. Hence, the expression "local Burgers vector"
should be avoided. It might be replaced by "local lattice vector".
c) The force always acts normal to the dislocation line. At first it
does not seem obvious that, in spite of a directed shear stress '£' and a
directed displacement b, the dislocation loop under stress should expand
uniformly in all directions, even against'£' and b. We have to remember
that a dislocation line is basically a boundary line between a "displaced"
and an "undisplaced" part ofthe crystal [see movement rule (5A-1}J. The
motion of a boundary line only makes sense in a direction vertical to
its orientation. At the border between two countries, for example,
nothing would be changed if all the border stones were moved within
the direction of the boundary line; besides, in our case the border stones,
i.e. the atoms, remain practically in their positions while a dislocation
moves. A force component (work per unit distance) in the direction of
the line is therefore meaningless. The Peach and Koehler formula
automatically eliminates the tangential component of the force.
d) The energy per unit length of the dislocation is partly concentrated
in the dislocation core; the greater part of it, however, is spread over the
surroundings, i.e. the stress field of the dislocation, and extends, in
principle, over the whole crystal. It diverges for a crystal as r tends to
 Chapters 5-6. Summary and Discussion 69

infinity. An exact delimitation of the energies of individual dislocations

is therefore not possible. The arbitrary definition of the effective radius
allows estimations to be made.
For the dislocation theory in general a kind of "uncertainty principle"
seems to apply. One either simplifies the formulae employed, thus
leading to inaccuracy, or one tries to carry out the calculation as exactly
as possible. In the latter case, however, the calculations involve a number
of parameters which are either only approximately known, or which have
to be estimated, and therefore inaccurate results are also obtained.

Suggestions for Problems

Chapters 5 and 6 have been kept short in this book since most of the
problems described are given in greater detail in other books, e.g. Read
(1953), Weertman and Weertman (1964) and Hirth and Lothe (1968).
These books also suggest many problems.
It might be particularly useful to practise the use of the movement
rule and the Peach and Koehler formula, and to become familiar with
the pseudo-vector character of the Burgers vector.
 7. The Interaction of Dislocations
7.1. Outline
Up to the present stage the properties of an individual dislocation in
a crystal have been investigated. The knowledge of the stress field of the
dislocation makes it possible now to understand the interactions between
two dislocations. This leads to the study of dislocation reactions, i.e. the
nodes formed by dislocations. At the end of the chapter dislocation
reactions in special crystal structures will be investigated.

7.2. Interaction between Two Parallel Screw Dislocations

The Peach and Koehler formula (6.3-4) indicates the force which
acts on a dislocation in a stress field. This stress field can originate
either from external forces or from other dislocations. We consider the
interaction between two straight parallel ruslocations of the same type.
One of them shall lie in the z-axis.
In the first example we deal with two right-handed screw dislocations.
In analogy to the considerations of the forces between charged particles
in electrostatics we assume the dislocation in the z-axis to be the source
of the stress field and investigate the action of the force due to this
stress field on the other dislocation. The Burgers vector and line sense
may point in the direction of the positive z-axis. We can here directly
apply the representation in polar coordinates. fib thus becomes [Eq.
(6.7-5)] :
(7.2-1)
Together with Eq. (6.3-4) the force per unit length of the dislocation
becomes:
Gb GbZ
/=Ibl· -2n
- [UoXU.] = -2- Ur
r· nr
(7.2-2)
Ur 'Uo, u.=unit vectors in the direction of the respective coordinate
axes.
Hence, the force between two equal screw dislocations is repelling
and decreases oc: 1/1'.
If the dislocation at the position r is of the left-handed screw type,
u. and / change their signs and the two dislocations of opposite signs
now attract each other.
 7.3. Interaction between Two Parallel Edge Dislocations 71

If the first dislocation, lying in the position of the z-axis, has right-
handed screw character with a Burgers vector b 1 and the other parallel
dislocation with b 2 is of mixed character, the force between them is
(Hirth and Lothe, 1968, p. 117):
f -- G(b1 • '1) (b 2 • '2)
2n r ur· (7.2·-3)
'2
"4 and being the unit vectors in the line sense. Eq. (7.2-3) fulfills the
invariance condition (5.2-1): F(l, b) =F( -I, -b).

7.3. Interaction between Two Parallel Edge Dislocations

Again we assume one edge dislocation lying in the z-axis acting as the
source of the stress field. The other edge dislocation may be in an arbitrary
position, but parallel to the first. Both dislocations are located such that
the additional half plane [Eq. (5.3-1)J points to the side of the positive
y-axis. The Burgers vectors of the two dislocations are assumed to be
equal. First we consider the case of the second dislocation being able
to glide on a plane parallel to the xy-plane, but not to climb. Hence, the
force on the second dislocation has to point in the x-direction and, ac-
cording to the Peach and Koehler formula [Eq. (6.3-4)J it is produced
by the shear stress component Gxy .
X(X2_y2)
Ix = b . Gxy = 7:0 b2 lx2 + y2)2- (7.3- 1 )
with
G
7:0= 2n(1-v) (7.3- 2)

The following segments can be distinguished (Fig. 7.3/1):

1.x>O Ixl>IYI Ix>o
2. x>O Ixl<IYI Ix<O
3. x<O Ixl<IYI Ix>o
4. x<O Ixl>IYI Ix<o.
This shows that, in the segments 1 and 4 the force is such that the dis-
locations repel one another, whereas they attract one another is segments2
and 3. Thus, for two equal parallel edge dislocations there is a stable
situation for glide at x = O. In the case of two edge dislocations with
opposite signs stable situations also exist, but here they appear for
Ixl=IYI·
The physical reason for the stability of the arrangement of two
dislocations placed on top of each other lies in the fact that they form
part of a possible interface between two crystals, slightly tilted with
respect to one another (Section 11.3).
72 7. The Interaction of Dislocations

If, in addition to glide, any other motion of the dislocations is

allowed, dislocations of the same kind repel one another whereas dis-
locations of opposite kind attract one another. The force vectors, how-
ever, do not necessarily lie in the plane defined by the two dislocations.
If the dislocations are in an arbitrary orientation with respect to one
another, the effect of every element of one dislocation on every element
of the other has to be calculated (Hirth and Lothe, 1968, p. 113).

I*1;l,,--r---:;xI
, yt / !
, I /-
-----~--~---~~-~-
I " " , I b //1 <Pl
I ,/ I '
-----t-~/
1/ I 'I ~
-----~~--+--~~-~Jl~
/ I
/,,
, I', 0
, ", 1

I, I, II
4 I 3 I 2 I
I I I
I I I
I
I, ~
I
I
I
.L
I
II _ _ II

I I I
I I )
I I I

1--,,-1
I

_ -
I T
I
I
I
I

0l
b

~~~ --f

Fig. 7.3/1. Force between two parallel edge dislocations [Eq. {7.3-1}] under
the condition that only glide is permitted. (The circles mark stable positions)

7.4. Dislocation Reactions

If two dislocations attract each other, a dislocation reaction may take
place with the formation of a single dislocation. It does not necessarily
happen over the whole length of the dislocations, so that dislocations
can bifurcate and form nodes. In these cases, the following condition
stated by F. C. Frank applies. If we consider the dislocation segments
 7.4. Dislocation Reactions 73

with their appropriate line sense and the Burgers vectors with the corres-
ponding signs, Frank's node condition states:
The sum of the Burgers vectors flowing into the node (line sense I)
equals the sum of the Burgers vectors flowing out of the node.

L b. (in) = 1.: b i (out) . (7.4-1)

• i

This node condition is a law of conservation of Burgers vectors (con-

tinuity), analogous to the conservation of electrical currents in a node
of conductors.
The node condition can also be formulated as follows:
If the line senses, and accordingly the signs of Burgers vectors are
chosen such that all the dislocations flow into the node, then the sum
of the Burgers vectors is zero.
The node condition can be proved as shown in Fig. 7.4/1, by moving
a Burgers circuit along the dislocations; the total Burgers vector has to

(d---:;S---B-
remain unchanged as long as no dislocation is cut.

---------------
Fig. 7.4/1. The proof of Frank's node condition. (Displacement of the
Burgers loop)

If a dislocation reaction occurs, the node condition furnishes the

balance of the Burgers vectors; it does not indicate, however, whether
this reaction will take place or not. This is determined by the increase
or decrease in energy of the crystal during the reaction. Neglecting all
the other influences, the energy per unit length of a dislocation is ap-
proximately proportional to b2 [Eq. (6.4-10)]. According to this criterion
two dislocations 1 and 2 will combine to give one dislocation 3, if
b~+b:>b~. (7.4-2)
From the node condition, it follows that
(7.4-3)
This means, geometrically, that the vectors b1 , b 2 and b s together have
to form an acute-angled triangle (Fig. 7.4/2a). An equilibrium exists if
(7.4-4)
74 7. The Interaction of Dislocations

i.e. if the three Burgers vectors form a right-angled triangle (Fig.7.4/2b).

In the case of
(7.4-5)

the Burgers vectors form an obtuse-angled triangle. A dislocation re-

action would result here in an increase in the energy of the crystal, so
it is not favoured (Fig. 7.4/2c).

:2 ~
111
, __
~ __
•

Fig. 7.4/2 a-c. The relative stability of dislocation reactions: a) stable,

b) labile, c) unstable

The energy difference between a screw dislocation and an edge

dislocation has not been considered in these estimations. Taking this
difference into account [Eq. (6.4-12)], the limiting angle for the stability
in the Burgers vector triangle may differ from 90°.
 7.S. Dislocation Reactions in Special Crystal Structures 75

7.5. Dislocation Reactions in Special Crystal Structures

From the geometrical definition of the Burgers vector it follows that
it must be a translation vector of the crystal. Moreover, the energy
considerations [T,..., b2 , Eq. (6.4-10)J show that the most favorable
Burgers vector is the smallest possible translation vector, since the
energy of a dislocation with a double Burgers vector would be four
times that of the simple dislocation. The smallest possible translation
vectors in different crystal structures are given in Sections 3.3 etc.
A discussion of several typical examples follows:

Face-Centred Cubic (FCC) Structure

1. The basic reaction is here of the type (Fig. 7.5/1):

(1/2) [110J+(1/2) [011J=(1/2) [101J* (7.5-1

, )

Fig. 7.5/1. Dislocation reactions in the f.c.c. structure. 1. Stable, 2. labile,

3. unstable

with the b2-balance (= b T G b) :

a2 /2 + a /2 > a /2.
2 2 (7.5-2)
The reaction will thus happen in the "direction of the arrow".
2. A situation where the reaction may go either way is:
(1/2) [110J+(1/2) [110J=[010J (7.5-3)
with the b2-balance
(7.5-4)
* We write the Burgers vectors in dimensionless coordinates, i.e.
(1/2) [11 oJ instead of (a/2) [110]. Thelattice constant, a will appear for
ibi2 through the metric tensor G. This notation is of special advantage for
non-orthogonal systems.
76 7. The Interaction of Dislocations

3- The following reaction will only occur with difficulty, smce it

creates an unstable situation:
(1/2) [110J+(1/2) [110J=[110J (7.5-5)
with the b2-balance:
(7.5-6)
The corresponding reactions also apply for the rock salt and the diamond
structure, as these two structures contain the same translation unit as
the face-centred cubic structure.

Hexagonal Structure
The hexagonal structure has a preferential plane, the basic plane
(001). The smallest translation vectors are of the type [1 OOJ. A possible
reaction is as follows (Fig. 7.5/2):
[100J + [010J = [110J (7.5-7)

o o
Fig. 7.5/2. Dislocation reactions in the c.p.h. structure. 1. Stable, 2. unstable

with the b2-balance [with G from Eq. (3.7-1)J:

(7.5-8)
and an improbable reaction:
[100J + [110J = [210J (7.5-9)
(7.5-10)
In several hexagonal materials glide is, in principle, also possible on
"prism planes" (100) and "pyramid planes" (102); however, the
Burgers vectors in the c-direction become very large.
The corresponding reactions in the basal plane occur also in graphite.
However, because of the pronounced layer structure of graphite, the
dislocation reactions in prism- and pyramid planes are practically
excluded.
 7.5. Dislocation Reactions in Special Crystal Structures 77

Body-Centred Cubic Structure

In the bcc structure the smallest Burgers vectors are of the type
(1/2) <111 >, the next larger ones of type <100> and even larger ones
of type <11 0>.
The most important dislocation reaction is of the following type:
(1/2) [111J+(1/2) [111J=[001J (7.5-11)
with the b2-balance:
(7.5-12)
The reaction
(1/2) [111J + (1/2) [111J = [11 OJ (7.5-13)
with
(7.5-14)
is unstable (Fig. 7.513). In certain rare cases small segments of <110>-
dislocations have been observed (Hale and Brown, 1968).

Fig. 7.5/3. Dislocation reactions in the b.c.c. structure. 1. Stable, 2. unstable

 8. Partial Dislocations
8.1. Outline
We have seen that the Burgers vectors are the smallest possible
translation vectors of the lattice. When a dislocation moves through
the crystal, the Burgers vector represents a displacement (of one part
of the crystal with respect to the other) such that the crystal which was
perfect before the passage of the dislocation is again perfect afterwards.
In close-packed and related structures this shift may become divided
into two steps, and in special layer structures even into more than two.
The passage of such" partial steps" or "partial dislocations" (abbrevi-
ated to "partials") involves the formation of stacking faults. Partial
dislocations thus appear as boundaries of stacking faults. In the following
we consider the situation in the face-centred cubic and in the hexagonal
structures.
In the fcc structure the stacking sequence of partial dislocations is
not arbitrary. The determination of this sequence will be discussed in
Section 8.3. As an example of the application of the considerations on
partials, the phase transformation of cobalt will be discussed in
Section 8.4.

8.2. Partial Dislocations

in the Face-Centred Cubic Structure
We consider the face-centred cubic structure as the ABC-stacking
of close-packed layers of spheres (Section 3.3) and assume that a dis-
location glides between a B- and a C-Iayer such that the lower layers
(A and B) remain fixed and the C-Iayer is shifted (vector m of the
movement rule (5.4-1) points upwards) (Fig. 3.2/3). As the dislocation
sweeps over the B-plane, a C-atom will be shifted from one C-position
into the next. This shift can take place in two steps, the first step leading
from a C-position into an A-position, and the second from this A-
position into the next C-position. Steps of this kind are, for example
(Fig. 8.2/1-solid arrows):

(1/6) [211]+(1/6) [121]=(1/2) [110]

(1) (2) (3) (8.2-1)
Co~A + A~Cl = CO~Cl·
 8.2. Partial Dislocations in the Face-Centred Cubic Structure 79

The left hand side of the equation represents the Burgers vectors of the
two partials, which together form the total dislocation with the Burgers
vector (1/2) [110].
The b2-balance of this reaction is:
(8.2-2)
From this point of view the splitting of partial dislocations is favoured,
i.e. they repel each other. It can also be seen from Fig. 8.2/1 that the
corresponding Burgers vectors form an obtuse-angled triangle.

A 8 c [1101
[ill] /[itf)
£ 0 + £ 0
/
/
/

>
0 0 £
~4(3)+
/
(ZI r'(ZI / 11121
+A
£ £(31\ ,11111
I ,
(11 V(11 I \
0

£
+
0
C1

+
0

A 0
£
[1211 !
'[liD1
\[2ff]
a
z
• C
+B
OA

b
Fig. 8.2/1 a and b. Partial dislocations in the f.c.c. structure. Full arrows:
displacement of the part above the plane of the paper; dashed arrows:
displacement below the plane of the paper. a) projection in the [111]-
direction. (The arrows in the right-hand diagram indicate the crystallo-
graphic directions.) b) situation in space (G glide plane)
80 8. Partial Dislocations

Between the two partial dislocations, a stacking fault is formed which,

in its turn, increases the energy of the crystal. The two partial dislocations
are held together by the stacking fault ribbon.
The splitting of a perfect dislocation into two partials can be estimated
as follows: The force per unit length between two screw dislocations
[Eq. (7.2-3)] is balanced by the change in total energy of the stacking
fault, i.e. by the change in the surface of the stacking fault [times the
stacking fault energy r (Section 3.10)] on changing the spacing r of the
two partials.
I. L ~-:L G(b1
:-'»
• 4) (b s . 's) d(L . y. r)
dr =Lr (8.2-3)
4n CIt r R::j

(8.2-4)
with
1 < or. < (i-v) = correction for the deviation from the screw orientation.
The more accurate formula
r = _G_ [(b . 1) (b . I ) + [b 1X '1] . [b x
2n y 1"'1 2 2 1- v
s 's]] (8.2-5)

is derived in Hirthe and Lothe (1968) p.298. Fig. 8.2/2 shows a split
dislocation. (For the numerical values of the stacking fault energies
see remark in Section 3.10.)

Fig. 8.2/2. Stacking sequence of a dislocation split into partials

Another possibility for the splitting of a perfect dislocation is:

(1/2) [101] = (1/6) [121] + (1/3) [111].
According to the b2-criterion
a 2 /2 = a 2 /6 + a 2 /3
this reaction is doubtful. The Burgers vector (1/3) [111] does not lie in
<
a 111 )-glide plane. Such a dislocation can therefore not glide. It is
called a "sessile" partial in contrast to a "glissile" partial. The sessile
 8.3. The Sequence of Partial Dislocations 81

partials are also called Frank partials and the glissile (1/6) 112)- <
partials are termed Shockley partials. Frank partials occur where layers
are inserted or removed (condensation of interstitial atoms or vacancies).
Several possible stacking fault configurations were discussed in
Section 3.9. The boundary lines of such configurations inside the crystal
are various types of partial dislocations.

8.3. The Sequence of Partial Dislocations

Eq. (8.2-1) shows that the displacement caused by a total dislocation
can be split up into two partial steps. After the first partial has passed,
the stacking order has changed from
t
ABCABCABC (8·3-1 )
KKKKKKK
to
t
ABCABABCA (8·3-2)
KKKHHKK.
The boldface part is that region of the crystal that has moved. The
inversion of the sequence of the partials would, after the first step,
produce the following stacking order:
t
ABCABBCABC (8·3-3)
KKKLLKKK
i.e. the labile sequence L (=ABB, etc.) would occur, with two close-
packed layers of spheres on top of one another. This shows that the
sequence of partial dislocations is not arbitrary, i.e. the vector addition
of the Burgers vectors of the partials is not commutative. In Fig. 8.3/1

0 A

+ B
.& C

/
/
+"\
\
\
\

Fig. 8.3/1. Possible displacements by partial dislocations. Full lines: upper

part displaced, dashed lines: lower part displaced
6 Bollmann, Crystal Defects
82 8. Partial Dislocations

all the possible steps on the (111) glide plane that are allowed for the
upper half of the crystal are indicated by solid lines. With respect to
the movement rule (5.4-1), these are the steps for which m points in
the [111 ]-direction. All these steps together form a hexagonal arrange-
ment (analogous to the graphite structure, Section 4.9). If m points in
the opposite ([111]-) direction, the dotted steps (Figs. 8.2/1 and 8.3/1)
become possible. In the first case the stacking order changes from (8.3-1)
to (8.3-2). However, if the "bottom" part is displaced and the "top"
part kept fixed, then the stacking order after the first step is

CABCACABCA (8·3-4)
KKKHHKKK.
The splitting of dislocations therefore depends on the sign of m.
Four equivalent {iii} glide planes exist in the cubic structure and
we must now determine the stacking sequence of the partial dislocations
in all these glide planes. The four {iii} planes can be joined to a tetra-
hedron, the "Thompson's tetrahedron" (Thompson, 1953). The edges of
this tetrahedron are formed by the Burgers vectors of the type 1/2 <110),
of the perfect dislocations, and the connections from the corners to the
face centres of the triangles are composed of the Burgers vectors of the
partial dislocations ofthe type 1/6 (112). In order to fulfill the conditions
for the splitting of dislocations and also to meet the movement rule,
m has to point from the respective tetrahedral surfaces towards the
interior of the tetrahedron. Otherwise the partial Burgers vectors, taken
in the correct order, would externally surround the triangle of the
tetrahedral surface. Thus, two possible tetrahedra exist for the eight
different <111) orientations of m (Figs. 8.3/2-4). The total Burgers

o o
\
\
\
\
\
\
I
I

/
I
!
o o
Fig. 8.3/2. Composition of Thompson's tetrahedra for the representation of
dislocation splitting on all four {111 }-glide planes. 'In has to point into the
tetrahedra
 8.3. The Sequence of Partial Dislocations 83

x
Fig. 8.3/3. Position of Thompson's tetrahedra in the cubic elementary cell

vector b results from the addition of the partial Burgers vectors b(l)
and b(2); the first term of the vector addition, b(l) is assigned to that
partial which points in the direction of v of the.movement rule
[Eq. (5.4-1)]. I and v can always be defined in such a way that m cor-
responds to one of the four <111) directions of Fig. 8.3/4. When, on
splitting the" perfect" Burgers vector b, the partial b(l) points to the
centre of a tetrahedral face, the sequence of partial dislocations can be
determined geometrically in an unambiguous manner by means of the
Thompson tetrahedron.
Here follows a description of an algebraic procedure for deter-
mining the sequence of partials. This procedure can be useful, for ex-
ample, in connection with computer calculations (Bollmann and Perry,
1967).
We define:
the line sense I,
<
the (total) Burgers vector b [of the type (1/2) 11 0)],
the glide plane given by m, of the form <111).
6·
84 8. Partial Dislocations

The different steps are:

1. An auxiliary vector c is calculated
c=(1/3) [bxm]. (8·3-5)
o

o o
Fig. 8.3/4. Thompson's tetrahedron net

It is of the type (1/6) <112).

2. The vectors i£ and yare detennined:
i£ = (1/2) (b +c) y = (1/2) (b -c). (8·3-6)
i£and yare the Burgers vectors of the two partials, but it is not yet
known which is b(l) (attributed to the line 1(1) leading in the direction v),
with
v= [mxl] (8·3-7)
[from cyclic permutation of Eq. (5.4-1)].
3. The vector B is defined as:
B = [b(l) X b(2)] . (8·3-8)
It can be parallel or antiparallel to m, hence the sgn (B . m) depends on
the sequence of b(l) and b(2). This sign is detennined by:

(8·3-9)
 8.3. The Sequence of Partial Dislocations 85

The two factors on the right are the signs of the products of the co-
ordinates of c and m. Thus, with m and c given, x and y have to be
attributed to btl) and b(2) such that Eq. (8.3-9) is fulfilled.
As the vectors m and B are parallel and hence, their coordinates are
proportional, sgn (B. m) can be checked on a single coordinate, e.g.
sgn(B. m) =sgn(Ba ma)
(8·3-10)

Example
We assume a dislocation with
1=[121], b = (1/2) [11 OJ.
The screw component of this dislocation is right-handed [sgn (I· b) = (+),
Eq. (5.3-4)] and the extra plane of the edge component (p= [lxh],
Eq. (5.3-3)) is on the side [111]. This direction remains normal to the glide
plane and we choose m = [111] (Fig. 8.3/5). According to Eq. (8.3-7)
v = [m X lJ = [101]

I
-- I
I £. 1/6[11211

./
)...---- ---+----
/'
// //
x

Fig. 8.3/5. Example of the splitting of a dislocation

86 8. Partial Dislocations

and 1(1) is the partial dislocation which leads in the direction of v. From
Eq. (8.3-5)
c = (1/6) [112J
and from Eq. (8.3-6)
~=(1/6) [211J
y=(1/6) [121J.

Hence,sgn (f) m;) = (-) andsgn (IJ Ci) = (+) and thussgn(B.m) = (-)
• •
[Eq. (8-3-9)J, i.e. the sense of B is opposed to that of m. This condition
is fulfilled jf:
b(1)=y=(1/6) [121J
b(2)=~=(1/6) [211J.

In order to check the invariance properties of Eq. (8.3-9), we assume

I, b and v as given (Fig. 8.3/6, case b) while m, c and B follow from
them. We introduce the following changes:

1--+ -I, b--+-b, v--+v.

(Due to the pseudo-vector character of the Burgers vector this must

leave the physical situation unchanged.) Here m --+ -m and thus
sgn (If mi) --+ -sgn (11m;) while sgn (IJ c;) =sgn (11 [b xmJi) remains
unchanged as both band m have changed their signs. Thus:
sgn(B. m) --+ -sgn(B. m)
but, since m has changed, B stays unchanged.
From this, it follows that
b(l) --+ _ b(l) and b(2) --+ _ b(2) .

Since, at present, the (-m)-side is displaced, due to the movement of

the dislocation, the triangle formed by b(l), b(2) and b is inverted.
Further physically equivalent cases are:
d) 1--+1, b--+b, v--+-v,

e) 1--+ -I, b--+-b, v--+-v

and physically opposed is

f) 1--+ -I, b--+b, v--+v.

Case f can be compared with the others, for example by placing the
arrangements f and e (Fig. 8.3/6) side by side and letting the partials
annihilate each other in a stepwise manner.
 8.4. Transformation of the FCC-into the CPR Structure 87

In the physically equivalent cases b to e v -+ -v involves an ex-

change of the numbering of b(l) and b(2) [while (1· b) -+ ( -1· - b)
describe identical situations]. The orientation of the Burgers vector
triangle is coupled with the orientation of m and thus with v and 1.

1--1
12.--/}
liz [110J 1'-1'
\ /
\ /
\ / 1'- -
\1/6 [121J 1/6[211J / 1JJ.-0
\
\\ /
/
r; - t
/ fl-0
m~@~[mJ\ 1/6md
\ /
\ \ I
/

1~ 1. 1. --1 1 --1
/}~/} 1l.--12. 12-1l.
1'--1' l' --1' 1'-1'

1'~

IJJ.~ 0 IJJ. ~
0
r; ~

I r; =

0 fl- @

~Zl ~Z)
.Il. II

Fig. 8.3/6. Invariance properties of Eq. (8.3-13)

8.4. Transformation of the Face-Centred Cubic -

into the Close-Packed Hexagonal Structure
Certain metals, a typical example being cobalt, transform on cooling
from the fcc to the cph structure. Since the cobalt transformation
occurs at around 420°C, it cannot be diffusion-aided. The motive force
88 8. Partial Dislocations

for the transformation can be understood as being due to the decrease

in the stacking fault energy, y, with decreasing temperature; y becomes
zero at the transition temperature and negative below this temperature.
This means that below 420°C the H-sequence is perferred to the K-
sequence.

B C

K A
V I)L )/ 1)/ K B
K C ~ ~ I K A
K B k,I~ 1/1IlL-V K C
K A
'r-~V'f- V I K B

V V V
K C K A
K B --K C
K A lL~ V I-rilL K B
I V I
--K C
K B vr If vr H A
H B

ABCABCABCAB BCABCABCABC
a b

A
K C
K B
K A
-K C
K B
H A
H B
H A Fig. 8.4/1 a-c. Geometrical con-
H B KH-+-o-++-o-+-H ditions for the transformation
of the f.c.c. into the c.p.h. struc-
A ture. a), b), c) sequence of dis-
ABCABCABCABCA placements
c

The transformation can be explained geometrically as a shift of every

second layer of the K-sequence into the stacking fault position (Fig.
8.4/1):
t
A BCABCABC (8.4-1 )
KKKKKKK
 8.4. Transformation of the FCC-into the CPH Structure 89

becomes
.j.
ABABCABCA (8.4-2)
HHKKKKK
and
.j.
ABABABCAB (8.4-3)
HHHHKKK
etc.
The question now arises as to how the transformation actually
occurs. Electron microscopic observations show that the transformation
is always incomplete. A survey of the problems can be found in Christian
and Swann (1964).
We discuss here several possibilities for the transformation mechanism:
1. A closed loop of a partial dislocation, i.e. a circular disk of stacking
fault, is formed spontaneously and starts to extend. This leads to a
decrease in the energy of the crystal due to the increase in the stacking
fault surface with negative stacking fault energy. On the other hand,
the energy is increased by the extension of the partial dislocation. As
the energy decreases ex; r2 and increases ex; r, it will rise at the beginning,
reach a maximum value, and will then decrease. The radius of the
stacking fault disk at the energy maximum (~E = 0) is the critical
radius, re. For r>re the loop can extend spontaneously. r< may be
obtained from
(8.4-4)
with
y = absolute value of the stacking fault energy,
r = 1Ine
· . Gb 2
tensIOn ~ - 2 - .
This leads to
Gb 2
r~-­
e 2 (8.4-5)
with
G ~ 7.5 .1011 dyn
cm 2 ,
a a2
b=6<112), b2 =C;'
a=3·55A,
erg
y~ 5 cm 2 •

(In order to obtain a finite negative value for y, the material has to be
undercooled before start of the phase transformation. In fact, it usually
starts at around 390°C.)
90 8. Partial Dislocations

With these values, rc becomes F:::! 1,600 A. A stacking fault disk of

0.3 (J. diameter should thus be formed spontaneously before it is able
F:::!
to grow on its own. The spontaneous formation of such a large loop is
practically excluded.
2. The possibility of a "reflection mechanism" occurring at the
crystal surface has been put forward in order to explain the formation
of the stacking faults necessary for the transformation (see Christian
and Swann, 1964). It was based on the following concept: Upon cooling,
dislocations already existing in the fcc cobalt split into partials. The
included stacking fault ribbon expands until it reaches the crystal surface
where it produces a glide step. It was thought that, by dynamic effects
new partial dislocations could be formed at the crystal surface in the
surroundings of the glide step. These newly formed stacking faults could
traverse the crystal in opposite directions and so trigger the formation
of new stacking faults, and so on. The merit of this mechanism would
be that the partial dislocation could be formed almost as a straight line
and so would not have to overcome a critical radius of curvature. How-
ever, it is difficult to understand how new partial dislocations could be
formed at the crystal surface, since the glide steps increase rather than
decrease the surface energy. This is the reason why this concept has
been abandoned.
3. A "pole mechanism" has been worked out by Seeger (1953,1956).
It is based on the assumption of the existence of a dislocation which is
oriented essentially normal to a {iii }-glide plane and which has a
screw component of the type (2/3) <111). A second dislocation of the
type (1/2) <11 0) branches off the first in the glide plane (Fig. 8.4/2).
An example of such a combination is:
(1/2) [121J=(1/2) [110J+(1/2) [211J. (8.4-6)
The dislocation in the glide plane is split into two partials
(1/2) [110J=(1/6) [211J+(1/6) [121J (8.4-7)
the first of which (1/6) [211 J moves in the sense of a right-handed screw
in the [111J-direction and the second in the [HIJ direction. (This
sequence can be confirmed as an application of the methods described
in Section 8.3.) The effect of the assumed screw component is to displace
every second layer by the Burgers vector of the partial dislocation, thus
transforming the crystal from the cubic to the hexagonal structure.
The difficulties of this mechanism, as were already discussed by
Seeger, are firstly that the two partial dislocations repel one another
after half a revolution (Section 7.2), and that kinetic energy is necessary
to surmount this barrier; this effect would explain the undercooling
 8.4. Transformation of the FCC-into the CPR Structure 91

111[211] l/Z[211] l/z[io11

1/6[121] 1/6 [1211
1/z[11 0]

liz [1211 '16 [211] liz [121] 1/6[211]

l/Z [0111

a b c

Fig. 8.4/2a-c. Seeger mechanism. a) Starting condition, b) pole mechanism:

height of screw path = 2 layers, c) pole dislocation as two separate stable
dislocations

which is necessary for the transformation. Secondly, it is obvious that

the dislocations (1/2) [121J and (1/2) [211] are not stable:

(1/2) [121] = (1/2) [11 OJ + (1/2) [011] (8.4-8)

(1/2) [211] = (1/2) [110] + (1/2) [1 01J. (8.4-9)
The dislocation combination necessary for the transformation would
therefore not be formed at all, unless by chance a node of the type
(1/2) [011] = (1/2) [110] + (1/2) [101J (8.4-10)
was in the vicinity of a dislocation (1/2) [110] (Fig. 8.4/2c), with the
partial dislocations winding around the free dislocation.
If the transformation mechanism were stopped, e.g. by other dis-
locations, stacking faults, etc., another dislocation arrangement of the
same type would be necessary for the continuation of the transformation
at a new location. Electron microscopic observation of a fully recrystal-
lized material shows, however, that practically no dislocations are present
inside the grains. This means that the dislocation combination required
for the pole mechanism in cubic cobalt is extremely unlikely, although
possible in principle.
4. As a further possibility, the "reflection mechanism" was recon-
sidered (Bollmann, 1961), with the modification that the new stacking
faults were not formed at the crystal surface but internally, at differently
oriented stacking faults which were assumed to be already present. Of
course, also in this case, it is not a proper reflection, but rather the
creation of a state of strain, which, in order to be relaxed, favours the
92 8. Partial Dislocations

formation of new stacking faults moving in opposite directions (Fig. 8.4/3).

The Burgers vectors of the new partials include an angle of 120 with 0

that of the incoming ones. Hence, for example, two partials have to
move out in order to cancel the long range strain resulting from the
junction of the stacking faults. A mechanism of the type given in
Fig. 8.4/4 results from these assumptions, whereby the different steps
8 __ .--~.....~.....~...
A -+-_f
'----flJ----j1-\

8 _...-_~-~-~-\
A--I7~~~~~~~~~--'_~
C --.t--tI>----fl>----tl>----tli--\
8_~-,~~-;~~~ __'_~~~~

~
[111J [110J_
[00 1] [111] ~-+--lI

?Il ~-=-~:~=;~~
[11il ~-+--lI

Fig. 8.4/3. Geometrical situation at the intersection of two stacking faults

Q b c d

~ ~
Fig. 8.4/4. Phase transition mechanism by "reflection" at stacking faults
 8.4 . Transformation of the FCC- into the CPH Structure 93

are thought to take place on a statistical basis. Fig. 8.4/4a shows that
at slow rates the stacking fault (2) would come to rest in an equilibrium
position at a certain distance from the stacking fault (1). Only with the
help of additional kinetic energy of the partial dislocation could new
stacking faults be formed .
The essential points of such a mechanism are:
a) The basic configuration of converging stacking faults is frequent,
since dislocations present in the cubic phase would tend to split. Rem-
nants of former stacking faults can often be observed (Figs. 8.4/ 5,
8.4/6).

Fig. 8.4/ 5. Evidence tha t the final transformation in direction C took place
across stacking faults already present in directions A a nd B

b) The newly formed partial dislocation is essentially straight since

it had to be formed along the straight line junction of two stacking
faults, without having had to surmount a critical radius (analogous to
the" reflection at the surface", No.2).
c) The transformation does not have to be continuous as is the case
with the "pole mechanism". Many faults are possible and are also
frequently observed (Figs. 8.4/5 and 8.4/6).
d) On slow cooling, a framework of stacking faults can stabilize
the cubic phase (Fig. 8.4/7), especially if it is sufficiently narrow
to prevent new partial dislocations from building up enough kinetic
94 8. Partial Dislocations

Fig. 8.4/6. Probable remnants of former stacking faults

Fig. 8.4/ 7. Hexagonal grain (H) adjacent to a cubic grain (e) which is
stabilized by a framework of stacking faults

energy to trigger the mechanism. On the other hand, on rapid cooling

(quenching in liquid air), a more complete transformation is induced,
 8.4. Transformation of the FCC-into the CPR Structure 95

Fig. 8.4/8. Practically complete transformation induced by quenching in

liquid air. G boundaries of the original cubic grain

<
whereby domains in the four different 111 ) -orientations are produced
in the original cubic grains (Fig. 8.4/8).
5. A comparable mechanism was suggested by Hirth (1969), in
which it is assumed that an initially present stacking fault is cut by a
perfect dislocation, so producing a step in the stacking fault, i.e. a
partials dipole (Fig. 8.4/9). (A dislocation dipole is a pair of parallel
dislocations of opposite sign.) Under the effect of the stress field, new
dipoles would originate from the first and, as a result of the negative
stacking fault energy, they would be forced apart, thus creating new
layers of the hexagonal structure.
If, in the original assumption, a partial dislocation cuts the stacking
fault instead of the perfect dislocation, the dipoles would have to be
formed at a stacking fault. In this case the mechanism would be close
to the previous one, except that the transformation proceeds parallel
to the previously existing stacking fault, while in mechanism No . 4 the
newly formed stacking faults are parallel to the incoming stacking fault.
6. Gedwill et al. (1964) observed that cobalt whiskers (hair-like
single crystals) transform more readily in the longitudinal than in
the transverse direction. The advantage from the energetic point of
view becomes obvious if it is remembered that an oblique cut of a cylinder
96 8. Partial Dislocations

is an elongated ellipse with the smallest radius of curvature at both

extremities (Fig. 8.4/10). If the transformation starts at such an ex-

I
+1/6[1121 I
• 0-
I -116[1121
+0
I
•
I
b

j
+ •
(3)
0-
I
I (2)
+0
(1) f
+ • O-
(2)
+0
I
f (1)
.- c
+ .. O-
f (3)
+9

I
Fig. 8.4/9a-c. Dipole mechanism due to Hirth. a) Dislocation impinging on
a stracking fault, b) dipole of partials formed at the limits of the" dislocated"
stacking fault, c) formation of new partial dipoles. (The numbering sequence
reflects the time sequence)

Fig. 8.4/10. Transformation of cobalt whiskers (schematic representation)

 8.4. Transformation of the FCC-into the CPH Structure 97

tremity, then the shear stress produced by the line curvature (Eq. (6.5-3)]
and the increase in the extent of the stacking fault act together to
accelerate the transformation. In contrast to case 1, the transformed part
here lies outside the curved partial, and hence no critical radius has to
be surmounted. The shear stress produced by the line curvature and the
transformed surface per loop (i.e. the decrease in energy) are both larger
in the case of the longitudinal transformation than for the transverse
transformation. These considerations also hold when the shape of the
whisker is prismatic instead of cylindrical. Thus, it is to be assumed that
this kind of transformation nucleates at the surface.

7 Bolhnann, Crystal Defects

 9. Dualistic Representation
of Dislocation Reactions

9.1. Introduction
In Section 7.4 it was shown that reactions take place between dis-
locations, i.e. they form nodes and coalesce if the energy of the con-
figuration is thus lowered. At the nodes, Frank's node condition for the
Burgers vectors (continuity of Burgers vectors) has to be fulfilled
[Eq. (7.4-1)]:
(9.1-1)

The Burgers vector is attributed to a dislocation but it is not localized

on it. Here we choose a representation in which the Burgers vector is
completely separated from the dislocation and placed into a separate
space (Bollmann, 1962, 1964). The space containing the dislocation
lines-the real crystal-is termed the L-space, and that of the Burgers
vectors the b-space. While the L-space contains the dislocation lines
together with their nodes, the b-space contains those Burgers vector
configurations which arise from the fulfillment of the node condition.
It will be shown that a strict duality exists between the two configura-
tions. The situation is comparable to the force polygons of structural
frameworks-the so called Cremona plans.
As a first example we discuss the intersection of two right-handed
screw dislocations in the bcc lattice with the Burgers vectors (1/2) [111]
and (1/2) [Hi] (Fig. 9.1/1 a). In Fig. 9.1/1 b the dislocation segments are
numbered and shown together with their Burgers vectors. In Fig. 9.1/1 c,
the line sense and the Burgers vectors of the segments 2 and 3 are
inverted. Frank's node condition is in this case:

(9.1-2)

The Burgers vectors are now transferred to the b-space and added: since
their sum is zero they must form a closed parallelogram configuration
(Fig. 9.1/1 d). We shall now formulate a series of duality relations which
will allow configurations in the L-space to be "translated" into those
in the b-space and "vice versa".
 9.1. Introduction 99

The first relation, which we term D 1, is as follows:

L b
D1 An L-node corresponds to a closed b-polygon.

This statement is obviously valid for arbitrary dislocations.

The coordinate system of the b-space is oriented in the same manner
as that of the L-space, but for reasons of presentation, the units are
usually chosen differently.
Instead of numbering the different dislocation segments we can also
attribute numbers to the L-fields bounded by dislocations. We already
introduce here a two-coordinate cipher system which will later be useful

L b
(2) (1)
001

(3) (0) 'lz[1ii)

b a

(2) (1)

~ #

~.
D01t ~
00'1
c
~110
(0) (2) (3)

110
d
(2) (1)

(0)
f
11
~
(2) 1 (3)
00

10

g h
"V" 01

Fig. 9.1/1. Introduction to the dualities between the configurations of the

dislocation lines and the Burgers vectors
100 9. Dualistic Representation of Dislocation Reactions

for dislocation networks. The corresponding ciphers in the b-configuration

are found at the b-nodes (Fig. 9.1/1e, f). The number of Burgers vectors
connected by a b-node equals the number of dislocation segments at-
tributed to the corresponding L-field. Thus, D 2 becomes:
D2 To every L-field corresponds a b-node.
Also, D3 is evident:
A dislocation segment The Burgers vector of that
separates two L-fields. segment connects the two
corresponding b-nodes.

Nothing has been said until now about the sense of the dislocation line
and of the Burgers vector. If a dislocation segment bounds an L-field,
its sense determines a sense of rotation around that field, which may be
clockwise or counter-clockwise. In an analogous manner the Burgers
vector may "flow" out of a node or into it, i.e. it can be fixed at the
node either by its tail or by its tip. The relation between the sense of
rotation and the flowing in or out of a node is remniscent of the relation
between curl and divergency in electrodynamics. In order to settle a
convention, we assume the principal sense of rotation around an L-field
to be counter-clockwise. Thus, we see from Fig. 9.1/1.g, h that, if the
dislocation segment follows this principal sense of rotation, the corre-
sponding Burgers vector flows out of the corresponding node. Therefore
we can formulate D 4 as:

D4 If a dislocation segment then the corresponding

flows in the principal sense Burgers vector flows out
(counter-clockwise) around of the corresponding
an L-field. b-node.
and vice versa (see also Section 11.2).

An analogous relation exists between the sense of rotation around a

b-polygon and the flow in or out of an L-node

D4' If a dislocation flows out of then the corresponding

an L-node. Burgers vector points
clockwise around the
corresponding b-polygon.

It is to be noted that, in the L-space, only the sense of the dislocation

line and not the orientation of the Burgers vector is of importance. On
the other hand, in the b-space, only the action of "flowing in or out"
 9.2. The Significance of the b-Net 101

counts and again not the orientation of the Burgers vector. A Burgers
vector drawn in an L-configuration serves only as an aid for visualizing
the edge- or screw character. When the L- and the b-configuration are
drawn, the line sense and the sign of the Burgers vector can be neglected.
They can always be reconstructed from the diagrams and from the
numbering scheme.

9.2. The Significance of the b-Net

According to the movement rule CEq. (5.4-1)J, the Burgers vector
can be interpreted as a displacement of that part of the crystal marked
by m if the dislocation moves in the direction v. Let us assume that, at
the beginning, the crystal is free of dislocations. The field 00 is then
extended over the whole area. The dislocations are now allowed to enter
the crystal one after the other and to form the dislocation configuration
step by step. In each freshly created field, one part of the crystal is
shifted with respect to the other. D 4 was chosen such that, with the
FS/RH sign convention, the b-configuration represents a map of the
relative displacements of that part above the corresponding L-field. The
part of the crystal below the L-configuration is considered as being fixed.
Hence, the vector m of the movement rule has always to point upwards;
this is the case if I points in the direction of principal sense. In this way,
b flows out of the node 00 and into the neighbouring node (e.g. 00 --? 10,
Fig. 9.1/1g, h). However, if I points in the opposite direction to the
principal sense, then m points downwards. In order that the b-con-
figuration describes the displacements of the upper part of the crystal,
either I and b have to be inverted or, in other words, b has to be con-
nected from its tip to the b-node 00, i.e. it has to flow into the node 00.
Fig.9.2/1 presents schematically the displacement of the upper part
with respect to the lower part the latter being kept fixed.

10

-11~OO
01

Fig. 9.2/1. Representation of the b-net as a map of the relative displacements

102 9. Dualistic Representation of Dislocation Reactions

9.3. The Prediction of Dislocation Reactions

Based on the b2-criterion we know from Section 7.4 that a dislocation
reaction might take place if the Burgers vectors form an acute-angled
triangle. We again consider our example from Section 9.1, with the
two screw dislocations {1/2} [111J and {1/2} [n 1J, where the b-polygon
forms a rhombus {Fig. 9.1/1 d}. We divide this rhombus into two acute-
angled triangles, and translate the configuration to the L-side. The
b-nodes 1 0 and 01 are then directly connected, which according to D 3,
means that a direct separation {i.e. a dislocation} exists between the
correspondingly numbered L-fields. This b-connection is the Burgers
vector of the new dislocation line. The signs are attributed by D 4. The
two b-nodes {10 and 01} now join three Burgers vectors, which ac-
cording to D1, D2 and D3, means that the corresponding L-fields are
bounded by three dislocations {Fig. 9.3/1}. The new dislocation is that
reaction product which is energetically favored.
{1/2} [111J + {1/2} [n 1] = [001J
+ 3a 2
{9·3-1}
b2 .• 3a2
2 2
2
>a.
It is also oriented as a right-handed screw dislocation.

L b

110
10

• 00
01

I
10

.. 11<1>00
01

Fig. 9.3/1. Stable dislocation reaction

 9.3. The Prediction of Dislocation Reactions 103

Hence, we can predict possible dislocation reactions by subdividing

the b-configuration into acute-angled triangles and translating the new
b-configuration to the L-side. On the L-side, naturally, only the topo-
logical connections, and not the exact geometrical L-configurations are
obtained.
The subdivision of a b-configuration into obtuse-angled triangles
would predict energy consuming reactions. The acute-angled or obtuse-
angled triangle criteria are, strictly speaking, only meaningful if all the
dislocations are of the same character. Fig. 9.3/2 presents an example in

01Ql0
00

01
h
11

00

.. 01<:])10

11
Fig. 9.3/2. Indifferent dislocation reaction

which edge and screw orientations are present. As the energy of an edge
dislocation is 50% higher than that of the screw dislocation, it may
happen in this case that the reactions do not occur, although the b-net
has been subdivided into acute-angled triangles. It is to be noted that,
in this case, a right-handed and a left-handed screw dislocation intersect.
According to the numbering system in the L- and the b-configurations,
it can be seen that the rotation sense is inverted in the two plots.
The energy of a dislocation is not only given by the Burgers vector
and the edge and screw character, but also by its length. Dislocation
reactions may also occur if energetically unfavorable Burgers vectors are
produced, if the total energy can be lowered. In the bcc structure (iron)
small segments of <110 )-dislocations have been observed as reaction
<
products of two 1 00 )-dislocations (Hale and Brown, 1968).
104 9. Dualistic Representation of Dislocation Reactions

In order to study a situation such as that given, for example, in

Fig. 9.3/3 a, the line sense of every dislocation segment is first fixed.
It is advantageous to draw the Burgers vector corresponding to every
line segment. Then the L-fields are numbered according to a two-co-

l b

13 12 11 10

-LIll 03 02 01 00

a b

13 12 11 10

/5lS5!
03 02 01 00

d c

Fig. 9.3/3. Zig-zag reaction

ordinate system. Both, the dislocation lines and the Burgers vectors are
referred to the same coordinate system. Now, on the b-side, a starting
point (e.g. the point 00) is chosen, and according to D4, we move round
the field 00. The corresponding Burgers vectors are accordingly joined,
either by their tails or by their tips to the b-node 00. The other ends of
the Burgers vectors are then numbered according to D 3; every end
receiving the number of the corresponding neighbouring field. After-
wards the construction continues in an analogous manner by moving
round the neighbour fields one after the other, drawing the Burgers
vectors and again numbering the ends, until every L-field is attributed
to a b-node and the Burgers vector of every dislocation segment has
found its place in the b-configuration. When all the fields have been
dealt with in this way the whole b-configuration is constructed. Different
projections may be required in complicated cases.
 9.3. The Prediction of Dislocation Reactions t05

Once the b-configuration is drawn, it is subdivided, as far as possible,

into acute-angled triangles and the new configuration is translated to the
L-side. The number of Burgers vectors meeting in a b-node is equal to
the number of dislocation segments bounding the corresponding L-field.
Fig. 9.3/3 shows the so called zig-zag reaction.

l b

-
12
13

.. 13 12 11 10
~
03 02 01 00

a b

~1J •
02/ 12
~_ _.;.;.11_ _:
00 I 10
c

Fig. 9.3/4. Intersection of dislocations containing identical Burgers vectors

The example in Fig. 9.3/4 is a case where dislocations of different

orientations, but with the same Burgers vector, intersect. The situation
can be obtained from the foregoing example Fig. 9.3/3 by slowly turning
the Burgers vector of the inclined dislocation line against the horizontal
orientation. In this manner, certain b-nodes are merged so that now
two numbers are attributed to the same b-node. As every b-node is, by
definition (D2) attributed to a single L-field, the fact that a b-node is
marked by two numbers means that two fields which were originally
drawn separately become joined to one field, i.e. there is no displacement
between these fields. We may formulate this as D 5:

D5 Two L-fields' originally drawn ... if the numbers attributed

separately, are joined to one to the L-fields meet at a
single field ... single b-node.
106 9. Dualistic Representation of Dislocation Reactions

The third example in Fig. 9.3/5 represents an unstable situation. This

configuration can be stabilised in two ways by the formation of direct
connections in the b-net; this corresponds to dislocation reactions in the
L-net. An extension of D 2 is evident from this example:

D6 A series of parallel corresponds to a chain of Burgers

dislocation lines vectors connected in
succession.

L b
--
05
.QL
•
.. 02 01 00
03
..
~
\

-02\
..
-(00')
05 04 03
01
..

05
04
..
02 01 00
I •
-
•
t2
01
.. ~
05 04
•
00

--
05
04
•
01 00
O} • •
~
/01 05 04 03
•
00

Fig. 9.3/5. Stabilization of an unstable dislocation arrangement

Since the dislocations are not joined at nodes, the b-polygon chain is not
closed.
The procedures given in this section can also be extended to partial
dislocations, as is done in the following section.
 9.4. Dislocation Reactions Involving Partial Dislocations t 07

9.4. Dislocation Reactions Involving Partial Dislocations

In Section 8.3 we showed that, if a dislocation in the face-centred
cubic structure (ABC-stacking) is split into partials, the sequence of the
partials is fixed. In the hexagonal structure (the ABAB-stacking) both
sequences can appear, according to whether the dislocation is located
above an A- or a B-layer. Dislocation reactions can also take place if the
dislocations are not on the same glide plane. In this case the dislocations

116m2]
1f6[121l~ 116[211]
22

b 02

B B
..
c c
c B
d

e
Fig. 9.4/t. Reaction of partial dislocations (triangular node)
108 9. Dualistic Representation of Dislocation Reactions

align themselves parallel to one another instead of merging, and dis-

locations which would cancel each other constitute dislocation dipoles, i.e.
a pair of parallel dislocations of opposite sign.
Here we investigate the crossing of two split dislocations. The first
case (Fig. 9.4/1) can arise in the cubic as well as in the hexagonal structure.
The dislocations may lie-but not necessarily so-in the same glide
plane. When the perfect Burgers vectors and the glide plane are known,

21

22 00
11

12
b

I
a

B B

c
c
d

e
Fig. 9.4/2. Reaction of partials in the hexagonal structure
 9.5. Thompson's Notation of the Burgers Vectors 109

the sequence of partials has first to be determined according to Section 8.3.

The Burgers vectors and the line sense of the partials are given for every
dislocation line segment. Now, in order to construct the b-net, we number,
and move round, the L-fields in the manner already described (Section 9.3).
The b-net is a map of the relative displacements of the part of the crystal
above the plane of the paper with respect to the lower part which is
assumed to be fixed. Let us assume the first layer above the plane of
the paper to be a B-Iayer and the structure below the plane of the paper
to consist of a CA-stacking. As the upper layer can only sit in positions B
and C (while A is forbidden since it is labile), the fields: 00, 20, 02, 22
have to remain in the B-position, and the fields: 10, 01, 12, 21 in the
C-position i.e. the stacking fault position. It is seen in the b-net that
the field 11 would remain in the A-position which is forbidden. Hence,
this field has to contract to a point. Furthermore, the fields 21 and 1 0
have to join or to remain separated by a dislocation dipole, as they
correspond to a common b-node (D 5). Therefore, after retranslation,
we obtain the L-configuration Fig. 9.4/1 e.
The second case (Fig. 9.4/2) can only occur in the hexagonal structure
and then only if the two dislocations are in different positions (B above
A and A above B). The configuration of this example is distinguished
from that of Fig. 9.4/1 by the fact that in one of the dislocations the
sequence of the partials is inverted. The b-net is again constructed in
the usual manner. The fields: 00, 20, 02, 22 again contain no stacking
fault, the fields 21 and 01 involve .. C over A" stacking faults and the
fields 1 0 and 12 .. Cover B" stacking faults. The fields 21 and 11 have
to be joined. We subdivide the b-net into acute-angled triangles and
retranslate the configuration. The partials 12/22 and 22/21 strongly
repel one another, since the corresponding b-triangle represents the
extreme case of an obtuse-angled triangle (180°). (See e.g. Hirth and
Lothe, 1968, p. 327.)
Further reactions involving partial dislocations will be treated in
Section 11.4.

9.5. Thompson's Notation of the Burgers Vectors

Thompson (1953) introduced the following notation in order to
attribute a Burgers vector to every dislocation: As was shown in Sec-
tion 8.3, the comers of Thompson's tetrahedron are denoted by the
capitals A, B, C, D, and the centres of the opposite faces by the Greek
letters IX, {J, y, <5 (Fig. 8.3/4). Two letters are attributed to every dis-
location such that, on looking directly along the line (which fixes the
line sense), the sequence of the letters is read, in the usual way, from left
to right; for example, the letters AC denote the Burgers vector which,
110 9. Dualistic Representation of Dislocation Reactions

in the tetrahedron, points from A to C (Fig. 9.5/1). If the dislocation is

viewed from the opposite side, the line sense, as well as the sequence of
the letters, is inverted, which takes into account the pseudo-vector
character of the Burgers vector. The perfect dislocations are denoted by

AC

Fig. 9.5/1. Thompson's notation for a single dislocation

Roman letters, the partials by mixed Roman and Greek. A special

"letter-algebra" was developed for the prediction of dislocation reactions,
and which is widely used in the literature (e.g. Hirth and Lothe, 1968,
p.301, Amelinckx and Dekeyser, 1959). In Thompson's notation, the
example Fig. 9.3/1 would correspond to Fig. 9.5/2.

[>0
C

Fig. 9.5/2. Thompson's notation for two dislocation nodes

Thompson's notation is translated into the dualistic representation

by replacing the letters in the dislocation segments by the Burgers
vectors themselves, and by constructing the b-net according to Sec-
tions 9.3 and 9.4. If it is felt necessary, the letters can be reintroduced at
the end. We shall restrict ourselves to the dualistic representation
because, as a plan of the displacements between the L-fields, the b-net
gives a clearer view of the situation than the letters in the L-net.

Chapters 7-9. Summary and Discussion

In Chapters 7 to 9, we started from the interaction between two
parallel dislocations where we saw that dislocations of the same sign
repel while those of opposite sign attract one another (Section 7.2). If
 Chapters 7-9. Summary and Discussion 111

only glide is allowed for parallel edge dislocations, then certain stable ar-
rangements exist (Section 7.3).
Dislocations which attract one another can also merge partially or
completely and form branches, i.e. they can react among themselves.
Here, the important law of continuity of the Burgers vectors, i.e. Frank's
node condition, has to be fulfilled (7.4). Besides this condition, the b2_
criterion (which, however, is only approximately valid) helps to decide
whether a dislocation reaction might take place or not. Quantitative
examples are given in Section 7.5.
Partial dislocations, representing partial displacements of a perfect
dislocation, are treated in Section 8.2. When their configurations are
considered they are seen to be limits of stacking faults. According to the
b2-criterion partial dislocations of a perfect dislocation repel each other,
but they are held together by a strip of stacking fault, the width of
which is determined by the stacking fault energy (surface tension).
The sequence of the partials in the fcc structure is not arbitrary, i.e.
the addition of the partial Burgers vectors not commutative. The sequence
can be determined geometrically by means of Thompson's tetrahedron
and the movement rule [Eq. (5.4-i)], or algebraically by means of
Eq. (8.3-13).
As an example of partial dislocations and stacking faults the phase
transition of cobalt from the fcc- to the cph-structure is discussed in
Section 8.4.
For the study of more complex dislocation reactions, the dualistic
representation was developed, whereby the Burgers vectors and the
dislocation lines are completely separated. Frank had already separated
the Burgers vector from its dislocation by representing it in the reference
lattice and not in the crystal. In principle, this idea is further developed
here. On interpreting the node condition on a geometrical basis, a strict
duality between the configurations of the dislocation lines and those of
the Burgers vectors appears (Section 9.1). The b-net represents a plan
of all the displacements on one side of the dislocation network with
respect to the other side which is considered to be fixed (Section 9.2).
The possibilities for dislocation reactions thus become directly visible
(Section 9.3). Reactions of partial dislocations can also be treated in
this manner (Section 9.4).
In order to interpret a given dislocation arrangement which has been
observed by electron microscopy, for example, the following practice
may help: A transparent paper is put on top of the photo. Then, in
every field bounded by dislocations a representative point, if possible
the central point of the field, is marked. Now all the neighbouring points
are connected across the dislocation lines. Afterwards, the configuration
is taken out of the dislocation network and rotated by 90°. It re-
112 9. Dualistic Representation of Dislocation Reactions

presents topologically, in most cases, the structure of the b-net although

the sign is not determined. This procedure can be checked on the con-
figurations of Sections 9.3 and 9.4 and the justification is given in the
summary of chapter 11.
Suggestions for Problems
Many exercises can be found in the book "Elementary Dislocation
Theory" by Johannes and Julia R. Weertman (1964) in which the
subject of interactions between dislocations is more stressed than in the
present book. Mathematically more sophisticated problems are treated
in the book of Hirth and Lothe (1968).
The bll-criterion can be more exactly formulated for dislocation
reactions between dislocations of different character, whereby the limit-
ingangle can deviate from 90°. In addition, other dislocation reactions
such as are indicated in Section 7.5, can be discussed.
Data about the different possibilities of partials in the fcc and the
bcc structures is to be found in the book of Weertman and Weertman
(1964).
Further exercises on the dualistic representation of dislocation
reactions can be set up by the combination of problems, as have been
presented in Sections 9.3 and 9.4. It is, however, advantageous to
relate these problems to dislocation networks (Chapter 11).
 10. Short Description of the Relations
between Dislocations and Point Defects,
and Dislocation Dynamics
10.1. Dislocations and Point Defects
This section shall present a short description of notions which are
important in dislocation theory, but which are not applied in this
monograph. The details are found in the literature (e.g. Nabarro, 1967;
Hirth and Lothe, 1968).
a} Dislocation loops can be formed by the condensation of vacancies
or interstitials between densly packed layers. A supersaturation of
vacancies may be produced in a material if it is quenched from a high
temperature, at which thermal equilibrium existed. On slow heating,
the vacancies become mobile and condense in disks or disappear at
dislocations or grain boundaries where the edge components of the
dislocations climb (Fig. 10.1/1). In a material with high stacking fault
energy (e.g. aluminium) the condensed vacancies form loops of perfect
dislocations while in a material with a low stacking fault energy (e.g.
gold) small stacking fault tetrahedra are formed. Real holes, resulting
from condensed vacancies, have been observed in AI-foils by Yoshida
et al. (1963).
b} A distribution of point defects, not in the thermal equilibrium, can
be produced by irradiation (neutrons, protons, electrons, fission products,
y-rays, etc.). Here also the point defects collect in groups upon heating,
and the disturbed crystal lattice is thereby partially repaired. The dif-
ferent stages of recovery are strongly temperature-dependent. Problems
of radiation damage are discussed in the book of Damask and Dienes
(1963), for example.
c} When vacancies or interstitials condense on screw dislocations,
helical dislocation lines may be formed.
d} Two notions which often appear in dislocation theory, especially in
dislocation dynamics, are the "kink" and the "jog".
If a dislocation lies between two close-packed layers, then its energy
will depend on the orientation (Fig. 3.2/1), whether it lies along rows
of spheres (a configuration which is repeated every 60° in the (110)-
orientations} or in an arbitrary orientation. As a model, the glide plane
can be -represented as a row of parallel potential hills and valleys. A dis-
8 Bollmann, Crystal Defects
114 10. Dislocations and Point Defects, Dislocation Dynamics

Fig. 10.1 / 1. Loops of condensed vacancies in aluminium . The irregular shape

of the dislocation is due to climb of the edge components induced by
condensed vacancies. (Photo: J. Silcox)

location will tend to align itself along the valleys. If it changes its path
across a potential hill it will be bent (Fig. 10.1/2), and such local bending

E
#//////////////1///////

1///////;;

1/'////11//11//////1//////
y
Fig. 10.1 / 2. Representation of a "kink"
 10.1. Dislocations and Point Defects 115

is termed a "kink". When a dislocation moves in its glide plane, the

kinks move along the line. This kink movement produces a friction
force-the so called "Peierls-Nabarro force"-which impedes disloca-
tion movement and contributes to the resistance to plastic deformation.
The other notion is the "j og ". This is another case of a local bend
in the dislocation line but this time with a component perpendicular to
the glide plane (Fig. 10.1/3). A jog can be understood, for example, as
a step in the extra plane of an edge dislocation. If, during the climb of
an edge dislocation, the extra plane grows or diminishes, jogs move along
the dislocation.

Fig. 10.1/3. Example of a "jog"

e) Kinks and jogs can be formed when two moving dislocations

intersect. We investigate two right-handed screw dislocations which are
normal to one another (Fig. 10.1/4). After the dislocations have crossed,
a segment connecting the two lines is formed, in which the material is
compressed. In this manner a chain of interstitials is formed, and energy
is consumed. Thus, after intersection, the dislocations can no longer
move freely. Each dislocation now has a jog which cannot glide, since
the movement would involve a component perpendicular to the line
segment of the jog as well as one perpendicular to the Burgers vector.
If Fig. 10.1/4 is reconstructed for other examples, it is seen that:
1. The intersection of a right-handed and a left-handed screw dis-
location produces vacancies, and
2. in the case of one or both dislocations being of pure edge character,
a kink is formed without volume change, so that both dislocations can
continue their paths relatively unimpeded.
If two dislocations intersect, dislocation No.1 obtains a jog (or a
kink) of the amount b 2 , and dislocation No.2 one of b 1 • r shall be the
direction of the movement of dislocation 1, where dislocation 2 is con-
sidered as fixed. Then, according to Homstra (1962) (and Hirth and
Lothe, 1968, p. 392), the sign of the displacement of the jog with respect
S*
116 10. Dislocations and Point Defects, Dislocation Dynamics

11

Fig. 10.1/4. Formation of

interstitials by the inter-
section of two screw dislo-
cations
b

to '1 is:
(10.1-1 )
A jog in a screw dislocation acts as a local edge component with the
corresponding extra plane. If the dislocation moves together with the
jog, the extra plane of the jog either grows or diminishes, and vacancies
or interstitials are thus produced. r' shall be simultaneously the direction
of movement of the jog in dislocation 1 and the total amount of the
displacement with respect to a coordinate system fixed in the crystal.
The movement of the jog can have two components, one perpendicular
to dislocation 1 (the same as r) as though it were rigidly fixed to the
 10.2. Dislocation Dynamics 117

dislocation, and the other along the dislocation. The number of point
defects, N, produced on the path r' is, again according to Homstra
(1962) :
bl [r' X b 2 J
N =sgn (b 2 ) --v-- (10.1-2)

where sgn (b 2 ) follows from Eq. (10.1-1).

sgn (N) = ( +) interstitials,
= ( -) vacancies,
V = atomic volume.
If one or both dislocations are of pure edge character, N = o.
10.2. Dislocation Dynamics
A large part of dislocation theory deals with dislocation dynamics.
The papers of a symposium held on this subject are collected in the book
"Dislocation Dynamics", edited by A. R. Rosenfield et al. (1968).
Dislocation dynamics treats the dynamical behaviour of dislocations
during plastic deformation. One aim of studies in this field is the under-
standing of stress-strain curves. The relation between stress and strain
depends on the velocity of deformation, on the temperature and on the
history of the specimen. Even if it is not yet possible to give an exact
quantitative explanation of these very complex phenomena, the study
of dislocation dynamics contributes substantially to their understanding.
 11. Dislocation Networks-Subgrain Boundaries
11.1. Introduction
In this chapter we shall investigate the case of two joined crystals of
the same kind, but slightly rotated with respect to one another. In this
context, the word "slightly" means that the two crystals are practically
in register over areas which are large compared to the unit cell, and that
only occasional corrections are needed in the form of dislocations. Such
a boundary, consisting of a network of discrete dislocations, is termed a
"subgrain boundary".

(2)
.
III! III HI
p

(I)

(a) (b)

Fig. 11.1/1. Fitting of two crystals. a) Perfect crystal divided into two parts,
b) both parts separated and crystal 2 rotated

-~
ilrHrrHl
~b(C)
~(I)
Cc) (d)

Fig. 11.1/1. c) Geometrical fit, d) physical fit

 11.2. Frank's Formula 119

We visualize the formation of a subgrain boundary in the following

manner: First we assume a perfect crystal divided into two parts: 1 und 2
(Fig. 11.1/1). In both parts, we assume a free vector;r, i.e. ;r(1) and ;r(2).
In the perfect crystal both vectors shall be identical.
;r(2) - ;r(1) = 0. (11.1-1)

Instead of a single pair of vectors, we may imagine two vector fields in

which every pair of related vectors fulfills condition (11.1-1).
Now we separate both parts and rotate crystal 2 by the (small)
angle e around the polar vector 9 (191 = e,
orientation: right handed
screw), leaving crystal 1 fixed. Along with crystal 2, the whole vector
field is rotated, so that now the related vectors are no longer identical, i.e.
(11.1-2)
where, in the difference vector, b(C), the superscript (C) stands for
" continuous". We shall see that b(C) is related to the Burgers vectors
and that it is a function of the polar vector 9 and of the vector ;r(1) (and
respectively ;r(2)). We now cut the two crystals, while preserving the
orientations, such that they can be placed on top of one another.
As long as we do not assume that atomic forces act between the two
faces, the two superposed crystals form a moire pattern at the interface
(Moire-Models Nos. 1-3). The locations where the crystal fit is perfect
appear as bright areas, while those where there is misfit appear as dark
areas. Such a situation is termed" geometrical fit".
If atomic forces act between the two crystals, then the bright regions
of perfect fit will extend, and the dark bands, corresponding to misfit,
will contract to dislocation lines, thus forming a dislocation network at
the interface. This situation is called" physical fit".
In the next section the function b(C) (9, ;r(1)) is investigated.

11.2. Frank's Formula

The function b(C) (9, ;r(1)) as indicated by F. c. Frank (1950) can be
written as:
(11.2-1)
with
(11.2-2)
e
Because of the small angle we do not distinguish between ;r(1) and ;r(2).
In the form (11.2-1), Frank's formula is a purely geometrical statement.
If ;r...L 9, then 1b(C)1 = l;rl j(). The important point, however, is the meaning
which Frank attributes to b(C). b(C) is approximately equal to the sum
of the Burgers vectors of the dislocations crossed by ;r if ;r is chosen in
120 11. Dislocation N etworks-Subgrain Boundaries

the boundary. We may illustrate this, for example, with the aid of the
moire model NO.1 (Fig. 11.2/1}. The relation is only approximate
because the Burgers vectors of dislocations can only acquire discrete
values, while x and consequently b(C) are continuously variable vectors.

Fig. 11.2/ 1. Demonstration of Frank's formula (b(C) """ n . b(D), n number of

intersected dislocations}
 11.2. Frank's Formula 121

According to Eq. (11.2-1), b(C) is perpendicular to ~ as well as to ().

Frank's formula is an approximation. It is mathematically exact only
for () = 0, but in this case it loses its meaning. Also, it is obvious that
for larger angles (), b(C) is always perpendicular to (), but then ~(l) and
~(2) can no longer be identified. The generalisation of Frank's formula
will be discussed in Section 12.3. For the time being, the theory of sub-
grain boundaries will be developed on the basis of the approximation
[Eq. (11.2-1)].
Referring to what was mentioned above, we may write Frank's
formula as:
(11.2-3)
where b(D) is equal to the Burgers vectors of those discrete dislocations
which are intersected by ~.
The Burgers vectors are translation vectors of the crystal lattice (or
better: coordinate differences). As Frank's formula does not distinguish
between the two lattices, the notion of the Burgers vector remains, in
this context, somewhat indeterminate. It will be further clarified in
Section 14.2.
We may interpret Eq. (11.2-1) as an imaging relation between the
crystal space (x-space) and a b-space, where the correspondance between
the two spaces is continuous.
Let us discuss this imaging relation more closely. We place () in an
orthogonal coordinate system along the positive z-axis. Then Frank's
formula, written in coordinate terms, is:
bl = -{)x2
b2 = {)xl (11.2-4)
b3 == 0.
°
b3 == signifies that only those points of the b-space which lie in the
(bl, b2)-plane have an image in the x-space. This image is determined
exclusively by Xl and x 2 and is independent of x 3 , which means that
every b-point in the (bI, b2)-plane has, as its image, a straight line parallel
to the x3-axis, i.e. parallel to ().
According to Eq. (11.2-3) there exist points in the (bI, b2)-plane for
which:
(11.2-5)
To everyone of these discrete points corresponds an individual straight
line parallel to the axis of rotation in the x-space. In this manner,
Frank's formula becomes "quantized", i.e. it is defined only for discrete
b-points.
The x-lines can be separated from one another by cell walls. The cell
walls shall be of the Wigner-Seitz type (Section 2.2), i.e. planes per-
122 11. Dislocation Networks-Subgrain Boundaries

pendicularly bisecting the connection between two x-lines. In this way

a honeycomb type of cell structure is introduced into the crystal space.
We can interpret this construction as follows: Instead of being cut
and joined, the two crystals can be idealized as two interpenetrating
point lattices without space filling atoms. The interpenetrating lattices
are understood as being two sets of positions for one set of atoms. The
atoms can sit in either lattice and so "realize" the corresponding crystal.
We may now choose a boundary passing through the cell structure,
and then place the atoms on one side of the boundary into lattice 1,
and on the other into lattice 2. The intersection of the boundary with
the cell walls becomes the dislocation network. The net, which is
formed by the interconnections of those discrete b-points in the (bI, b2 )_
plane, the images of which (the x-lines) are intersected by the boundary,
becomes the b-net. The relations between the two configurations are
given by the dualities in Section 9.2. As soon as the boundary is fixed,
the rest of the cell structure loses its meaning. Outside the boundary the
space is now occupied by the undisturbed crystals 1 and 2. The inter-
sections of the x-lines with the boundary are the points where the two
crystals exhibit optimum matching. They are the centres of the fields
bounded by dislocations (bright areas in the moire model) and are

Fig. 11.2/2. Honeycomb structure representing the geometrical interpretation

of Frank's formula (0 = MS-points)
 11.2. Frank's Formula 123

termed "minimum strain points", or "MS-points". The cell walls

correspond to the locations of worst fit, i.e. the dark bands in the moire
models or the dislocations in the crystal (Fig. 11.2/2).
The b-net formed by all the Burgers vectors is a net of translation
vectors of the crystal, which means it is a section through the translatiot).
lattice of the crystal. Hence, a natural coordinate system of the b-net,
i.e. of the b-nodes, is given by the coordinate system of the crystal. In
this manner, the corresponding "numbering" of the L-fields is also
defined. In the configurations given in Section 9.1 the coordinates of
the L-fields were first arbitrarily fixed, and those of the b-nodes follow.
While in earlier sections the L- and the b-configurations were related
exclusively by geometrical-topological considerations, Frank's formula
gives an algebraic relation between the two configurations of a subgrain
boundary. For a given rotation (), the cell structure is a three-dimensional
representation of all the possible two-dimensional subgrain boundaries.
Naturally, the cell structure is independent of the chosen coordinate
system. The honeycomb-like cells are always parallel to the axis of
rotation and the plane of all the b-vectors is perpendicular to that axis.
With respect to the application of the duality D4 (Section 9.1) a
certain ambiguity remains in the attribution of the sign of the Burgers
vector in relation to the dislocation line sense. If we define lattice 1 as
the basis from which the rotation () starts and lattice 2 as the end product
of the rotation, and if we now choose a boundary within the honeycomb
structure, crystal 1 can be located on either side of the boundary. How-
ever, the sign of the dislocations in the boundary depends on the location
of crystal 1. If crystal 1 and 2 exchange places, the dislocations change
their sign (e.g. from a right-handed to a left-handed screw).
In order to take this fact into account the duality D4 can be re-
formulated in terms of the duality between curl and divergency, where
the curl-vector is a polar vector (right-handed screw) coupled to the
sense of rotation around the L-field (line sense), and the divergency is
positive if the Burgers vectors "flow" out of the node (and vice versa).
With these notions D4 becomes:

D4* If the curl vector of an L-field the divergency of the

points into crystal 2 corresponding b-node is
positive.

Hence, if crystal 1 and 2 exchange sides the sign of the curl vector
switches, i.e. the line sense alters while the Burgers vector remains
unchanged which means that the dislocations switch their sign too.
In the next section a classification of subgrain boundaries is developed,
based on this cell structure.
124 11. Dislocation Networks-Subgrain Boundaries

11.3. Classification of Sub grain Boundaries

A subgrain boundary is defined by:
a) the orientation of the rotation vector (J,
b) the value of the rotation angle () and
c) the position of the boundary within the crystal (for the time being
we assume a planar boundary).
a) We distinguish between a "special" and a "general" orientation
of (J within the crystal. As "special" we understand an axis of rotation
which is (in crystal 1) perpendicular to two linearly independent Burgers
vectors (minimal translation vectors). In this case the b-plane is densely
occupied by b-Iattice points which thus form a planar b-net. Such
special orientations in the fcc structure are the (111)- and the (100)-
and in the bcc case the <11 0)- and the <100 )-orientations. In the
" general" orientations of (J it is no longer possible to construct a planar
b-net in the b-plane which is defined by Frank's formula. This problem
will be discussed in Section 11.6.
b) The value of () determines the size of the cells; the larger (), the
smaller are the cells.
c) The orientation of the boundary plane with respect to the axis of
rotation may be:
1. perpendicular,
2. in arbitrary orientation,
3. parallel.
Accordingly, the boundary is called:
1. a pure twist boundary,
2. a partial twist boundary,
3. a tilt boundary.
The terms are summarized in the following table:

Boundary Orientation of ()
special general
1- () special pure twist general pure
boundary twist boundary
Arbitrary with respect to () special partial twist general partial
boundary twist boundary
II () * special tilt boundary general tilt
boundary
* For the tilt boundary finer distinctions are needed, which will be given
in Section 11.4.
 11.4. "Special" Subgrain Boundaries 125

11.4. "Special" Subgrain Boundaries

a) Special Pure Twist Boundary
In a special pure twist boundary the b-net is planar, and the b-net
and the boundary plane are parallel. From Eqs. (11.2-4) and (11.2-5),
follows that the b-net and the lattice of the MS-points are related by a

Cf 20 40 60 eo .~
e ,

Fig. 11.4/ 1. a) Special pure twist boundary on the {111 }-plane in the f.c.c.
structure; duality between the L- and the b-net. b) Hexagonal dislocation
network in iron. (Photo: A. Mascanzoni)
126 11. Dislocation Networks- Subgrain Boundaries

transformation in which the lattice of the MS-points is rotated by 90°

with respect to the b-net. From this fact follows that the L-net consists
of screw dislocations. Fig. 11.4/1 shows the duality relations between the
b-net and the L-net for a hexagonal dislocation network in the {111}-
plane of the fcc structure (or the basal plane of the cph structure). The
corresponding networks in the {100}-planes of the fcc- and the bcc
structures consist of squares. The {11 O}-plane in the bcc structure
furnishes a somewhat distorted hexagonal network (moire model No.1).

b) Special Pure Twist Boundary of Partial Dislocations

Boundaries containing partial dislocations are possible in the fcc

structure on {111}-planes, and in the cph structure on the (001)-plane,
since stacking faults can be formed only on these planes. On splitting of
perfect Burgers vectors into partial ones, the b-net, which was originally
composed of triangles, is transformed into a hexagonal network
(Fig. 8.3/1). Upon reverse-translation, an L-net consisting of triangles
is generated in which every second triangle contains a stacking fault
(Fig. 11.4/2) . The single dislocation reaction corresponds to Fig. 9.4/1.
An example in graphite is given in Fig. 11.4/3 .

Fig . 11.4/2. Dislocation network of partial dislocations

 11.4 ... Special" Subgrain Boundaries 127

Fig. 11.4/ 3. Example of a triangular network in graphite

c) Special Partial Twist Boundary

In the case of the special partial twist boundary the same cell structure
as for the pure boundary is cut by the now inclined boundary plane.
Since the b-net remains unchanged, some of the dislocations acquire a

O,2~
I I

Fig. 11.4/4. Bent partial twist boundary. (Bottom half is more of a twist-
while top half is more of a tilt boundary.) (Photo: M. Malone)
128 11. Dislocation Networks-Subgrain Boundaries

mixed character of edge- and screw components. The geometrical con-

figuration of the partial twist boundary is generated as a projection of
the pure twist boundary onto the inclined boundary plane, i.e. it is
derived from the pure boundary by an affine transformation. It is not
necessary that the boundary be planar. It may be a curved surface over
which the dislocations continuously change their character. An example
is given in Fig. 11.4/4.

d) Special Tilt Boundary

A tilt boundary is formed if the boundary surface is parallel to the
axis of rotation, i.e. parallel to the lines formed by all the MS-points.
Along these MS-lines the two crystals are in perfect register. Hence, a
tilt boundary will pass, as far as possible, through these lines. Again
different situations are possible:
1. The tilt boundary may be planar and pass through dense rows
of MS-lines. The intersections of the boundary plane with the cell walls
are again the dislocations. In this case the tilt boundary is formed by a
series of parallel dislocation lines. The dislocations are of pure edge
character and the boundary plane is perpendicular to the Burgers vectors
(see Fig. 12.3/4).

Fig. 11.4/ 5. Example of a stepped tilt boundary in iron.

(Photo : A. Mascanzoni)
 11.5. Foreign Dislocations 129

It was shown in Section 7.3 that two parallel edge dislocations could
form a stable array, if glide were permitted but climb excluded. We now
have the justification for this in that such arrangements are the origin of a
stable tilt boundary.
2. The orientation of the tilt boundary may be random with respect
to the cell structure. However, the boundary will not cut through the
cell walls in an arbitrary manner, but will fit into the MS-lines as a
stepped surface. Hence, such a boundary will be stepped and will consist
of different types of edge dislocation (see Fig. 12.3/4). It does not have
to be even approximately planar but can be curved and broken along
the whole length (Fig. 11.4/5).

11.5. Foreign Dislocations

The "special" twist boundary is characterized by the fact that all
the Burgers vectors lie in the same plane. Now we consider the case of
one dislocation in the network having a Burgers vector which points in
a direction out of that plane. We term such a dislocation a foreign
dislocation. If the b-net is constructed following the methods given in
Section 9.2, we see that the foreign dislocation produces a step in the
b-net (Fig. 11.5/1). Stability considerations of the type given in Fig. 9.3/5

foIA.--:F--:P'--:::iit""-7tj- - -- - - - -- - - - - - - --

~ ' Iii
[00; ;fll]
__
jI[] ~=j
----- -- -------

12
Fig. 11.5/1. Step in the b-net produced by a foreign dislocation
9 Bollmann, Clystal Defects
 130 11. Dislocation Networks-Subgrain Boundaries

can be applied to such steps. Overhanging steps should be transformed

into smoothly inclined ones.

11.6. "General" Sub grain Boundaries

In the case of the" general" subgrain boundary, the axis of rotation
is crystallographic ally oriented in an arbitrary manner. In Section 11.2
it was shown that the b-net should form a plane perpendicular to the
axis of rotation. However, this will be no longer possible with this axis
in a general orientation, since the Burgers vectors can only be oriented
in certain discrete directions. The planar b-net thus has to be approxi-
mated by a stepped one. In addition to a minimum of two linearly
independant Burgers vectors in the "special boundary", two more are
needed for the formation of two independant series of steps in the b-net.
Thus, a really general boundary contains at least four independant
series of dislocations. Naturally, boundaries with three independant
series can also exist.

L b L b

a b
Fig. 11.6/1 a and b. Summary of the different types of subgrain boundaries:
a) special-and b) general boundaries. 1. Pure twist -, 2. partial twist -,
3. tilt boundaries
 11. 7. The Energy of a Subgrain Boundary 131

It was shown in Eq. (11.2-4) that Frank's formula cannot be satisfied

for b-points with b3 =l= 0, i.e. such points do not have an x-line as an
image. In our procedure involving the stepped b-net we disregard this
condition. In Section 12.11 the justification will be given. (It will be
shown that, in addition to the rotation 0, a translation of lattice 2 with
respect to lattice 1 by an amount b3 will be required.)
As far as the orientation of the boundaries (pure-, partial-, etc.) is
concerned, the general- are similar to the special subgrain boundaries.
However, in the general boundary not all the Burgers vectors are
perpendicular to the rotation axis. Fig. 11.6/1 gives a summary of the
various relative orientations.

11.7. The Energy of a Subgrain Boundary

The energy of a subgrain boundary is calculated for the case of a
regular tilt boundary consisting of a row of similar parallel edge dis-
locations. In this manner we essentially follow Read (1953). The absolute
value of the Burgers vectors shall be b, and h shall be the spacing between
the dislocation lines. The angle of tilt () then becomes:

(11.7-1)

We consider a unit (e.g. 1 cm 2 ) of the boundary surface. According to

Section 6.4, the energy per unit length of an edge dislocation is:

E/cm =r= 4n~lb~v) In (r;) +const. (11.7-2)

=~+1;
1;. = elastic energy of the dislocation in its surroundings,
1; = core energy.
The number of dislocations per unit surface is:
N- 1 (c~L (11.7-3)
- h(cm)

Within a distance from the boundary which is much larger than h, the
strain fields of the dislocations will cancel out and the crystals will be
essentially undisturbed. Therefore, we choose ro = h and obtain for the
elastic energy of the boundary:

N· T,=
1
(~).~~-~ln(~)
h 4n(1-V) b
=Eo () In (1/()) (11.7-4)
=-K.)()ln().
9*
132 11. Dislocation Networks-Subgrain Boundaries

We assume the core energy per unit length 1;. to be proportional to b2 ,

i.e. proportional to the section of the cylinder which was cut out as
"bad" material. On enclosing one factor b into () and the other into Eo
we may write:
NI;.=Eo· (). A, (11.7-5)

where A is a constant. Thus, the dependence of the boundary energy on

the angle of tilt becomes:
E =Eo (}(A -In ()). (11.7-6)

The plot of this function starts vertically (dE /d () = 00 for () = 0) and

attains a maximum value for:

(11.7-7)

with a maximum value of the energy:

(11.7-8)

The maximum is attained for values of () between 15 and 25°. The

Eq. (11.7-6) is valied for all kinds of subgrain boundaries (not only tilt
boundaries), if the constants Eo and A are modified correspondingly. The
above behaviour of the energy has been experimentally confirmed
(Read, 1953, p. 192). Typical values are:
Eo R:i 1,200 ergs/cm 2 , Emax R:i 600 ergs/cm 2 , (}max R:i 25°.
Referred to the maximum values, Eq. (11.7-6) becomes:

E ()
---~- [ 1-ln~-
() 1. (11.7·9)

--
Emax - (}max (}max

1.0

0.8 ./
v
~
..:;
0.6
/
/
"-
...... 0.4 /
0.2 /
II
o 0.2 0.4 0.6 0.8 1.0 1.2 1.4
(J/(Jmax-

Fig. 11.7/1. Subgrain boundary energy [Eq. (11.7-9)]

 11.8. Some Considerations of Recrystallization 133

This equation is no longer valid at large angles, (), where the spacing
between the dislocation cores becomes so small that the boundary be-
comes a high angle boundary.

11.8. Some Considerations of Recrystallization

On highly deforming a metal single crystal (e.g. by rolling), the
hardness increases. On annealing for a fixed time at given temperatures,
the hardness decreases as the annealing temperature increases. In the
example (Fig. 11.8/1) of pure nickel, three different stages can be dis-

250
kg/mml '-... I
r-- _I
I1
200
1 1
en
V)

'"co 150 l 1
1
1
"E I
\ I
0

I
.c
~ 100
'"
-'"
u
:>
I
I
I
'J1
50
I I
L I
o 100 200 300 400 500 600 700 800 900 °61000
Temperature
Fig. 11.8/1. Hardness of pure nickel during recrystallization

tinguished. The first ranging from room temperature to 250°C is the

recovery stage, the second from 250 to 400 °C is that of the" primary
recrystallization" and at temperatures > 400°C there is the stage of
"secondary recrystallization".
Electron microscopic examination of the internal state of the material
reveals at room temperature a high dislocation density which, however,
is not homogeneously distributed (Fig. 11.8/2). Small dislocation free
areas are surrounded by a mass of highly entangled dislocations. During
the recovery, the material becomes increasingly more clearly subdivided
into cells (0:::::; 1 fl.) where the cell walls are formed by more or less
regular dislocation networks. This situation is termed" polygonization"
and hence, the recovery stage is due to polygonization.
At 250° distinct crystal grains appear at various spots which grow
out into their surroundings (Fig. 11.8/3)' and which are at least partially
separated from their surroundings by discrete high angle boundaries.
During the growth of these grains the dislocations are annihilated at the
boundary (Fig. 11.8/4). At 400°C all the grains have grown together, and
134 11. Dislocation Networks-Subgrain Boundaries

Fig. 11 .8/ 2. Internal state of nickel after 80% cold rolling

Fig. 11 .8/ 3. Formation of nuclei of recrystallization in nickel at 250 °C

 11.8. Some Considerations of Recrystallization 135

Fig. 11 .8/4. Final stage of primary recrystallization in nickel. The last

polygonized areas disappear

Fig. 11.8/ 5. State after the end of the primary recrystallization

136 11. Dislocation Networks- Subgrain Boundaries

nothing is left of the polygonized structure (Fig. 11.8/ 5) . Hence, the

material has lowered its hardness as well as its energy by eliminating
the dislocations. At this point, the primary recrystallization stage comes
to an end. In Fig. 11.8/5 lines marking the different stages of grain
growth can be seen (analogous to annual growth rings in trees) and they
are probably due to impurity segregation to boundaries existing at that
time.
During further rearrangement, the material can decrease its energy
only by decreasing the total grain boundary surface. This takes place
during the stage of secondary recrystallization when the small grains
disappear, the large grains grow and curved boundaries straighten out
(Fig. 11.8/6}.

Fig. 11 .8/6. Result of the secondary recrystallization

The question concerning the nucleation 0/ the crystal grains now

arises. Let us consider the energy content of a crystal grain in highly
deformed surroundings. This energy consists of two contributions: the
surface energy and the volume energy, i.e. the energy of the dislocation
tangles. The value of the first is increasing and that of the second is
decreasing while the grain grows. We idealize the grain as a sphere. In
this case we obtain:
E =E~ 4n r2 -E~ 43n r3 (11.8-1)
 11.8. Some Considerations of Recrystallization 137

with E~ = specific surface energy, i.e. energy per unit surface,

E~ = specific volume energy.
According to Eq. (11.8-1) the energy of a crystal grain would first
increase with increasing radius, attain a maximum at the critical radius
rC' and decrease afterwards. The critical radius rc follows from dE/dr = 0
as:
2E~
rc= -0-' (11.8-2)
Ev
One could imagine that, as in other nucleation problems, the grain would
first have to be forced to grow while increasing its energy, in order to
overcome the critical size; from this point on it could grow on its own.
Let us remember, however, that the specific energy of a subgrain
boundary depends on the misfit between the two sides of the boundary
[Eq. (11.7-6)]. Thus, with a small misfit E~, the critical radius rc can be
small. Here we assume E~ to be essentially constant. As the dislocation
density is very inhomogeneous in the deformed material it may happen
that at certain spots the local radius of curvature of a dislocation free
area is larger than rc so that the nucleus can grow on its own. With the
growth of the nucleus, the misorientation at the boundary, and hence
E~ and rC' may increase (Fig. 11.8/7). Nevertheless, the radius of curva-

Fig. 11.8/7. Schematic representation of the growth of nuclei

ture of the boundary may still be larger than rc whereby the grain
continues to grow. At an angle of misfit of around 20°, the specific
surface energy E~ attains its maximum value. From this point on the
138 11. Dislocation Networks-Subgrain Boundaries

critical radius remains essentially constant and the grain grows in an

unrestrained manner until it meets another grain.
In the case where originally polycrystalline material was deformed,
old high angle boundaries can act as nuclei for new grains. Here the
driving force is given by the difference in volume energy (dislocation
density) between the two sides of the old boundary. Once the boundary
moves, the volume swept out is dislocation free, i.e. only one side of the
boundary contains dislocations, and this sustains the driving force. As
the boundary is relatively straight from the beginning, no critical radius
has to be overcome (Bailey, 1960).

11.9. Limiting Cases

In Sections 11.2 to 11.6 it was shown how the theory of subgrain
boundaries can be developed by employing Frank's formula together
with the dualistic representation. Now we consider the limitations of
this theory.
It is known that the accuracy of Frank's formula diminishes with
increasing misfit between the two crystals. In addition, the formula is
only valid if both crystals are rotated with respect to each other, i.e.
for grain boundaries (with small misfit) but not for phase boundaries.
Chapter 12 treats a generalization for arbitrary crystalline interfaces
and Frank's formula proves to be a special case of a more general
equation.
A further problem arises if a dislocation branches out from a network
into one of the undisturbed crystals. On attempting to draw the b-net
for this case, we find that there is no longer a "one-to-one" relationship
between a b-net and an L-net (Fig. 11.9/1). The configuration of the b-net

L b

Fig. 11.9/1. Breaking away of a dislocation from a dislocation network

 11.10. Dualistic Representation in Three Dimensions 139

itself contains a "dislocation". The b-net is no longer topologically

planar (i.e. it can no longer be spread into a plane). It is also not possible
to define L-fields at the spot where the dislocation branches out. This
difficulty will be cleared up in Section 14.3.

11.10. Dualistic Representation in Three Dimensions

On trying to join several subgrain boundaries, it can be seen that it
is no longer possible to represent such an arrangement by a closed
b-net. As every L-net has its own mesh width, i.e. its own 0, the different
dislocation networks become interconnected by dislocations branching
out of one boundary and being incorporated into the other. Thus, a
common b-net would contain many "dislocations" (Section 11.9).
However, another possibility to extend the duality to three dimen-
sions becomes apparent if, instead of a dislocation line, the whole surface
of the subgrain boundary, and instead of the Burgers vector the O-vec-
tors are considered. Now, in place of the two-dimensional fields limited
by dislocations, the three-dimensional subgrains are numbered. The
b-space is replaced by a space of O-vectors where the composition of

---- - -------
c ----
e
I 2
0-------0
1 I 2
I 2
=~
I 2
= 0+---0

2 I
2 I 2 0----0

3
I 2 ~2

4
I 2
:I :I
Fig. 11.10/1. Three-dimensional duality between subgrains and rotation
vectors e. 1. Right-handed twist boundary, 2. left-handed twist boundary,
3. partial twist boundary, 4. tilt boundary
140 11. Dislocation Networks-Subgrain Boundaries

these vectors forms a three-dimensional framework. Hence, instead of

the b-nodes the nodes in this framework shall be numbered. The con-
vention shall be fixed such that, at a junction of two O-nodes (numbered
Nl and N 2 ), the node designated by Nl (arbitrarily chosen) shall be the
tail and the N2-node the tip of the O-vector whereby the O-vector follows
the normal right-handed screw convention. The subgrain Nl shall be
kept fixed and the subgrain N2 shall be rotated by an amount O. It can
be seen that the relative orientation of the two subgrains is invariant
while Nl and N2 are interchanged.

c e
3
1

1- 1<] 2

1<]
2

2
~ 3

1- [>1
2

>+- [>1
3

4
2

2\73
5
>+- 1
f,- \72
3

6
I

i
Fig. 11.10/2. Dualistic representation of the junction of three subgrains
(Abbreviations: Tw twist boundary, Tt tilt boundary, r right-handed
rotation, lleft handed rotation). 1. All three boundaries Tw r; 2. (1,2) Tt,
(2,3) Tw 1, (3,1) Tt; 3. All three boundaries Tw 1; 4. (1,2) Tt, (2,3) Tw r,
(3,1) Tt; 5. (1,2) Tw 1, (2,3) Tt, (3,1) Tw r; 6. All three boundaries Tt
 Chapter 11. Summary and Discussion 141

Fig. 11.10/1 represents different kinds of subgrain boundaries. The

boundary surface is oriented normal to the plane of the paper.
The following dualities exist between various entities in the C-space
(crystal space) and in the O-space:

C-space B-space

1. A subgrain (3 dimensions, corresponds to a point where several

abbreviated 3 D) 8-vectors meet (OD)

2. A subgrain boundary (2D) corresponds to a 8-vector (1 D)

3. An intersection of sub- corresponds to a polygon of 8-vectors

grain boundaries in a line (2D)
(iD)
4. An intersection of several corresponds to a polyhedron of 8-vectors
subgrain boundaries in one (3D)
point (OD)

Examples of possible subgrain boundaries meeting in a line are

given in Fig. 11.10/2.
This kind of duality between subgrains and 6-vectors is an ap-
proximation of the same type as is represented by Frank's formula.
Rotations can never be added exactly as vectors. In reality they are
noncommutative tensors. The smaller the angles 6, the better, however,
becomes the approximation.

Chapter 11. Summary and Discussion

In this chapter, the theory of subgrain boundaries has been treated,
i.e. the boundaries between two crystals of the same kind slightly ~
with respect to one another. These boundaries are formed of dislocation
networks. Their description is based on the interpretation of Frank's
formula [Eq. (11.2-1)J which represents an algebraic relation between
the b-space and the L-space. Frank's formula can be interpreted as an
imaging relation between a plane in the b-space and the complete
crystal-space (L-space). The b-plane passes through the origin of the
b-space and is perpendicular to the axis of rotation 6. For every point
in the b-plane, there exists a corresponding x-line parallel to 6 in the
crystal space.
If Frank's formula is quantized [Eq. (11.2-5)J, the discrete b-points,
which are possible combinations of Burgers vectors, are imaged as a
set of discrete x-lines parallel to 6. These lines are composed of minimum
strain points, abbreviated to MS-points. The separation of the MS-lines
142 11. Dislocation Networks-Subgrain Boundaries

by cell walls produces a cell structure. The possible dislocation networks

forming the boundary can be understood as sections through this cell
structure. A classification of subgrain boundaries is developed on this
basis (Section 11.3). For the special orientations of (), the situation is
uniquely described by the quantized Frank's formula, limited by the
fact, however, that Frank's formula itself is an approximation. The
procedure also enables the study of dislocation networks containing
partial dislocations (Section 11.4).
For the general orientations of (), the b-net becomes stepped, as
"foreign" dislocations (Section 11.5) are introduced into the boundary.
Frank's formula is violated by this procedure. Nevertheless, the duality
relations do hold for the correlations between Burgers vectors and dis-
location lines. The (b l , b2 )-coordinate axes of the b-net are perpendicular
to the (Xl, x 2)-axes of the L-net, since b stays perpendicular to x.
It was shown ill the Summary of Chapters 7-9 how the topological
properties of the b-net could be derived from the configuration of the
L-net. The justification for such a procedure lies in the fact that every
dislocation arrangement of some complexity will tend to approximate to
a subgrain boundary.
The above theory of subgrain boundaries is not valid for large angles
of rotation e, i.e. in the domain of general grain- and phase boundaries.
The theory will be extended into this domain in the following chapters.
On the other hand, the duality relations cannot be fully applied if the
L-net is no longer topologically planar (branching of a dislocation out
of a network, Section 11.9). This problem will also be solved in Sec-
tion 14.3. In Section 11.10 it was shown that it is possible to generalize
the duality relations between the individual subgrain boundary and its
b-net to a dualistic representation of a polygonized structure.

Suggestions for Problems

Many cases of special and general subgrain boundaries can be con-
structed and discussed in detail. The configurations given in the article
by Amelinckx and Dekeyser (1959) which are represented in Thompson's
notation, can be transformed into the dualistic representation. With
respect to the energy of subgrain boundaries, exercises can be found
in the books of Read (1953) and of Weertman and Weertman (1964).
 12. General Geometrical Theory
of Crystalline Interfaces
12.1. Introduction
A boundary surface between any two crystals will take up such a
position that the crystals exhibit more or less optimum matching. Our
task will be to formulate the notion of the" matching" of two periodic
structures in a correct mathematical manner. A start has already been
made in the case of the MS-points of the quantized version of Frank's
formula [Eq. (11.2-5)].
We idealize the two crystals as two interpenetrating mathematical
translation lattices, i.e. we disregard space filling atoms. Then we look
for positions of "best fit" in these interpenetrating lattices. It will be
shown in the next sections how this is achieved. When such positions
have been distinguished, a prospective boundary can be placed through
them. A special example was the "honeycomb" -structure of the sub-
grain boundaries (Section 11.2). For monatomic materials (e.g. fcc or
bcc metals) the interpenetrating lattices are understood as being two
sets of positions for one set of atoms, which, depending on the set chosen,
can form either crystal 1 or crystal 2. Once the boundary is fixed, the
atoms may be fitted into set 1 on the one side, and into set 2 on the
other. The boundary itself remains as an approximately two-dimensional
remnant of the two interpenetrating three-dimensional lattices.
The interpenetrating lattices with the positions of "best fit" re-
present a complete description of all the possible boundaries between
two crystals of given structures and orientations. As soon as the boundary
is chosen and "realized", the remainder of the interpenetrating lattices
lose their meaning.

12.2. The Coincidence-Site Lattice

Let us consider two interpenetrating translation lattices (lattice 1
and lattice 2). In our investigations, we usually assume lattice 1 to be
fixed and all the changes such as: translation, rotation, etc. are performed
with lattice 2. With the relative orientation of the two lattices given,
lattice 2 is translated in such a way that one of its points coincides with
a point in lattice 1. That point shall be termed a "lattice coincidence
site". An atom located there is in an unstrained position in both lattices.
For this reason we regard this point as a "point of best fit".
144 12. General Geometrical Theory of Crystalline Interfaces

We may ask whether other examples of lattice coincidence occur

within the interpenetrating lattices: In effect there exist either none, or
one (the one which was arbitrarily produced) or, due to the periodicity
of the two lattices, an infinite number of them. This infinity forms a
lattice of coincidence sites, i.e. a coincidence site lattice.
Taking the example of two equal, two-dimensional square lattices,
we examine the orientations (of lattice 2 if lattice 1 is fixed) for which
coincidence site lattices exist, and we investigate the structures of these
coincidence site lattices. We hereby fix the coordinate system such that
the coordinates of all the lattice points (of lattice 1) are integers (crystal
coordinate system Section 2.2). Initially, both lattices are placed into
the same orientation and brought to full coincidence. Then, choosing a
lattice point as origin, we rotate lattice 2 by the angle e in the positive
sense around the coordinate origin. In this way the pattern formed of
lattice points of both lattices varies with e. After a rotation of 90° both
lattices once more coincide completely. However, already with rotations
of up to 45°, we obtain all the possible patterns if we regard those
patterns as being identical which can be superposed by rotation. Thus,
the patterns for e and for 90° - e are identical.
Inside lattice 1 we delimit a sector of 45° by the two mirror axes
[1, OJ and [1,1]' Then, we subdivide this sector into two halves of 22.5°
(Fig. 12.2/1). One of these sectors is limited by the [1, OJ-axis and the

y
4
/
V
m2 /
V
;/ ~i'-
v:
X85
/
/ 51 31 3
/ V41 13\ 55

/
/
V 25 11 5 2!j....-' ~
/
V13 5 ~~ 53

/ ~5/, ~ 11 V
13 \31 ~25 Z
~~21 'j
~ 1.":;0 - ~
.....", ml x
~ t::-- I i
.KIll r<:::- ............ I 1
'- r-.. . .
'-,-
~

Fig. 12.2/1. Determination of the E-values for rotation on the {100}-plane

in the cubic structure (see Table 13.3/T1, No. 1-22)
 12.2. The Coincidence-Site Lattice 145

other by the [1, 1J-axis. Any point from outside such a sector can be
transferred into a point inside by a rotation of 0 < 45 0. The angle
between the mirror axis and the position vector of one of the two points
related in this way is 0/2. Fig. 12.2/1 shows that the structure of the
pattern does not depend on whether a point of sector 1 is rotated into
one of sector 2 by + 0 or vice versa. The orientation of the pattern,
however, is not the same. On rotation by positive angles 0 the correspond-
ing lattice coincidence sites lie in sectors 2 and 4, and for negative angles
o in sectors 1 and 3.
In our example of the square lattices the coincidence site lattice must
be itself a square lattice, since for every point rotated over the x-axis
a similar point must be rotated over the y-axis. A measure of the coinci-
dence site lattice is given by the surface of its unit cell. When considering
whether a point within a sector (e.g. sector 2) is a possible lattice coinci-
dence site close to the origin, it must be assured that its coordinates (x, y)
do not contain common factors, since this would mean that a fraction
of the position vector would again lead to a coincidence site. In particular,
the two coordinates should not be even.
For a chosen point (x, y) of lattice 1, for example (Fig. 12.2/2), the
surface of the square of the position vector is given by:

E'=X2 +y2. (12.2-1)

For (x, y) = (3,1) E'= 10.
Both coordinates 3 and 1 are odd numbers, hence the sum of their
squares is an even number. It is evident, however, that the square
(formed by the origin and three coincidence sites) contains a coincidence
site in its centre. [(3/2, 1/2) +(
-1/2, 3/2) = (1, 2)]. For this central point
+
E' = (1)2 (2)2 = 5 and this is the smallest unit for the given angle of
rotation
0=2 arc tan (y/x) (12.2-2)
(=36° 52.2' in our example).
If E' is even, then its value has to be divided by 2 in order to obtain
the smallest unit.
(12.2-3)

with IX = 2 if E' is even, and IX = 1 for E' odd. E is called the" generating
function" .
For the point (4,1), E'= 17 and the centre of the square [(2,1/2) +
(-1/2, 2) = (3/2, 5/2)J is not a lattice coincidence site, as its coordinates
are not integers. Care must be taken with large coordinates since, oc-
casionally, certain lattice coincidence sites can belong to several coinci-
dence site lattices.
10 Bollmann, Crystal Defects
146 12. General Geometrical Theory of Crystalline Interfaces

Fig. 12.2/2. Special coincidence site lattice [2'=5, 0=36 0 52.2',

tan (0/2) = 1/3J (Table 13.3/T 1, No. 19)

This example of a coincidence site lattice has been treated here in

such detail because, in Section 13.3, it will become important in a more
general form for the study of the periodic or nonperiodic structure of
patterns.
The coincidence site lattice is not yet the full solution of the problem
of our search for locations of best fit. It still contains serious deficiencies.
First of all, the knowledge of the position of a coincidence site tells
us nothing about its surroundings except that the pattern of lattice
points must be prieodic (with the periodicity of the coincidence site
lattice). The converse is not true, since periodic patterns may exist
without coincidence sites.
Secondly, as Fig. 12.2/3 shows, it is practically impossible to identify
a coincidence site in a more complicated pattern. The reason lies in the
fact that the concept of the coincidence site contains no measure of the
spacing between two points. Whether two points coincide or are 0.0001 A
apart may be physically insignificant. Mathematically, however, the
situations are fundamentally different.
Thirdly, let us assume the orientation of lattice 2 to be such that
there exists quite a dense coincidence site lattice. This lattice is com-
pletely destroyed by an infinitesimal change in the orientation of
 12.2. The Coincidence-Site Lattice 147

o
o 0
000
o 0
0 0 0 00 0
0
o
o 0 0 0 0 0 0 & Q, co
0
0 0°0 0
0
,,° 0
0
Q
0
000 0 0 0 0

o 0 0 00 &> 0 000 0 q., lib g 00 0° t:! 0 0 e 0 0 0 0

o 0 0 0
o 0 0 0 00 0 0 (!I 00 Q 0 0 00 ; CD &Q 00 0 0000 0 0 0

<b ~ 000 cf °0 e 0 0> 0 0000 () 0 & 0 0 0

Cb
°
DOD

~ ()::I CO Q Go 0 0 0 0 0 8 q,o ~ 0 0 00 0 t:P '0 0 0

00 0 0 0°0
00 I> 0 0 0

o 0 00 °0 0 0 rP 0 0000 c:» <>0 9 000 0 00 e Q co 0 0 0

00 0 00 0 0 0
o ~ 00 0 000 8 q, ~ 0 0 0 0 cP 0 • 00 cP 0 00 0 ~ Cb 0
o
0°00
o
0 00 0 <110
0000
0 [} 000
0
0 00
0
e 0°
0
0
00
Q co 0 0 0 0 " 0 C) 0 0 0 0 0 CP
0 0
o 0 0 -. 0 0 0 0 ~ 0 0 0 o .p 000 0 Q, '0 IJ 000 .<J 0 0 e 0 ti)
0 0 0 00 0°00
o 0 0 00 6 ~ 00 0 0 0 0 ~ 0 6) 0 0 0 0 0 0> ., Q '\) <S" 000 8 a
0 0 0 0 0
o 00 &> 0 00 0 0 Q, Cb (Ii0 000 f! 0 0 e 0 0
°0
f'D 0 00 0 c. 0 " 00
0
o 00000 . 00000
o 00 t!I 0 ~ 000 0 00 0 Q. 00 fS 000 8 ~ ~ 0 0 0 0 & 0 0
o 0 0 0 0
o 0 000 00 °0 , 0 fP 0 0 0 0 CD ~ (; 000 0 Q)
o 0 on 0
0
0
0
0
0 0
co 0
0°0
0
0
0 e ca
cP ~ 9 00 0 000 B "" 'I) 0 0 0 0 & 0 0 0 0 .p 000 8 Q, 0 a
o 0 0 0 0 0 0
Ol 0 0 00 • Q... f} 000 0 Cb
0
~ CC) 0 0 0 " 0 0 Q 0 0 °0 0 e,0
0°0 0 0000 00
o q, -.0 0 0 0 d' • 0 0 .p 000
0000 0 0°0
o
Q, '0 IJ 000 .J> ' "
0
0 8 e
o 0 000 'B 0 CQ 0 000 ~ 0 ~ 0 0 0 0 cp 0 Q 00 0 0 0 °
o 00 0 0 00 0
o 0 0 .p 0 o o. 8
Q, '0" 00 Q Q .s:> 0 cJ> 0 0 " 0 r! '" e
o 0 0
o
~ 0 4> 0 0 0 0 (1) Q. Q 00 0 000 8 ~ ~ 0 0 0
00 0 0 " 0
o 0 0 0 0 0
o 0 0 000 (j o~ & 0 0 dJ 0 00 ~ OJ 00 (; 000 0 0 0 0
o 0 Q 00 0 000 8 q, ~ 0 00 0 ct °0 0 0 0 0 0
o
o 0 0 0
0
0 0
0
c.
0
'b (} 000 0 Q:) e Q 0 0 0 0 0 0 0
o 0 0 0 0 0
o 0 0 0
0000 00

o 0 0
o

Fig. 12.2/3. Superposition of two square lattices. The picture contains four
coincidence sites

lattice 2, although the change in the physical situation is expected

to be infinitesimal. In addition, there are an infinity of orientations
for which no coincidence site lattice exists, namely those for which, in
the case of the square lattices, tan ((J/2) is irrational. Hence, the coinci-
dence site lattice is extremely discontinuous in its behaviour for changes of
the orientation of lattice 2. In contrast to this, the MS-points (Section 11.2)
are continuous functions of (J.
In order to be able later to generalize the concept of the coincidence
site lattice we have to express the facts somewhat more mathematically.
Each of the two lattices can be understood as a translation group con-
structed by means of three linearly independant basic vectors. We may
correlate the two lattices point by point by a linear homogeneous trans-
formation A
IAI =1=0. (12.2-4)
The index L means" lattice point". In our example, A is a rotation, but
it may be a general affine transformation (expansion, rotation, shear,
10·
148 12. General Geometrical Theory of Crystalline Interfaces

etc.}. The determinant IAI =1=0 means that both lattices shall be defined,
with respect to the number of dimensions, in the same space, i.e., for
example, a three-dimensional lattice shall not be imaged onto a two-
dimensional one.
The whole set of the lattice points may be called an "equivalence
class" or simply a "class". Hence, the two lattices are two related
classes of points [related by Eq. (12.2-4)]. Every lattice point is therefore
an "element" of one of the two classes. Lattice coincidence sites are thus
coincidences of elements of related classes. [They are not coincidences of
related elements, since by the relation in Eq. (12.2-4), a coincidence is
only possible at the origin.]
In the next section we shall generalize the notion of the coincidence
site lattice based on the above reflections.

12.3. The O-Lattice

In Section 12.2 the lattice coincidence sites were defined as being
coincidences of elements of related point classes, where the relation was
given by Eq. (12.2-4).
The whole space of a crystal does not only consist of lattice points
but also of the interstices between them; or expressed mathematically,
not only of the translation group but also of all their cosets. A coset is
formed of a position vector of a point within a unit cell plus the complete
translation group.
When we choose the crystal coordinate system as being given by the
unit vectors of lattice 1, then all the lattice points are indicated by
integral coordinates. Considering the (cyrstal) coordinates of an arbitrary
point e.g. (12.138, 7.243, -4.421) we designate as external coordinates
the integers (12, 7, - 5) and as internal coordinates (0.138, 0.243, 0.579).
Hence, the external coordinates designate the unit cell and the internal
coordinates the positions within that unit cell. The internal coordinates
Xi shall always have values such that o:s;; Xi < 1 .
In the crystal coordinate system all the points of a coset have the
same internal coordinates and so may be designated by them. Every
coset is represented inside every unit cell (Fig. 12.3/1). Together with
the translation group having internal coordinates (OOO), the infinite
number of cosets forms a "partition": a subdivision of the whole space
into "equivalence classes".
Through the equation
~(2)=A~(1)
IAI =1=0 (12·3-1)
every point in crystal 1 is related to a point in crystal 2, and hence every
equivalence class of crystal 1 is related to one in crystal 2. The related
 12.3. The O-Lattice 149

class in crystal 2 has the same internal coordinates in the coordinate

system of crystal 2 as the original class in crystal 1. While Eq. (12.2-4)
was only formulated for lattice points, Eq. (12.3-1) includes all the
points in the space. As such, the two interpenetrating lattices represent
two different partitions of the same space.

1. [0 ,01
2. [1/5,3/51
3. [2/5,1/51
4. [3/5,4/51
5. [4/5,z;51

Fig. 12.3/1. Equivalence classes (Table 13.3/T1, No. 43)

We now define" O-points", the generalized lattice coincidence sites,

as coincidences of elements of related equivalence classes, no matter what
may be the values of the internal coordinates of these classes.
Hence, the lattice coincidence sites are special O-points with the
internal coordinates (000). Roughly speaking, the O-points can be
understood as coincidences of internal coordinates or coincidences of
points which are in equivalent positions in the two crystals.
We shall now derive an equation for the O-points based on the above
definition. We start from an arbitrary point X(l) in crystal 1 with arbitrary
external and internal coordinates. We term as C(l) the equivalence class
of that point x(1). The corresponding point X(2) in the related class C(2)
is given by
(12.3-2)
1 So 12. General Geometrical Theory of Crystalline Interfaces

Alternatively, we can, starting from reel) (C(l»), find other elements e (C(l»)
of the class C(l) by adding translation vectors tel) of lattice 1 to re(l).
Hence:
(12·3-3)
If the points defined by Eqs. (12.3-2) and (12.3-3) coincide, we designate
that coincidence point by re(O) (Fig. 12.3/2):
re(lI) (C(II») =re(l) (C(l») +t(l) =re(O) (C), (12.3-4)

Fig. 12.3/2. Coincidence of elements of related equivalence classes

where C means that the point belongs simultaneously to both related

classes C(l) and C(II). Together with Eq. (12.3-1) we obtain:
re(O) = A -1 re(O) + tel)

or
(l-A-l) re(O) = tel) (12·3-5)
(I = identity = unit transformation).
As we shall see later (Section 12.5) that all the possible translation
vectors tel) of lattice 1 form the b-Iattice, we already change the notation
here; hence t(l):= bILl (L means lattice point). Thus:

1 (l-A-l) re(O) = bILl I. (12·3-6)

All the o-points are solutions of Eq. (12.3-6). We shall see that this
equation is the basis of the whole geometrical theory of crystalline
interfaces.
The general procedure for solving the equation is as follows:
1. Choice of the coordinate system.
2. Formulation of the transformation A which relates the two
lattices.
 12.3. The O-Lattice 151

3. Formulation of the basic Eq. (12.3-6) by matrix operations.

4. If the determinant II-A-11 *0, the solution is

(12·3-7)

The problems arising when II-A-11 =0 will be discussed in Sec-

tion 12.9.
In order to get acquainted with the procedure we calculate the
example of Fig. 12.2/2, where the difference between the coincidence
site lattice and the O-lattice becomes evident. As the basic system we
choose the orthogonal coordinate system of crystal 1. We formulate the
problem two-dimensionally. The transformation A of Eq. (12.3-1) in
this case is a rotation B:

A=B= (
COS 0 - sin 0) . (12·3-8)
sin 0 cos 0
From this follows:

A-I = (
COS 0
sin 0) (12·3-9)
- sin 0 cos 0
and

I-A-l = ( 1 -cos 0 -sin 0)

. (12-3-10)
( ) sinO i-cosO

If II-A-11 =\=0 the solution is given by Eq. (12.3-7). In our example the
determinant is:
II-A-11 =2 (i-cos 0) (12·3-11 )

and the inverse matrix (l-A-l)-l becomes:

1/2 (1/2) cotan(O/2))

(I_A-l)-l= ( -(1/2)cotan(O/2) 1/2 . (12.3- 12)

As in the crystal coordinate system the two basic translation vectors are
given by the coordinates (1, 0) and (0, 1), the unit vectors of the O-lattice
are the cloumn vectors of the matrix (12.3-12), i.e.

u~O)= (1/2, -(1/2) cotan (0/2)) and u~O)= ((1/2) cotan(O/2), 1/2).

The geometrical interpretation of this result is given in Fig. 12.3/3.

u~O) is determined by the point of intersection of a straight line, which
subtends an angle 0/2 on the negative y-axis, with the vertical line at
x = 1/2, and correspondingly al~O) with the x-axis Fig. 12.3/4 shows two
tilt boundaries selected from the same O-lattice.
152 12. General Geometrical Theory of Crystalline Interfaces

x.1!z
( \ \ ~

\ I \ ~ \ ~

-lr\ -t-I 1\ \\ ~ --
~
...-'-

\! .q= XIO) -
t-~ -- --1+ -- ....<'y-
~

l/Z

--
~
~
- ,.- 1\1 \
- 1-
--
~X

f\ ~r- ).~ T' \

81Z x(O)

- ~
!-(
\
1\ 1\ .1 e---- ~ ~

_\
I-
)-- ~ ~
f- I \ 1\
Fig. 12.3/3. Construction of the O-lattice according to Eq. (12.3-12)

In the special case of the coincidence site lattice Fig. 12.2/2,

cotan(0/2) =3 and

x~Ol
(xiOl) -_ (
1/2 3/2)
-3/2 1/2 b~Ll·
(biLl)
(12.3-13)

The unit vectors of the o-lattice are '140) (1/2, - 3/2) and u~O) (3/2, 1/2)
(see Fig. 12.5/1). Every second O-point is a lattice coincidence site. The
other o-points belong to the equivalence class [1/2, 1/2] * (Fig. 12.5/1).
In the case of cotan (0/2) = 4 the following equivalence classes
contain coinciding elements: [0,0], [1/2,0], [0,1/2] and [1/2,1/2],
which may be shown as an exercise.
We see that the O-lattice, in contrast to the general coincidence site
lattice, is a continuous function of the transformation A (e.g. of 0) and
that an O-lattice also exists where no coincidence site lattice is present.
A slight change of A leads to the instantaneous disappearance of a
coincidence site lattice, while the O-lattice undergoes a smooth change;
a property which is common to the MS-points (Section 11.2). It will be
shown in the next section that Eq. (12.3-6) can be understood as a
generalization of the quantized version of Frank's formula [Eq. (11.2-5)],
and that, hence, the MS-points may also be looked on as special o-points.
* The equivalence classes are indicated by boldface brackets [ ... ].
 12.4. The O-Lattice and Frank's Formula 153

°
I °
° ° °
- -t,t-'*---4--_°-1._o
° ° fo '0 ° -
X(O)_ - '-t--'-_-J~'--''''-'._.-_-,-_0
2 ° ° - Xm
-1-"'- U ° ° 1
, "'- -X(2)
° ° ~ 1
- -t-Ti-r-'*_°--JLo
° ° ° ° "° -
--f-"'--1,"---1H-_-L °
° I ~ "'-,.-t-
A 0' ° ° °
8
aJ bJ

cJ dJ
Fig. 12.3/4. Application of the O-lattice to the construction of a symmetrical
and an unsymmetrical tilt boundary

12.4. The O-Lattice and Frank's Formula

In Section 11.2 the theory of subgrain boundaries was developed
from Frank's formula. If we can show that Eq. (12.3-6) is a natural
generalization of that formula, then we find that there exists a con-
tinuous transition from low-angle boundaries to general crystalline
interfaces. In terms of coordinates Frank's formula is:
bi -Ox2
=

b2 = OXI (12.4-1)
ba = 0
1 54 12. General Geometrical Theory of Crystalline Interfaces

where b (bt b2 ba) represents the discrete points of the b-net and (J is
placed in the z-axis. In matrix notation this equation becomes:

(12.4-2)

On the other hand, according to the procedure given in Section 12.3, for
a rotation around the z-axis, A is given by:

A= CO -sin ()
s~() cos ()
0 ~) (12.4-3)

C)
and Eq. (12.3-6)

b~L)
b~L)
= r-co,o
sin ()
0
-sin ()
1 -cos ()
0
0)C~)
o
o
x~O).
x~O)
(12.4-4)

If () is smail, cos ()~1 and sin ()~(), and Eq. (12.4-4) becomes:

0)o (X~O»)
x~O) (12.4-5)
o x~O)

which is identical with Eq. (12.4-1). The b(LLvectors representing

translation vectors of lattice 1, and in Frank's approximation also of
lattice 2, are the Burgers vectors, and the ;x:(O)-points become the MS-
points. Therefore, the theory of subgrain boundaries is a special case in
the general theory of crystalline interfaces.

12.5. The b-Lattice

The basic Eq. (12.3-6)
(1 -A-t) ;x:(O) = b(L)

can be interpreting as being an imaging relation of a b-lattice onto the

two interpenetrating lattices 1 and 2, as was done in Section 11.2 in the
representation of Frank's formula. The b(L)-vectors are defined as the
translation vectors of lattice 1. We now define the b-lattice as the lattice
formed by all the possible translation vectors of lattice 1 when these are
rearranged such that they have a common origin (Fig. 12.5/1). Thus,
b(L) stands for the lattice vector of the b-lattice. The b-lattice, from the
point of view of geometry and orientation, is identical with the trans-
 12.6. Various Aspects of the O-Points 155

Crystal Space b Spac,

Fig. 12.5/1. The b-Iattice pictured as the lattice of all the translation vectors
of lattice 1

lation lattice 1. Therefore, the coordinate system of lattice 1 also fixes

that of the b-Iattice, and this, in tum, determines that of the o-lattice.
We shall see later that this dualistic representation will be essential
for the development of the theory.

12.6. Various Aspects of the O-Points

According to the basic definition, the a-points are positions where
the two interpenetrating lattices show "best fit" (Section 12.1). This
concept was specified in mathematical terms as the coincidences of
elements of related equivalence classes (Section 12.3); in this form it is
a generalization of the notion of lattice coincidence sites.
Furthermore, the a-points proved to be generalized MS-points (solu-
tion of the quantized Frank's formula, Section 11.2). If we try to
generalize Eq. (11.2-5) in order to obtain the basic Eq. (12.3-6), two
problems arise:
1. In Frank's formula, re(l) ~re(B) ~re was not exactly fixed. The
question now arises concerning the exact position of re for larger angles ().
(re = re(l), or re(B), or some point in between?)
2. According to Frank's formula, b remains perpendicular to re, and
hence the b-coordinate system is perpendicular to that of the re-vectors,
where b =re(2) _re(l). This condition can only be maintained if re is an
average value of re(l) and re(2). Otherwise the b- and the x-coordinate
156 12. General Geometrical Theory of Crystalline Interfaces

systems are no longer perpendicular. Here the question arises concerning

the meaning and orientation of the b-net.
Both questions can be answered if we assume that every lattice
coincidence site is an MS-point. {The converse does not have to be true.}
Comparing the dualism between the b- and the L-net given in Fig. 11.4/1
and the construction of the coincidence site lattice in Fig. 12.5/1, we
see that ;v =;v(O} has to be identified with ;V(2} under the condition that
b = ;V(2} - ;V(l} = btL} {12.6-1}
is a translation vector of lattice 1. The b-Iattice is thus defined as the
lattice of all the translation vectors of lattice 1 and the O-lattice as the
lattice of those points, ;V(2} for which:
{12.6-2}
Together with Eq. {12.3-1} we obtain:
;v(O} _ A-I ;v(O} = btL}
or
(12.6-3

~~~~-t---+,.L\

~~~~~~~~~~~~,~~~
-+--~~--~~~~~~~~'7_~-~~
,

\-)(
~,
Fig. 12.6/1. o-points as origins of the imaging between lattices 1 and 2
 12.6. Various Aspects of the O-Points 157

which is the basic Eq. (12.)-6). This is actually the way that the O-lattice
was found.
At this stage the O-points, the generalized MS-points, were localized
but their meaning was not immediately apparent; the meaning, however,
is clear from Fig. 12.6/1 (Fig. 12.6/2 shows the complete pattern).

Fig. 12.6/2. The complete pattern of Fig. 12.6/1, showing the symmetry of
the point arrangement (same pattern as Fig. 12.2/3)

By means of Eq. (12.1-4), lattice 2 was "created" out of lattice 1

by the transformation A where, for convenience, a lattice point was
chosen as the origin. It can be seen that the same lattice 2 in the same
position can be created out 0/ the same lattice 1 by the same trans/ormation A
but starting from other origins. Naturally, on starting from a new origin,
the pairing (the correlation) of the lattice points (of lattices 1 and 2) is
changed. Formulated in an equation, this signifies a coordinate trans-
158 12. General Geometrical Theory of Crystalline Interfaces

formation by translation. Lattice 2 was defined by Eq. (12.2-1)

(a:(2L) = Aa:(lL)}. The new origin shall be a:(0), i.e.

(12.6-4)
The point a:(2L) is here no longer related to a:(lL) but to another point
of lattice 1 i.e. a:(lL) +b(L) (i.e. a:(lL) +t(l»). From Eq. (12.6-4) it follows
that
(12.6-5)
and, applying Eq. (12.1-4)
_a:(0) =Ab(L) -Aa:(O). (12.6-6)
Upon multiplication of both sides by A-I, the basic Eq. (12-3-6) is
produced. Therefore, all the O-points are origins of the type described
above.
Naturally, the various meanings of the O-points indicated in this
section, such as:
Coincidences of equivalent points,
Generalization of MS-points, and
Origins for the relation between lattice 1 and lattice 2
are different manifestations of the same situation. However, which of
the various aspects has to be emphasized will depend on the problem
being treated.

12.7. The Actual Formulation of the Transformation A

We start with the general case where two crystals of different struc-
tures in a given relative orientation are joined. We assume that the
lattice constants a, h, c, (x, p, Y (Section 2.2) of the two lattices are
known. Starting from a cubic lattice of unit period 1 A we "produce"
lattice 1 by the transformation S(l)
a:(1) = S(l) a:(orth) • (12.7-1)
S(l)is formulated according to the general method of Section 2.2. Of the
nine coefficients, three can be chosen by fixing the orientation of the
unit cell (e.g. c in z-axis and a in (xz)-plane]. The detailed procedure is
given in (Eq. A1-26). Lattice 2 is produced in the same manner, but in
order to be free to choose its orientation we rotate it by R. Hence:
a:(2) =RS(2) a:(orth). (12.7-2)
By eliminating a:(orth), we obtain
a:(2) = {RS(2)S(lH} a:(1):= A;:v(l). (12.7-3)
 12.7. The Actual Formulation of the Transformation A 159

A is composed of three different transformations. Eq. (12.7-3) is valid

in the orthogonal coordinate system. From this follows:

(/_A-1) = (/-8(1) (8(1»)-18-1) (12.7-4)

and, if I/-A-1 1 +0, according to Eq. (12.3-7):

w(O) = (/-8(1) (8(8»)-18-1)-1 b(L). (12.7-5)
As this equation is formulated in the orthogonal coordinate system, the
unit vectors of the b-lattice are the column vectors of the matrix 8(1).
In this manner, the unit vectors of the O-lattice are determined as images
of the b-unit vectors.
If both crystals are of the same kind, i.e. if we deal with grain
boundaries, 8(8)=8(1):= 8, and hence A=R. The 8-matrix then only
plays a role in the determination of the b(LLvectors.
It will become useful to express all the vectors in the crystal co-
ordinate system (of lattice 1). Then, every matrix M is transformed
from the orthogonal- to the crystal coordinates (Index C) by:
MlC) = (8(1»)-1 M8(1) (12.7-6)
and all the vectors w by:
(12.7-7)
As we deal here with a coordinate transformation and not with a vector
transformation, the w in Eq. (12.7-7) has the same significance as in e
Eq. (2.2-5).
In the crystal coordinate system A becomes:

(12.7-8)
and the basic equation:

i.e.
(12.7-9)
As an example, we investigate the rotation on the (110)-plane in the
bec structure, which is illustrated by moire model No. 1. The unit
vectors of the crystal coordinate system are given by the vectors
[1/2, 1/2, 1/2] and [001] in the cubic system (Fig. 12.7/1), hence

8= ( l 0)
1'2
1 • (12.7-10)
160 12. General Geometrical Theory of Crystalline Interfaces

The unit vectors of the O-lattice in orthogonal coordinates follow from

the matrix (12.3-12) multiplied by the column vectors of S [Eq. (12.7-10)].

ui°)= [(1/4}(V2 +cotan(0/2), (1/4}(1-V2 cotan(0/2)]

u~O) = [(1/2) cotan (0/2), 1/2]. (12.7-11)

In crystal coordinates, these vectors become:

u~(O)(C) =[(1/2) (1 + (\12/2) cotan (0/2)), - (3V2/8) cotan (0/2)]

u~(O)(C) = [(V2/2) cotan (0/2), (1/2) (1- (V2/2) cotan (0/2)]. (12.7-12)

This result is obtained either by formulating the transformation A in

crystal coordinates
A(C)=S-IRS (12.7-13)
or by transforming the matrix (1 _A-I}-I. The column vectors of the
transformed matrix are now the unit vectors of the O-lattice because, in
the crystal coordinate system, the unit-b(LLvectors are of the form
<1,0). As a third possibility, the orthogonal u~O) can be transformed by
Eq. (12.]-7).
The procedure which was given in this section is useful for relatively
small rotations. For large rotations an additional point of view appears
which will be dealt with in the next section.

12.8. Choice between Various Possible Transformations A

In Section 12.7 the example of rotation on the {11 O}-plane in the
bcc structure was calculated. Fig. 12.8/1 shows a comparison between
the calculated O-lattice and the corresponding moire model (model No.1).
For a small rotation the agreement is good; for large rotations, however,
the moire model shows a configuration which is completely different
from the pattern calculated on the basis of the transformation A=R
[Eq. (12.3-8)]. The reason for this discrepancy lies in the fact that, on
increasing the angle of rotation above a certain value, R no longer
describes the relation between the nearest neighbours in the two lattices.
Visually, however, the nearest neighbours are automatically related in
the moire pattern. Also, from the physical standpoint the relation
between nearest neighbours is of primary importance.
In order to correlate the nearest neighbours for larger rotations the
following vector relations (in coordinates of their respective crystals)
must exist:
;V(I) [1,0] to ;V(2) [1, -1]
and
 Calculated 0 - Lattice
e Dire' -M odel

•t xl1J2 .

10·

x~lI
1 xiO)
GO·
..ft':::"2 x iI)
': .:.':'.' ': . _ _ 1
'::: . . :::
:::,;.; . ,
• XIO)
1

50' 29'

60·

70' 32'

xli)
1
80· ': ,':'

Fig. 12.8/ 1. Comparison between calculated O-lattices and the moire models
No.1. [A : Eq. (12.7-3), A' : Eq. (12.8-9) J
11 Bollmann, Crystal Defects
162 12. General Geometrical Theory of Crystalline Interfaces

First ;r(1) is transformed into ;r(1)' in lattice 1 by the transformation U

and from there into ;r(2) by the rotation R (Fig. 12.8/2).

(12.8-1)

(12.8-2)

Fig. 12.8/2a and b. Point pattern for e = 50° 29' o-lattices for: a) A = rota-
tion, Eq. (12.3-8), b) A' = pure shear, Eq. (12.8-9)

This means that we choose new unit vectors in crystal 1, and starting
from these, we apply the rotation S-1 RS (in crystal coordinates) III
order to arrive at ;r(2).

(12.8-3 )

U is a "unimodular" transformation, i.e. a transformation which leaves

the volume of the unit cell unchanged (det (U) = 1) although it may,
however, change its shape. The orthogonal transformations, and especially
the symmetry operations, which leave the volume as well as the shape
unchanged, form a subgroup of the unimodular transformations and
these, in turn, form a subgroup of the general linear transformations.
 12.8. Choice between Various Possible Transformations A 163

Referred to the orthogonal coordinate system (Fig. 12.8/3):

A' = 8A(C) 8-1 = 8 (8- 1 R8U)8- 1 = R8U8-1 . (12.8-4)

s
--=--

Fig. 12.8/3. Construction of the transformation A' [Eq. {12.8-4)J

A' 8 = R 8 U -+ A' = R 8 U 8-1

The case of the small rotation is described by U = 1. Then, in crystal

coordinates, Eq. (12.8-3) corresponds to (12.7-8) (with 8(2)=8(1)=8)
and in orthogonal coordinates (12.8-4) to (12.3-8).
In crystal coordinates
(1 _A'(C)-I) = (1 -U-1 8- 1 R-1 8) (12.8-5)
and in orthogonal coordinates
(1 -A'-I) = (1 -8U- 1 8-1 R-l). (12.8-6)
Explicitly the matrices are:

A ,(C) = ( cos (j + (V~/2) sin (j cos (j - (~/2) sin (j)

(12.8-7)
- cos (j + (V2/4) sin (j (3 V2/4) sin (j

1 _ A'(C) -1 = ( 1 - (3 V~/4) sin (j cos (j - (V 2 /2) sin (j) (12.8-8)

( ( )) - cos (j + (V2/4) sin (j 1 - cos (j - (V212) sin (j
11·
164 12. General Geometrical Theory of Crystalline Interfaces

and in orthogonal coordinates

I (1!2) cos o+ 0V2!4) sin 0 (V2!2)COSO-(1!2)SinO)
(

A= -(3V2!4)cosO+(1!2)sinO (1!2) cos o+ (V2!2) sin 0 (12.8-9)

and
(I-A'-I) =

1-(1!2)cosO-(V2!2)sinO (V2!2)COSO-(Y2)SinO) (12.8-10)

(
o
-(3V2!4)cos + (1!2) sin 0 1-(1!2)cosO-(3V2!4)sinO'
The determinants, as volume ratios of related unit cells, are independent
of the coordinate system of the transformation. Since A' is a rotation
multiplied by a unimodular transformation, it follows that
det(A') = 1. (12.8-11)
However, the determinant
II-A'-II =2 -cos 0 -(5V2!4)sin 0 (12.8-12)

while the simple rotation:

II-A-II =2(1-cos 0). (12.8-13)

, ,,
,,
2
,
/
2, '
,,
,,
/

/
/
, 1
,, , /

,, , /

/
, /
/
/

.- /

0
1 en1 ;...",
~I .~I ""1
;'1 =1 0..... 1
1
""1 Lrl,

-1
0 10 20 30 40 50 60 70 80 90
Fig. 12.8/4. Determinant II - A-II. 1. Eq. (12.8-12), 2. Eq. (12.8-13)
 12.8. Choice between Various Possible Transformations A 165

Both determinants are indicated in Fig. 12.8/4. The determinant

!I-A-I! is the ratio of the volume of the b-unit cell (in this case the
surface) to the O-lattice unit cell. The smaller this determinant, the
larger becomes the unit cell of the O-lattice and the closer are the cor-
related points in lattices 1 and 2.
The nearest neighbours in both lattices are related by that transformation
A selected from various possibilities for which the determinant !I-A-I!
acquires the smallest absolute value.
From Fig. 12.8/4 we see that for () less than 29° 30', the transformation
A relates the nearest neighbours, and above this value the relation is
given by the transformation A', which contains U. Fig. 12.8/2 shows
that in the two crystals which are of identical structure, different unit
cells have to be chosen in order to relate the nearest neighbours.
Fig. 12.8/5 gives the determinant !I-A'-I! for a rotation on the
{112}-plane. In this case four different transformations are required for
a rotation of more than 90°. This example may be studied on the moire
model No.2 (see also the Table 13.3/T1).
In the general case where two different crystals have to be fitted
together the procedure corresponds to that given in Section 12.7. Starting

Oet II_A-1)

, ,,
,
/ /
... ...
...

,, /
.,.. ...
... ... ' ,

- I
~,
....
•
«>
I

~+---~--.---r---r---r---'---'---r--.

o 10 20 30 40 50 60 70 80 gil
8-
Fig. 12.8/5. Determinant II - A-II for rotation on the {112}-plane
166 12. General Geometrical Theory of Crystalline Interfaces

from an orthogonal unit cell

;:v(I) = 8(1) ;:v(orth) • (12.8-14)
In order to determine ;:v(2) it is, in general, more convenient to first
redefine the coordinate system of the orthogonal unit cell by U e.g.
Eq. (12.8-2) before transforming that into the unit cell of lattice 2.
;:v(2) = R 8(2) U ;:v(orth) • (12.8-15)
Hence, in orthogonal coordinates
A' = R8(2) U8(1)-I. (12.8-16)
For crystals of the same type 8(2) = 8(1) = 8 and A' becomes Eq. (12.8-4).
;:v(2) could also be represented by

;:v(2) =RU' 8(2) ;:v(orth) (12.8-17)

and
A' = RU' 8(2) 8(1)-1 (12.8-18)
but here U' would have the less convenient form

U' _ ( 1/2 V2/2) (12.8-19)

- -3V2/4 1/2
instead of Eq. (12.8-2).

-1
a

U'
~

-1
b

Fig. 12.8/6. Difference in the meaning of V, Eq. (12.8-2), and V', Eq. (12.8-19)
 12.8. Choice between Various Possible Transformations A 167

The difference in the operation of U and U' is shown in Fig. 12.8/6.

Summarizing, the transformation A can be expressed in its general
form either by Eq. (12.8-16) or by (12.8-18).
It is often the case that relations between the two lattices may be
known which do not relate the nearest neighbours, e.g. twin laws are
described as a rotation by 180 around a twin axis or as mirror images
0

across certain planes. We now present a method which allows to

arrive at the transformation A, indicating the nearest neighbour
relations, while starting from an arbitrary point to point relation A
between the two lattices. The calculation is carried out in the crystal
coordinate system.
We first transform a unit vector of lattice 1, e.g. uil ) and arrive at
a lattice vector of lattice 2, i.e. ;r(2L)
;r(2L) =AU~I). (12.8-20)

Then, by another transformation U we transform U~l) into that trans-

lation vector t~l) (of lattice 1) which lies closest to ;r (2L) :
t~l) = U U~I) . (12.8-21)
The transformation which relates t~l) to ;r(2L) is the transformation A
relating nearest neighbours.
;r(2L) = (AU-I) t~I):= A;r(IL) (12.8-22)
and
(1 -A-I) = (1 - UA-I). (12.8-23)
The problem is now to find the transformation U. For this purpose we
determine that unit cell of lattice 1 which contains ;r(2L), i.e. we separate
the external- and internal coordinates of ;r(2) (Section 12.3). We decom-
pose ;r(2L) into to and ;r.
;r(2L) = to +;r. (12.8-24)
(This is the reason for working in the crystal coordinate system.)
Now we have to find that corner of this unit cell which is nearest
to ;r(2L). We subdivide the unit cell into Wigner-Seitz cells (W-S-cells)
(Section 2.2) around every corner and look for that W-S-cell which
contains ;r(2L) .
In order to solve this problem we first have to find the equation of
the W-S-cell wall between two arbitrary points t' and t (Fig. 12.8/7).
Initially, we formulate the problem in orthogonal coordinates. The
projection of the vector ;r onto a vector through the origin parallel to
t' - t must be equal to the projection of the vector t (1/2)(t' - t), i.e.+
( t' - t ) [( t' - t) t' - t ] (12.8-25)
It'-tl .;r = t+-2- · It'-tl
168 12. General Geometrical Theory of Crystalline Interfaces

Fig. 12.8/7. Elements for the calculation of the Wigner-Seitz cell walls

then
(t'-t)· ~-(1/2)(t'+t). (t'-t) =0 (12.8-26)
and
[(t'·~) - (1/2) (t'. t')J - [(t·~) - (1/2) (t . t)J = O. (12.8-27)

This is the equation of the W-S-wall. We abbreviate it to

f(~, t)' -t(~, t) =0 (12.8-28)

where
t (~, t) = (t . ~) - (1/2) (t . t) . (12.8-29)
In general coordinates t (~, t) becomes:
t(~, t) = (tTG~) - (1/2) (tTGt). (12.8-30)
Hence, we can separate the equation of the W-S-wall (placed between
two points t' and t) into two similar expressions, which individually
depend on only one of the points.
If we choose a point ~ not on the Wigner-Seitz wall but on that side
of the wall on which t' lies, then the expression (12.8-28) becomes
positive, i.e. t(~, t') > t(~, t), and vice versa if ~ lies nearer to t.
This leads to the procedure for determining the corner of the unit
cell which is nearest to ~(2L). We have to determine t (~, t) for all the
following t-vectors: [100J, [010J, [11 OJ, [001J, [101J, [011J and [111J,
while for ~ the internal coordinates of ~(2L) are chosen. That t for which
t (~, t) has the largest positive value is the one which is nearest to ~. The
sum of t(l) is given by the t chosen in the above manner together with
 12.9. Solutions of the Basic Equation 169

the external coordinates of ~(2L).

t(1) =to +t. (12.8-31)

If ui1) is ui then t(1) is the first column vector of U (because in the
1 ),

crystal coordinate system, U~l) = [1 OOJ}. The second and the third
column vectors of U are determined in the same way, with U~l) and Uhl ) .
If two of the values I (~, t) are the same, then the point ~(2L) lies
at the same distance from two different points. (This situation corre-
sponds, for example, to the meeting point of the curves in Fig. 12.8/3).
If all the values of I (~, t) are ~ 0, then t = [OOOJ and t(l) = to' If several
t-vectors are closest to ~, t should be chosen such that U is unimodular
(det (U) = 1), as only then the two lattices are in a 1 to 1 relation.
This procedure can be programmed on a computer.
An equivalent procedure is to find the smallest value of the square
of the distance to the corner t:
(t -~)TG(t -~) (12.8-32)
where for t all the eight corners of the unit cell have to be checked.

12.9. Solutions of the Basic Equation

After having formulated the transformation A in Section 12.8 we
now discuss the solutions of the basic equation.
(12.9-1)
If we set
(1 _A-l):= T (12.9-2)
Eq. (12.9-1) may then be written as
T~(O)=b(L) (12.9-3)
or, in index notation:
(12.9-4)
and in the expanded form:
t~ Xl (0) + t~ X2 (0) + t~ = bl (L)
X3(0)

t~ Xl (0) +t~
X2 (0) +t~
X3 (0) = b2 (L) (12.9-5)
t~ Xl (0) +t~ x 2 (0) +fs x 3 (0) = b3 (L).

The rank of the matrix T may be 3,2, 1 or 0 in the (trivial) limiting case.
In the case where the rank (T) = 3 the solutions ~(O) are given by
(12.9-6)
In crystal coordinates the b(LLvectors are of the form <100), and hence,
the column vectors of the matrix T-l are the unit vectors of the O-lattice.
170 12. General Geometrical Theory of Crystalline Interfaces

In the examples of rank(T) <3 the Eqs. (12.9-5) may be solved by

standard linear algebraic methods. Here we present methods which are
specially suited to the solution of our problems involving three, two
or one dimension.
The Eq. (12.9-5) can for given bi(L) be understood as equations of
three planes. In the case of rank (T) = 3 the three normal vectors on
these planes are linearly independent and the planes intersect in a point,
which represents the solution. The normal vectors on the planes are the
row vectors of the matrix T, i.e. (tf, t~, t~). If the equations are formulated
in the crystal coordinate system, then the coordinates of these row
vectors are referred to the reciprocal lattice related to the crystal lattice
(since here the lower indices vary). (In the orthogonal system the crystal
space and the reciprocal space are of the same structure, but of different
dimensions A -+:> A-I.)
In the case of rank (T) = 2 the three planes are parallel to a straight
line, and thus the three normal vectors lie in a plane. For arbitrary
values of bi(L), the three planes would intersect in pairs in three parallel
lines and the three equations would not have a common solution. A
change of bi(L) means a translation of the corresponding plane. In order
to be able to solve Eq. (12.9-5) the three planes must intersect in one
line. In this manner, if two b-coordinates are given the third is also fixed.
Assuming that bI and b2 are given, b3 has to be of the form

(12.9-7)
or
(12.9-8)

For the moment we consider all bi as continuous variables. Eq. (12.9-8)

represents a plane in the b-space passing through the origin, and we term
this b-plane a two-dimensional b-subspace. This means that for
rank(T) =2, solutions to Eq. (12.9-5) can only be found for b-Iattice
points which lie in that b-subspace.
The solutions, :£(0) of Eq. (12.9-5) are no longer single points, but are
now all the points on the common line of intersection of the three planes.
For every b(L) in the two-dimensional b-subspace there exists a straight
x(OLIine (or shorter O-line). We already came across such a situation in
the interpretation of Frank's formula (Section 11.2), where the b-sub-
space was the plane b3 = 0 and the O-lines were straight lines parallel
to the z-axis of rotation.
On solving Eq. (12.9-5) we not only have to determine the O-lines
but also the normal to the b-subspace c;, [Eq. (12.9-8)]. First we look for
a vector through the origin parallel to the O-lines. Once this vector is
known, the positions of the O-lines are determined by the values of b(L).
 12.9. Solutions of the Basic Equation 171

As was mentioned above, the O-lines are the lines of intersection of

three planes. The lines of intersection are perpendicular to the three
normal vectors [Eq. (12.9-5), row vectors (tL t~, t~) (i=1, 2, 3)J. A vector
",(0) parallel to the O-lines must result from the vector product of two
of the row vectors
(12.9-9)

(i) and (k) indicate the fixed indices. t(;) and t(k) are, as already mentioned,
vectors in the reciprocal lattice 1 and v(O) is a vector in the crystal
coordinate system of lattice 1. The vector product procedure [Eq.
(12.9-9)J may be verified by formulating the equations of the planes, e.g.:

t~ vl + t~ v + t~ v
2 3 = 0
(12.9-10)
t~ vl + t~ v +t~ v
2 3 = 0.

It is then assumed that v 3 = 1, and vl and v 2 may be calculated. It turns

out that the components are proportional to the vector product of the
row vectors tr) .
The next task is to determine the normal to the b-subspace. For this
purpose we consider Eq. (12.9-5) as a general relation between the x- and
the b-space. For the case of rank (T) = 2, the three-dimensional x-space
is imaged onto the two-dimensional b-subspace, and consequently the
image of every vector ;£, and in particular the images of the ;£-unit
vectors ;£=<100> must lie in the b-subspace. This means that the
column vectors of T lie in the b-subspace and hence their vector product
forms the normal vector (cl C2 c3 l, the coordinates of which are specified
in the reciprocal coordinate system.

(12.9-11)

Summarizing, we obtain for rank (T) = 2 the orientation of the O-lines

(vector 1,(0)) from the vector product of two row vectors of the T-matrix
and the normal on the b-subspace, i.e. the coefficients of the equation
for the b-subspace [Eq. (12.9-8)J from the vector product of two column
vectors of the T-matrix.
The vector v and c can be normalized to unit vectors:

I v I 2 =(V T GV) (12.9-12)

Icl2 = (CTG-l c). (12.9-13)

In order to determine the exact position of the O-lines we require

additional particular solutions of Eq. (12.9-5). We obtain these, for
example, by setting x 3 = 0 and calculating Xl (P) and x 2 (P) (P= particular)
172 12. General Geometrical Theory of Crystalline Interfaces

from the first two equations

I(P) _ bl(L) t~ _b 2 (L) t~
X - t11 t22 -
t12 t 12

2(P) _ b 2 (L) tl-bl(L) ti (12.9-14)

x - tl t2 _ tl t2
1 2 2 1
x 3 (P) =0.

The bi(L) have to be the coordinates of lattice points of the b-lattice

and they have to lie in the b-subspace, i.e. they must satisfy Eq. (12.9-8).
Then the O-lines become:
Xl (0) = A VI +
Xl (P)

X2(0) = A v 2 +
x 2 (P) (12.9-15)
x 3 (0) =A v3 +x3 (P)

with A as a continuous parameter which determines the coordinates of

a point on an O-line.
In the case of an O-line lattice (rank (T) = 2) a special representation
of the situation perpendicular to the O-lines will be of advantage. Here
every O-line appears as a point, and the essential features involved in
the choice of a boundary (such as line density, cell structure, etc.)
become directly visible. We call this the standard form.
In order to find the standard form we start from the basic equation
(12.9-16)
The O-line v(O) [Eq. (12.9-9)J and the normal to the b-subspace c
[Eq. (12.9-10)J are assumed to have been already determined and
normalized. We now apply a rotation R such that c falls on the b3-axis
(the z-axis, respectively) [Appendix Ai, Eq. (A1-63ff.)J
(12.9-17)
It is useful to formulate (12.9-16) in orthogonal coordinates since R then
has the form (A1-60).
Now the b-subspace lies in the (orthogonal) bl b2-plane. The O-lines
were also rotated by this operation but they are still arbitrarily oriented
in the x-space. They can even be parallel to the (xy)-plane [the (bl b2 )-
plane respectively]. The orientation of the rotated O-lines is then given by:
(12.9-18)
N ow we determine a second rotation Q which brings v' (0) onto the z-axis
Qv'(O) = v" (0) (12.9-19)
l.e.
v" (0) = [0, 0, v~'(O)J.
 12.9. Solutions of the Basic Equation 173

Next we set
(12.9-20)
hence:
;r(0) = Q-l ;r' (0) . (12.9-21)
Along with Eq. (12.9-17), this gives
(RTQ-l) ;r'(0) =b'(L). (12.9-22)
This equation states that every point in the (b 1 b2 )-plane has as its
image an x-line parallel to the z-axis. The representation of the O-lines
is independent of x 3 • We can now reduce the system of equations from
three to two dimensions by simply ignoring the third component.

1:-_:_llL:1 1:1:
I
=
(12.9-23)

Thus, the matrix takes the form

(RTQ-l)i =t'1 (i, k=1, 2) (12.9-24)
and the Eq. (12.9-22) becomes

(12.9-25)
The solutions are
;r' (0) = T'-1 b' (L). (12.9-26)
Here we have obtained the standard form for rank (T) = 2 which re-
presents the O-lattice in the projection perpendicular to the O-lines. The
orientation of the O-lines in the orthogonal, or in the crystal coordinate
system is given by v(O) [Eq. (12.9-9)].
In the case of rank (T) = 1, the three planes represented by Eq. (12.9-5)
are parallel. In order that solutions exist, the three planes have to
coincide. Hence, only one b-component can be chosen freely, since the
others are thereby fixed. This means that, in this case, the b-subspace
is a straight line. For every b-lattice point on that line the solution of
Eq. (12.9-5) is an O-plane. All three row vectors of T are parallel and,
as such, are normal vectors of the O-plane. On the other hand, the
column vectors of T are also parallel and lie in the b-subspace, i.e. the
b-line. Each of the three equations of (12.9-5) is an equation of the
O-plane if b(L) is a lattice vector of the b-lattice and lies in the b-line, i.e.
the b-subspace.
By a procedure analogous to that used for rank (T) = 2 we can find
a standard form for rank (T) = 1. Here a normalized row vector which
represents the normal to an O-plane is denoted by v(O), and a normalized
1 74 12. General Geometrical Theory of Crystalline Interfaces

column vector which determines the b-subspace by c. We now rotate by

an amount R the vector c onto the b3-axis, and by an amount Q the
vector v'(O) onto the z-axis. We again obtain an equation of the type
(12.9-22). In this case, however, it represents the relationship between
the points on the b3-axis and the z-position of the planes perpendicular
to the z-axis, i.e. it represents a relationship between two one-dimensional
sets of points. The equations are reduced to one dimension by

(12.9-27)

Hence:
t'g (x' (0))3 = (b' (L))3 (12.9-28)
and the standard form for rank (T) = 1 is:
'(0))3 _ (b' (L))3
(
x - t'i . (12.9-29)
We designate the O-points, O-lines and O-planes as O-elements. An
O-element is the image of one b-lattice point.
The results of this section are presented in Table 12.11/T 1. The
O-elements which are those regions where the two lattices show best
fit, are points, lines, or planes depending on the rank (T). O-planes are,
for example, twin planes. The (trivial) limiting case of rank (T) = 0
signifies that the transformation A is equal to I i.e. we are dealing with
a perfect crystal containing no interface.
In the case of rank (T) = 2 we can interpret the vector v(O), which
fixes the orientation of the O-line, as an invariant eigenvector of the
transformation A. An eigenvector maintains its orientation during the
transformation but may change its length by a factor A.
A;r =A;r. (12.9-30)
If, in addition, it conserves its length we call it an invariant eigenvector
(A = 1), i.e.
(12.9-3 1)
We write
(12.9-3 2)
(i.e. we refer to the end of the transformation ;r(0) instead of to the
initial ;r). Thus:
(12.9-33)
and, along with Eq. (12.9-31) this gives:
(l-A-l) ;r(0) =0 (12.9-34)
which is the homogeneous part of the basic equation.
 12.10. The Behaviour of the O-Lattice upon Translation of Lattice 2 175

In the case of rank (T) = 1 the transformation A contains two linearly

independent invariant eigenvectors.
We have assumed in this section that be b-subspace contains b-Iattice
points. The problems which arise when this assumption is not valid will
be discussed in Section 12.11.

12.10. The Behaviour of the O-Lattice

upon Translation of Lattice 2
If, in one of the moire models, lattice 2 is translated with respect to
lattice 1, the O-lattice also experiences a translation. We shall now derive
the analytical relations between the two translations.
We start again from the principle of the coincidence of elements of
related equivalence classes. Lattice 2 shall be displaced by the vector d(2).
~(l) and ~(J) +
b(L) are elements of the special class C(l). That element
+
of the class C(2) which is related to ~(l) is no longer A~(l) but A~(l) d(2).
Let us imagine for the moment that ~(1) is a lattice point. Since lattice 2
is displaced by d(2) there is no longer a lattice point at the position
A~(l); it is displaced to the position A~(l) + d(2). Thus, the equation
for the coincidence of equivalent points is

(12.10-1)

~(Od) means" displaced ~(O)". From this it follows that:

(12.10-2)
and
~(Od) =A-1 ~(Od) _A-1 d(2) +b(L) (12.10-3)
( l - A - l ) ~(Od) =b(d) (12.10-4)
with
b(d) = b(L) + d(b) = b(L) _ A -1 d(2) (12.10-5)
d(b) is the displacement of the b-lattice. Hence:
d(b) = - A -1 d(2). (12.10-6)
If ~(Od) is decomposed as follows
~(Od) =~(O) +d(O) (12.10-7)
then:
(12.10-8)
and hence:
(12.10-9)
or
(12.10-10)
176 12. General Geometrical Theory of Crystalline Interfaces

This latter relation is valid in any case for rank (T) = 3 and will have to
be rechecked for the other cases (Section 12.11). In summary, if lattice 2
is translated by d(2) , then the b-lattice is displaced by d(b) , with

(12.10-11)

and the O-lattice is displaced by d(O), with

Td(O) =d(b). (12.10-12)

Upon translation of lattice 2 the O-lattice is also translated. The geo-

metrical structure of the O-lattice, however, is conserved [as long as
rank(T) =3J.
We can also describe the situation in another manner by attributing
individual coordinate systems to everyone of the four lattices. The
coordinate system of lattice 1 (and the b-lattice, respectively) and of
lattice 2 are related by the transformation A, and those of the b- and
of the O-lattice are related by the transformation T. If lattice 2 is
translated by a vector d(2) with the coordinates (d1 , d 2, d3 ) in the co-
ordinate system of lattice 2, then the b-lattice is translated by d(b) with
the coordinates (-dl, -d2 , -d3 ) in the b-lattice coordinate system, and,
correspondingly, the O-lattice by d(O) (-d1 , -d 2 , _d3 ) in terms of its
own coordinate system. Apart from the sign, the numerical values of all
the d-coordinates are the same but in different coordinate systems
(Fig. 12.10/1). These relations can be checked with the aid of the moire

b2

b',
-,
,,
iiT ,!...

Fig. 12.10/1. O-lattice translation resulting from the translation of lattice 2

 1 2.11 . Translation of the 0-Lattice for the Case of Rank (T) <3 1 77

models. By introducing the translation we have extended the range of

the transformations A to nonhomogeneous transformations.
Special problems appear for translation in the case of rank (T) < 3.
These will be discussed in the following section.

12.11. Translation of the O-Lattice

for the Case of Rank (T) < 3
In Section 12.10 the relations between a translation of lattice 2 and
the translation of the b- and the O-lattices were derived. They are as
follows:
(12.11-1)
and
(12.11-2)
If rank (T) = 3:
(12.11-3)
Thus, when the O-lattice is a point lattice, it is uniquely determined.
Let us now consider the case of rank(T) =2, where only those
b-lattice points which lie within the b-subspace (i.e. the b-plane) have
O-line images. Hence, we have to split the translation of the b-lattice d(b)
into two parts, one parallel and the other perpendicular to the b-plane.
(12.11-4)
If = 0, then d(b) translates a b-lattice point within the b-plane by
d(b.lJ
d(bID, and accordingly the corresponding O-line is translated by d(O).
This means that the O-line lattice is translated as a whole entity, and the
problem may be treated as a two-dimensional one.
However, the action of d(bJJ is different. In the case where a certain
b-lattice point was originally located within the b-plane and had an
O-line image, then, because of the displacement by d(b.lJ it leaves the
b-plane and the corresponding O-line disappears (Fig. 12.11/1).
As an example we discuss the rotation of a cubic lattice around the
[001]-axis. That part of the b-lattice which lies in the b-subspace is the
square point lattice in the (001)-plane. The interpenetrating lattices
form layers of points in integral x3-positions. The O-lines are parallel to
the x3-axis. Every point on an O-line is a coincidence of equivalent points
in both lattices, and as such they represent possible origins for the trans-
formation A which is here assumed to be the rotation (homogeneous
transformation). If lattice 2 is now translated, for example, by d(2) =
[0,0, 1/4J, then the b-lattice is translated by [0,0, -1/4]' The b-sub-
space (b 3 = 0) no longer contains b-lattice points. On the other hand,
the two crystal lattices, where lattice 2 is now displaced with respect to
12 Bollmann, Crystal Defects
 Calculated B-Latt ice
e Moire-Model
O-Lattice (Enlarged)

a ,
I xiI)
2'

10' +t-

r
b L

60' ,
:t'

C -1 0 1 2

SO'29 \\
\ ~
d -I 0 1 2

~\\
e

\ .
.J t';------.- .........

. .
'

No O-Lines
., .
, ,
, ~-----',-..,...

<,
f

J'~.
. o
,
, ,
,
Fig. 12. 11 / 1. Comparison of the calculated O-lattice-translation with the
moire model NO.1
 12.11. Translation of the O-Lattice for the Case of Rank (T) <3 179

lattice 1, cannot be imaged onto one another by the pure rotation, i.e.
there are no coincidences of equivalent points i.e. no O-lines.
An analogous situation appears in the case of rank (T) = 1. Here the
translation d(b) of the b-Iattice has to be split into one component parallel
to the O-line and another perpendicular to it. The translation of the b-
lattices provokes also here a corresponding translation or disappearance
of the O-lattice.
Now we are in the position to solve the important problem which
arises if the b-subspace contains few b-Iattice points or none at all. In
this case, the homogeneous transformation A could produce an O-lattice
consisting of only a few O-elements. The O-lattice would be "diluted".
On the other hand, the b-vectors between the b-Iattice points (i.e. the
Burgers vectors of the dislocations between the O-elements) would
become very large. In reality, however, a "dense" O-lattice will appear
with elementary b-vectors. For this reason, O-elements will also have to
originate from b-Iattice points in the close surroundings of the b-subspace.
We project these b-Iattice points perpendicularly onto the b-subspace,
where the component b(LIi) determines the position of the O-element,
and the component b(L.l) corresponds to a translation of lattice 2 with
respect to lattice 1. The O-lattice is determined here by a nonhomo-
geneous transformation where the homogeneous part is the same for all
O-elements, while the translation may be different for each O-element.
We consider the case of rank(T) =2 and split b(L) as follows
(12.11-5)
with
b(LIi) = b(L) - (Cb(L)) c (12.11-6)
c is the unit vector normal to the b-plane (= b-subspace)
b(LJ.) = (Cb(L)) c. . (12.11-7)
Now the basic equation becomes
(l-A -1) x(O) = b(L) - (cb) (L)C (12.11-8)
and the translation
d(bJ.) = _b(LJ.) = _ (Cb(L)) C = - A - l d(2). (12.11-9)
The change of sign: d(bJ.) = _b(LJ.) arises because the b-Iattice point is
translated from outside into the b-subspace, which is considered to be
fixed; hence, d(bJ.) points towards the b-plane, while b(LJ.) points away
from the b-plane (Fig. 12.11/2a).
Thus:
(12.11-10)
12·
180 12. General Geometrical Theory of Crystalline Interfaces

b
Fig. 12.11/2a and b. Projection of the b-lattice points onto the b-subspace.
a) Perpendicular projection, b) projection along the O-lines in a general case

It was proposed earlier (Bollmann, 1967, p. 376) that the b-Iattice

point be projected along the O-line. Here, however, lattice 2 would also
have been translated along that direction, thus eventually leading to
physically unstable situations (Fig. 12.11/2b and Section 9.3) especially if
the O-lines were steeply inclined to c. Therefore, it is advantageous to
apply the shortest translation, i.e. the one perpendicular to the b-sub-
space, since the elastic energy is proportional to the square of this
displacement. When A is a rotation, the two cases are identical.
For the case of rank (T) = 1, the unit vector s shall point in the
direction of the b-line (Section 12.9). Then:
b(LID=(sb(L)) s (12.11-11)
and
(12.11-12)
Hence
(12.11-13)
and
(12.11-14)
By this procedure we have supported the stepped b-net of Section 11.6.
In the general case, the b-plane (b-line resp.) in the b-Iattice will be
approximated by a stepped surface (line resp.) and the O-lattice will be
calculated by Eq. (12.11-8) [(12.11-13) resp. J, the standard form may be
obtained by applying (12.9-25) [(12.9-28) resp.J.
Table 12.11/T 1 summarizes the procedures for the solution of the
basic equation.
 ....
!V
....
....
>-l
Table 12.11jT1
~
(f1

Pi"
rt-
Rank(T) O-Elements Determination of the b-Subspace Determination of the Determination of
(nD = n-dimen- O-Elements b-Subspace 0-Elements if b(L) o·
;;
sional) does not lie in the o>+>
b-Subspace rt-
::r
(p

3 O-point (OD) ;r(O) = T-l b(L) b-space (3D) b(L) arbitrary o

2 O-line (1 D) Vector product of row b-plane (2D) Normal to b-plane Normal projection of ~
vectors of T +partic- =vector product of b(L) onto the b-plane ri'
(p
ular solution column vectors of T
8'
...,
O-plane (2D) Normal to O-plane b-line (1 D) Column vector of Normal projection of st
b(L) on b-line (p
= row vectors of T = b-line
T + particular solution D
(f1
(p

0 O-space (3 D) Total x-space b-point (OD) Point b [0, 0, OJ Translation into b-point o>+>
(perfect crystal) (lattices 1 and 2 ::0
coincide perfectly) po
;;
:>;'

-:l
1\
w
....
....00
182 12. General Geometrical Theory of Crystalline Interfaces

12.12. Subdivision of the Crystal Space into Cells

The transformation A only relates the nearest neighbours in the
"region" of an O-element. The different O-elements regions can be
separated by cell walls. If the unit cell of the O-lattice is large compared
to that of the crystal lattice, these cell walls have a physical meaning.
The lines of intersection of the grain- or phase boundary with the cell
walls are the dislocations. Hence, the cell walls have to be carefully
chosen.
We introduce the following criterion for the cell wall: We consider
two O-points on adjacent O-elements as origins, particularly the point
[OOOJ and the point ~(O). The general vector b was defined by Eq. (11.1-2)
as:
(12.12-1)

i.e. as the difference vector of corresponding points in the two lattices.

~(2) is regarded here (somewhat arbitrarily!) as variable ~, i.e.

(12.12-2)

Now we compare the absolute values of these b-vectors, on choosing

the two O-points as origins. ~ determines a point on the cell wall if

(12.12-3)

The imaging relation between the b-space and the crystal space is
given by
b(~) = (l-A-l) ~:= T~ (12.12-4)
and
(12.12-5)

Under the condition defined by (12.12-4), we can interpret Eq. (12.12-3)

as an imaging of Wigner-Seitz cells (Section 2.2) of the b-Iattice onto the
crystal lattice.
In arbitrary (crystal) coordinates:

(12.12-6)

bT = transposed vector of b (line vector),

G = metric tensor.
The condition (12.12-3) becomes:

(12.12-7)
In matrix form:
(12.12-8)
 Chapter 12. Summary and Discussion 183

with
(b - b(L)f = bT _ b(L)T. (12.12-9)
From (12.12-8) follows that
bTGb -bTGb +b(L)TGb +bTGb(L) _b(L)TGb(L) =0. (12.12-10)
The first two terms cancel one another. Since all the terms are scalars
they are equal to their transpose. Because (AB)T =BT AT:
bTGb(L) = [bTGb(L)V =b(L)TGTb =b(L)TGb. (12.12-11)

The last step is possible because G is always symmetrical and consequent-

ly G T =G. Now Eq. (12.12-10) becomes:
b(L)TGb - (1/2) b(L)T Gb(L) =0 (12.12-12)
or, using Eq. (12.12-4):
(b(L)TGT) ~ - (1/2) (b(L)TGb(L)) =0. (12.12-13)
This is the general equation of a cell wall between two O-elements. The
cell wall bisects (but does not have to be perpendicular to) the connection
between the O-elements [~= (1 /2) ~(O) fulfills the equation J. [See also
Eqs. (12.8-25) to (12.8-30).J
Fig. 13.3/3 a presents an example of cell walls calculated for the case
of a {11 O}-plane in the bcc structure for a rotation of 38° 56.6'. We have
chosen ~(2) as variable ~, but we could equally have chosen ~(1) or some
point between ~(2) and ~(1). In general, the orientation of the cell wall
depends on that choice.

Chapter 12. Summary and Discussion

The geometrical theory of general crystalline interfaces consists
essentially of two parts. The first part is the determination of the
O-lattice, its behaviour on translation of one crystal with respect to the
other, and the calculation of the cell structure. The structure of the
O-lattice is located within the two interpenetrating translation lattices.
The second part of the theory concerns the problems connected with
the choice of the crystal boundary. The boundary is placed in the
O-lattice, passing as much as possible through O-elements, i.e. through
the locations of best fit. Then crystal 1 is occupied by atoms on one
side of the boundary and crystal 2 by atoms on the other side. Only
at the boundary itself do some remnants of the O-lattice structure exist.
The O-lattice is a general three-dimensional description of all the possible
two-dimensional boundaries.
In this section, only the first part, i.e. the construction of the O-lattice,
is summarized. The essential steps are as follows:
184 12. General Geometrical Theory of Crystalline Interfaces

1. Both crystal lattices are related by a linear transformation such

that every lattice point of lattice 2 is the image of a lattice point of
lattice 1 (Sections 12.2 and 12.3)

2. The O-elements are given as solutions of the basic equation

where the b(L)-vectors are translation vectors of lattice 1, and as such

are lattice vectors of the b-lattice (Section 12.3-12.5).
3. In the proximity of the origin, the transformation A must relate
the nearest neighbours in the two lattices. If there is a choice between
several possible transformations then that transformation which gives
the smallest absolute value of the determinant II -A -11 is the one which
has the greatest physical significance. The formulation of the trans-
formation A is given in Sections 12.7 and 12.8.
4. Depending on the rank of the matrix (I -A -1) : = T, the solutions
for x(O) maybe points [rank (T) = 3], lines [rank (T) = 2], or planes
[rank(T) =1]. For rank(T) =3 images exist for all the lattice points of
the b-lattice, while for rank (T) = 2 and rank (T) = 1 only those points
within a b-plane and within a b-line, respectively, are imaged (Sec-
tion 12.9).
5. A translation of lattice 2 by d(2) in terms of the coordinates
of lattice 2 involves a translation of the b-lattice by _d(b), where _d(b)
has the same coordinate values in terms of the coordinates oftheb-lattice
(apart from the sign) as d(2) in lattice 2. _d(b) induces a displacement of
the O-lattice by _d(O), the coordinates of which, in the coordinate system
of the O-lattice, are again numerically identical to the others in their
respective coordinate system (Section 12.10).
6. If the b-subspace is only sparsely occupied by b-lattice points, it
can be approximated by a stepped b-net. This also includes the b-lattice
points in the close proximity of the b-subspace. This means that, for the
transformation A, a translation is added to the homogeneous trans-
formation (Section 12.11).
7. In order to determine the cell walls between the O-elements, it
was stipulated that, at the cell wall

Ib (x)12 = Ib (x _x(O))12
which leads to Eq. (12.12-13) and is equivalent to an imaging of Wigner-
Seitz walls of the b-lattice onto the O-lattice (Section 12.12).
 Chapter 12. Summary and Discussion 185

The equation for the cell walls is plausible but not completely sure.
Since the intersections of the interface with the cell walls represent the
dislocations (as was already shown in Section 11.2) it is possible that
elastic anisotropy has to be considered. Here we have chosen X(2) as
variable x. Once the choice of X(1), or a point between X(2) and X(1) is
made, the orientation of the cell wall may vary, but it must always
bisect neighbouring O-elements.
For given structures and orientations of the two crystal lattices, all
the elements of the O-lattice are determined with the aid of the indications
given in this chapter. The applications are dealt with in the in following
chapter.
Suggestions for Problems
There is no lack of possible problems in this chapter. The calculations
of Sections 12.7 and 12.8 can be carried out explicitly on given examples,
such as moire models Nos. 1, 2, and 3. The transformations which relate
nearest neighbours can be determined. The calculated O-lattices can
be quantitatively compared with the moire models. Furthermore, the
examples of rotation of moire model No.1 ({ 11 O}-plane, bcc structure)
by 50° 29' and 70° 32' can be reduced to the standard form. Similar
possibilities exist with moire model No.2, for () =34° 03' and 62° 58'.
(The necessary indications are to be found in Table 13· 3fT 1.)
In addition, with respect to the translation and the cell structure, a
sufficient number of examples can be found in connection with the moire
models (Sections 12.10 to 12.12): From the behaviour upon translation
by () = 60° of moire model No.1 ({ 11 O} bcc), for example, it can be shown
that the coordinate system of the O-lattice is left-handed (I TI < 0).
However, care must be taken when using the moire models, since
translation operations carried out by means of sheet a (lattice 1) cor-
respond to d(1) =d(b) instead of d(2) (d(1) = -A -1 d(2)). Rotation, on the
other hand, is done with sheet b (lattice 2).
 13. Applications of the O-Lattice Theory

13.1. Outline
Up till now the two adjoining crystals have been idealized as two
interpenetrating point lattices. The O-lattice with its cell structure
introduces an ordering into the interpenetrating lattices. The next step
is the choice of the boundary. Once the boundary is chosen, the atoms
will occupy, on the one side of the boundary the positions of crystal 1,
and on the other side those of crystal 2. The boundary itself is a remnant
of the O-lattice structure.
It shall be emphasized once again that the O-lattice is a description
of those positions where the two crystals fit in an optimum manner. The
cell walls, on the other hand, are the positions of worst fit, i.e. the posi-
tions where the correlation between nearest lattice points in the two
lattices changes between two O-elements (origins!).
A crystal boundary will be placed, as far as possible, through O-ele-
ments. However, not all the possible boundaries are equivalent, since
the energy might vary. The problems related to the boundary energy
will have to be discussed.
Often the problem is not only to find some boundary for some given
relative orientation of the two crystals, but also to first determine the
crystal orientation such that the best possible boundary can be drawn.
Here, neither the relative orientation of the crystals nor that of the
boundary is known at the beginning. We require a criterion for the
optimum boundary, which is basically the criterion for the minimum
boundary energy. In most cases it is not possible to calculate accurately
enough the boundary energy. It will be shown that in certain cases an
energy minimum can be determined without knowledge of the value of
the energy.
We distinguish roughly two ranges of problems according to the
dimensions of the O-lattice unit cell compared to those of the crystal
lattice.

1. The unit cell of the O-lattice IS large compared to that of the

crystal lattice, and

2. the O-lattice unit cell and the crystal lattice unit cell are of similar
dimensions.
 13.1. Outline 187

In the first case the two crystals deviate only slightly from one an-
other. The crystal boundaries are formed by dislocation networks where
the intersections of the boundary with the O-lattice cell walls (Sec-
tion 12.12) are the dislocation lines. On choosing the crystal boundary,
a surface, which may eventually be stepped, is traced in the b-lattice;
this surface contains all those b-lattice points which correspond to the
O-elements intersected by the boundary. These selected b-lattice points

Fig. 13.1/1 . Features in a grain boundary (stainless steel) as expected from

the O-lattice theory. (Photo : G. Dupouy, F . Perrier, Toulouse)
188 13. Applications of the O-Lattice Theory

form the b-net of the dislocation network in the boundary. The Burgers
vectors are attributed to the dislocation lines by the duality relations
of Sections 9.1 and 11.2.
No optimum boundaries exist for subgrain boundaries, i.e. low angle
boundaries between identical crystals. Here the optimum state is the
perfect crystal, since the minimum energy of a subgrain boundary is at
() =0 (Section 11.3). However, optimum dislocation networks are
possible in phase boundaries. An example will be discussed in Section 13 .2.
The second example of when the dimensions of the O-lattice- and
the crystal unit cells are similar, is the case of high angle- or the general
phase-boundaries. Here the conventional notion of the dislocation
loses its meaning since, if such a boundary is considered as a dislocation
network, the dislocation cores would overlap. Nevertheless, optimum
boundaries are also possible, but the emphasis has to be placed on the
pattern of the lattice points: whether this pattern is periodic or not, and
in the former case whether the period is large or small. It is to be ex-
pected that, for a small period, small groups of atoms come together
and thus give rise to an energy minimum in the sense that any deviation
from that boundary orientation, as well as from the relative orientation
of the crystals, causes an increase in the boundary energy. The problems
arising in connection with the pattern of lattice points will be treated
in Section 13.3.
Fig. 13.1/1 shows an electron micrograph of a grain boundary in
stainless steel taken with the 1.5 MeV microscope at Toulouse. The
boundary shows the type of features as are excepted from the O-lattice
theory. Further electron microscope observations have been reported
and discussed by: Gleiter (1969), Ishida et al. (1969) and Levy (1969).
Calculations on grain boundary energy were done by Fletcher (1967).

13.2. Example of an Optimum Phase Boundary

In this section we present an application of the O-lattice theory
(Bollmann and Nissen, 1968), concerning the solution of a special
mineralogical problem which, as such, should belong in a mineralogy text
book. However, the exercise was carried out in order to test the O-lattice
theory, because the situation was very general without any evident
simplifications, and because measured data were available.
The problem concerns an exsolved alkali feldspar (Fig. 13.2/1), a
lamellar two-phase system consisting of a monoclinic and a tric1inic
phase. The tric1inic phase is twinned, which is evident from the vertical
banding of that phase. The thickness of the lamellae is about 1,000 A.
The mineral acts as a natural optical grating which produces irridescent
 13 .2 . Example of an Optimum Phase Boundary 189

2 OOO'&'
Fig. 13.2/ 1. Cryptoperthite (feldspar, moonstone) m monoclinic (homogeneous)
t triclinic (twinned)

colours. For this reason it is often used as a gem stone under the name
" moonstone" .
Chemically speaking, this moonstone is essentially (K, Na) (AI, Sils0 s
and its structure is given in Fig. 13.2/2. Because of the fact that in the

-!SA Of( 00
O!,i---,!,.---!;i-3-!;-i---C'l!;-i oSiAl
Fig. 13.2/ 2. The structure of feldspar
190 13. Applications of the O-Lattice Theory

composite crystal the lattice constants could not be determined with

sufficient accuracy, the values of the two most likely pure phases were
used. They are given in Table 13.2/T1.

Table 13.2/T 1
Structure la (A)
Lattice 1 8.561 (6) 12.996(2) 7.193(4) 90° 116° l' 90°
monoclinic
("Ortho-
clase" Or 92)
Lattice 2 8.135(3) 12.788(3) 7.154(2) 94° 14' 116° 31' 87° 43'
triclinic
(Low albite
Kodarma)

First of all, an attempt was made to fit the monoclinic and the
triclinic phases together. The monoclinic phase was taken as lattice 1.
The transformation A was formulated according to Section 12.7. In
order to determine the matrices 8(1) and 8(2) the a-axis was placed along
the x-axis and the c-axis in the (xz)-plane. (For the procedure see
Eqs. (A 1-17) to (A1-26).]
For the determination of the optimum boundary, a rotation R of
only a few degrees is required, since the two phases are very similar.
The complete rotation can be expressed by three separate rotations
around the X-, y- and z-axes. As long as the rotations are small enough,
they can be superposed as polar vectors, i.e.

(13. 2-1)

Thus, in terms of orthogonal coordinates:

A = R.RyRX 8(2) 8(1)-1. (13.2-2)
First the determinant II-A-11 =ITI was calculated as a function of the
angles ¢x, ¢y, ¢., and it proved to be indefinite (Fig. 13.2/3). In the space
with the angles of rotation as coordinates the surface ITI = 0 resembles
a single shell hyperboloid. Inside the hyperboloid ITI > 0, and outside
ITI<O.
No optimum is apparent in this representation. For all the orienta-
tions ¢i on the surface ITI = 0, the O-lattice is a line lattice, but also
here the line density can strongly vary.
We have to remember that the determinant ITI is a measure of the
volume of the O-lattice unit cell, while what we require is a surface
measure in order to be able to determine the optimum boundary. As has
 13.2. Example of an Optimum Phase Boundary 191

-1
3 !px

Fig. 13.2/3. Adaptation: monoclinic-triclinic II - A-II = 0 as a function of the

rotation <I> of lattice 2 (units: degrees)

already been mentioned, the energy is not known. For energy calculations
to be carried out, values of material constants would be needed, the
accuracy of which would be doubtful. It is preferable to adhere strictly
to the geometrical approach. From the considerations of subgrain
boundaries, we know that the energy first increases with increasing
dislocation density (Section 11.7). Therefore, we replace the energy
function by a purely geometrical parameter, which we can assume to
vary monotonically with the energy in the range of interest. It does not
necessarily have to be proportional to the energy. We only require that
it increases when the energy increases, and vice versa.
For a series of parallel dislocations of spacing d and with an absolute
value of the Burgers vector b, we replace the energy by the parameter

(13·2-3)

If the boundary is formed by a dislocation network consisting of two

independent series of dislocations, then:

(13.2-4)

This function has a positive definite value, and remains continuous even
when one of the dislocation series disappears, i.e. when one of the d;
becomes infinite.
192 13. Applications of the O-Lattice Theory

This geometrical parameter is determined in the following manner:

In the case of rank (T) = 3 the O-lattice is a point lattice. We assume as
possible boundaries the three faces of the O-lattice unit cell (parallele-
piped). These boundaries pass through the O-points, and their inter-
sections with the cell walls are the dislocations. In order to simplify the
problem, we assume the cell walls to be parallel to the faces of the
O-lattice unit cell (Fig. 13.2/4). The geometrical parameter 1l is speci-
fied as:
(13· 2-5)

xlO)
-3

Fig. 13.2/4. Elements required for the determination of the geometrical

parameter [Eq. (13.2-8)J

~ and Fa follow from the cyclic permutation of the indices. d2 is given by:
d2 - _I ~2 (0) m~Ol
X Im~Oll
1-__
m~Ol
1_1 (0)
~2 X~3
(0)1
(13·2-6)
and:
Ib~L)12 Ib~L)12 'Im~OlI2
(d 2)2 = Im~Ol xmls°ll2 (13· 2-7)
and together with Eq. (13.2-5):
p. _ Ib~L)12Im~OlI2 + Ib~L)12Im~OlI2
1- Im~Ol x mls0ll2
(13·2-8)
and, correspondingly Fa and Fa.
 13.2. Example of an Optimum Phase Boundary 193

The orientation of the crystal lattices (<Px <Py <Pz), for which the smallest
of the above parameters exhibited a minimum, was understood as being
that orientation for which the O-lattice furnishes the optimum boundary.
The orientation of the boundary was given by the two x~O)-vectors which
were contained in the expression defining that parameter. ~ proved to
be the smallest of the parameters, and its value is given as a function
of the angle <Pi in Fig. 13.2/5. The surprising result is that this parameter

~z
3

2
, ,,
,,
,,
-1 ~~ ______ ~-Y

-1 3
!Px
Fig. 1 3.2/5. Geometrical parameter P 1 for the monoc1inic-tric1inic adaptation
(coordinates as in Fig. 13.2/3)
exhibits two minima. The value of ~ at the minima is about 26 units.
The function is continuous on the surface ITI = 0, and the minima lie
very close to that surface, i.e. the O-lattice of the optimum boundary
is practically a line lattice. The orientations of the crystals and of the
boundary determined above are completely different from the measured
values.
It is to be noted that the triclinic phase is twinned. According to
F. Laves this twinning is an averaging process which takes place in the
triclinic phase in order to achieve a pseudo-monoclinic character. The
twinning of the triclinic phase can be explained by one of two different
twin laws, the "Albite law" (twin axis b*) and the "Pericline law"
(twin axis b). (The details are found in the indicated article.) Accordingly
the matrix 8(2) was averaged, so that now two monoclinic phases had
to be fitted together.
t3 Bollmann, Crystal Defects
194 13. Applications of the O-Lattice Theory

Fig. 13.2/6 shows the configuration corresponding to Figs. 13.2/3 and

13.2/5, for the fitting of the two monoclinic phases. Here, the deter-
minant ITI has a negative definite value and its absolute value shows a
distinct minimum, as in the case of ~. Both minima coincide in the

-1 +-Y--r--~-Y
-1

Fig. 13.2/6. a) II _A-II and b)P1 for the monoclinic-monoclinic

adaptation

<f;-space and, in addition, the value of ~ is 3 units, i.e. about eight times
smaller than in the case of Fig. 13.2/3 which indicates that here also
the boundary energy is markedly smaller than in the first case. Fig. 13.7
shows the calculated optimum boundary. The calculated dislocation
network might well represent the period in the boundary, but not the
details. The dislocations will split into partials since a Burgers vector
of 13 Awill probably not be stable. In addition, the boundary is disturbed
by the twinning but is nevertheless stable. The comparison between the
c·alculated and the measured values shows good agreement (Fig. 13.2/8).
The differences in the averaging process carried out according to the two
twin laws do not appear to be significant. However, in order to reduce
the twin boundary surface to a minimum, the Albite twinning law is
preferred in this case.
The calculation brings to light another property of the optimum
boundary. The optimum transformation A (<f;~opt)) has three eigenvectors
wjth the corresponding eigenvalues Ai. The optimum boundary contains
those two eigenvectors, with the eigenvalues Ai, which show the smallest
deviation from 1. In the boundary, the crystal experiences no rotation
only extension, and this extension is minimal.
The result of the calculation is surprising if we consider under what
assumptions it was obtained. The only numerical data used was the two
sets of six lattice constants each of the two crystals (Table 13.2/T1); no
 13.2. Example of an Optimum Phase Boundary 195

0- Lattice (pericline average) b -lattice

Phase boundary
Lattice (1 )'\attice (2)
b3~
Front view 1 )---t1 12, -+--,f--+--+-~
1 1 1
1 I

I
1-
I I I I i
I I -0--0--0--0--0-
I I I I I
Top view I I I I I
11 I I I
I I I I 1020
- 0 - - 0 - - 0 - -0 --a-

+-Hu;:J"
t I I I I
1I I I I
I I
I 1

i
I I I •
Q2
I 11 _~o_o~ ~09~~0_0? :001 000
I I I I

250 ~
I-----i

Fig. 13.2/7. o-lattice of the optimum boundary

Fig. 13.2/8A-H. Comparison between the calculated and the measured data.
Relative positions of the unit cells, measured rotation: 55-60'; calculated:
56'. A: measured eigenvector (EV) [3,0, 1J; B: calculated EV [2.56,0,1];
C: measured EV [1,0, 6J; D: calculated EV [1,0, 11.8J; E to F: range of
the measured normals to the phase boundary (6, 0, 1)-(8, 0, 1); G: calculated
normal to the boundary (12.6, 0, 1); H: calculated position of the boundary
[1,0,11.8J
13*
196 13. Applications of the O-Lattice Theory

material constants, no atomic forces, not even atomic positions entered

into the calculation. This means that the result is independent of these
entities, and that the same result would be obtained if the unit cells
were filled with atoms in a completely different manner. The result
shows, in essence-at least in this case-that the best geometrical fit
is equally the most favorable from an energy point of view.

13.3. The Periodicity of the Pattern of Lattice Points

In this section we investigate the possibilities of minimum energy
boundaries in the case of large differences in orientation and structure
of the two crystals. Here the standard concept of the dislocation fails,
but the periodicity of the pattern formed by the lattice points of both
crystal lattices is considered as an essential aspect.
We have defined the o-points as origins of the transformation A
between the two lattices on the one hand, and as coincidences of elements
of related equivalence classes on the other (Section 12.6). The first
aspect dominates for dislocation boundaries. The o-points within the
crystal boundary are the fixed points, and origins for the relaxation
between the two crystal lattices. The relaxation can be imagined to
start simultaneously from all the O-points, until it comes to a halt at the
cell walls (dislocations). It is not essential to know the exact position of
the O-points as long as the O-lattice unit cell is large compared to the
crystal unit cell. It is also immaterial whether the pattern of lattice
points is periodic or not.
The situation is modified, however, if the unit cell of the O-lattice is
comparable in size to that of the crystal lattice. Here it becomes im-
portant to formulate clearly the notion of the pattern of lattice points.
We sqbdivide the complete pattern into pattern elements. A pattern
element is the arrangement of crystal lattice points within one cell of
the O-lattice (bounded by the cell walls). If the complete pattern is
periodic, then it consists of a finite number of different pattern elements;
otherwise it consists of an infinity of them. For a given transformation A
between two given lattices, the pattern element depends only on the
position of the o-point (origin) within the unit cell of lattice 1, i.e. on
the internal coordinates of the O-point. This position is determined by the
related equivalence classes, the elements of which coincide at that
o-point. Thus, the number of different pattern elements is equal to the
number of related equivalence classes containing coinciding elements
i.e. the number of different internal coordinates of O-points.
We shall now develop a computational method for the determination
of the number of different pattern elements. For this purpose we introduce
the notion of the reduced O-lattice, which is understood as a single cell
 13.3. The Periodicity of the Pattern of Lattice Points 197

of lattice 1 in which all the internal coordinates of the O-elements are

plotted. (The reduced O-lattice can contain O-points, O-lines or O-planes,
but here we consider only O-points.) If the pattern is periodic, the reduced
O-lattice contains a finite number of points; otherwise this number is
infinite.
Here again we come across the problem which we already met in
connection with the coincidence of lattice points (Section 12.2). A pattern
may be periodic, but an infinitesimal change in the relative orientation
of the crystals can destroy the periodicity completely, although the
physical change would be expected to be infinitesimal. Periodic and non-
periodic patterns lie as close together as rational and irrational numbers.
The problem of the existence of a coincidence site lattice and of a
periodic pattern are identical if at least the O-point at the primary
origin is a lattice point. However, a pattern can be periodic without
containing a coincidence site lattice, if the primary origin of the trans-
formation was not chosen as a lattice point. The analysis of the periodicity
of the patterns gives a deeper insight into the meaning of the coincidence
site lattice.
The relation between periodic and non-periodic patterns becomes
somewhat more continuous if, instead of an infinite O-lattice, a finite
piece of it is investigated, which, for example, contains 100 O-points.

liz
-~~-~-

0 0 0

0 0

0 0 0

0 0

0 0 0

a li1 b

c d

Fig. 13.3/1 a-c. Examples of reduced O-lattices. a) Rotation {100} (Table

13.3/T1, No. 13); b) Rotation {110} (Table 13.3/T1, No. 31); c) Rotation
{iii} (Table 13.3/T1,No. 67); d) Rotation {112} (Table 13.3/T1, No. 72)
198 13. Applications of the O-Lattice Theory

Its reduced O-lattice can in this case also contain a maximum of 100
points. If, however, the period of the pattern is smaller, then the reduced
O-lattice contains less than 100 points. It is even possible that it consists
of only very few points (Fig. 13.3/1); occasionally even of a single point.
Starting from a highly periodic pattern consisting, for example, of two
pattern elements, and changing the crystal orientation slightly so that
the pattern is no longer periodic, the reduced O-lattice will consist of
100 points, but these will now be arranged in two small groups around
the original two points. However, if the observed area of the O-lattice is
extended to infinity, these small groups of points will extend, fuse and
finally be distributed homogeneously over the whole unit cell of lattice 1,
which houses the reduced O-lattice. If we consider the orientation of the
crystal as variable, and limit the O-lattice to a finite range, then the
fact that the reduced O-lattice consists of small groups of points can be
considered as an indication that the orientation of the crystals is close
to a periodic pattern.
The search for periodic patterns (e.g. such as may be produced by the
moire models) can be performed by means of a computer in the following
manner: First, 10 X 10 O-points are calculated in crystal coordinates,
according to Eq. (12.8-5):

(n, m =1 ... 10) (13·3-1)

where xiO) are the column vectors of the matrix (/- U-lS-IR-lStl . Of
the internal coordinates of the O-points, only the first digits after the
decimal point are extracted; e.g. of the coordinates (3.725, - 2.533), the
internal coordinates are (0.725,0.467). Now a 10 X 10 matrix M(j, K) is
defined in which, at the beginning, all the elements are zero. Then, the
element M (7, 4) is given the value 1. The same procedure is repeated
for all the 100 calculated O-points. If, in the course of the calculation,
M(7, 4) appears again, its value remains 1. The value of an element
only indicates whether it is occupied or not, but not by how many points.
The addition of the values of the matrix elements furnishes the number N
of occupied positions.
10 10
N = z:, z:, M(j, K).
J~1 K~l
(13.3-2)

The situation corresponds to a subdivision of the reduced O-lattice into

100 cells. When only very few positions are occupied, then the pattern
is very nearly periodic. The result of such a calculation is shown in
Fig. 13.3/2.
Not only is the number of pattern elements important, but also their
size. The size of an O-lattice cell is given by 1/I/-A -11 =1/1 TI. Thus, the
 13.3. The Periodicity of the Pattern of Lattice Points 199

8
20 25 , 30 35 I 45
I

M <=> i:o N
"" ;;; ~ on 0
'"
.....
..... ~ ,., ~
i1::l ""
""

~-------2~0------~25------~30~-----3~
5 ----~--------4~5-- 8

Fig. 13.3/ 2. Number of pattern elements N(fJ) for rotation on the {11 O}-pla ne
(compare with Fig. 12.8/4)

t otal period of the pattern becomes

N'
=m
N
' (13-3-3)

N' is the number of crystal units per period of the pattern. The crystal
unit is the volume (the surface resp.) of the unit cell of lattice 1.
The exact values of N, N' and ITI can be determined by means of a
generating function £ in an analogous manner t o that given in Sec-
tion 12.2. E can be generalized as follows:
£' = Iml ~ +m2 u 2 12
(13·3-4)
= [m~I~ 12 +m~ lu212 +2ml m2(~ u 2)]
E = (1/oc) E'. (13·3-5)
Ul , u 2 are the unit vectors of the crystal unit [column vectors of the
S-matrix, Sections 2.2, 12.7, Eq. (A 1-26) J.
~,m2 are the coordinates of the considered point of lattice 1 (according
to Fig. 12.2/1).
oc = 1 or 2 depending on whether E' is odd or even.
200 13. Applications of the O-Lattice Theory

The same conditions as in Section 12.2 are valid, i.e. no common

factors of m1 and m 2 , and the point under investigation shall be nearest
or the second nearest coincidence site to the origin. Fig. 12.2/1 shows the
geometrical situation.
The practical procedure for determining periodic patterns (Table
13.3/T1) and their reduced O-lattices consists of the following steps:
1. In a first computer programme, a table was calculated as a function
of the angle of rotation () (in steps of 0.1°). The table contained the
determinant ITI, the number N of elements =1=0 of the matrix M(j, K),
~~O) and J'~O) in crystal coordinates. Also, the values of N'=N/ITI would
have been useful.
2. In a representation of the type given in Fig. 12.2/1, the value of E,
which in most cases is identical with N', was determined. Here:

() =2 arctan (y/x) (resp. 2 arctan (x/y)) . (13·3-6)

3. From that value in the table which lies closest to the () calculated
by (13.3-6), ITI is taken, and an approximate value of N = E . ITI is
determined. As the table was calculated only in steps of 0.1°, the cal-
culated N might not be an exact integer, but the nearest integral value
will be the correct one. The exact ITI is obtained from N/E=N/N'.
From the coordinates of ~}O), the internal coordinates of the O-points
and hence the reduced O-lattice can be determined.
4. In a second computer programme, on the basis of the exact
()-values, ITI, the complete matrix M(j, K), and the exact coordinates
of the unit vectors of the O-lattice in crystal- and in orthogonal co-
ordinates were calculated. The matrix M(j, K) gave a direct image of
the reduced O-lattice.
The difference between E and N' shows up if ITI =0 (and hence
N' = (0) while E remains finite. It must be stressed that E is a measure
of the coincidence site lattice and N' a measure of the O-lattice. The
coincidence site lattice which represents the period of the pattern is
exclusively determined by the pattern itself, while the O-lattice depends
on the correlation A between the two lattices. All the different O-lattices
of a given configuration contain the same coincidence site lattice
(Fig. 12.8/2). In the case of ITI = 0, a series of coincidence sites lie within
the same O-element. They are not separated by "dislocations". N' makes
this distinction, E does not. Therefore, N' is a more realistic indication
of the energy of a configuration.
If the pattern is represented by different transformations A, then the
total period (the coincidence site lattice) is independent of the actual
value of A. However, the transformation which correlates the nearest
 13.3. The Periodicity of the Pattern of Lattice Points 201

neighbours subdivides the total period into the smallest number of

pattern elements N.
The results of the calculation for rotation on the {1 00 }-, {11 0}-, {111}-
and {112}-plane in the bcc structure are given in Table 13.3/T1. This
table contains the matrices S, G and U, the rotation angle 0, the deter-
minant ITI, the number of pattern elements N [Eq. (13.3-2)J, the number
of crystal units per period of the pattern N' [Eq. (13.3-3)J and the co-
ordinates m I , m 2 [Eq. (13.3-4)]. The unit vectors of the crystal unit in
terms of the orthogonal coordinates u l and u 2 are the column vectors of
the S-matrix. (The length of the cube edge is taken as unit length.) The
table further contains the coordinates of the points of the reduced
O-lattice. A pattern consisting of a single pattern element (coincidence
site lattice) is shown in Fig. 13.3/3.
If a periodic pattern really is a minimum energy pattern, it has to be
one of those with the smallest N'. These can be found in the table.
Fig. 13.3/4 shows examples of periodic patterns from the moire
models.

Table 13.3/T 1

No.
1° I
/I-A-l/
IN
N'
I m Im
i 2
Equivalence classes of o-points
(reduced O-lattice)

Rotation on {lOO}-plane, S = I, G = I N defined in Eq. (13.3-2)

U=I N' defined in Eq. (13.3-3)
ml , m 2 defined in Eq. (13.3-4)
8° 47.8' 0.023529 2 85 13 1 (a): (0,0) (1/2, 1/2)
2 9° 31.6' 0.027931 4 145 12 1 (b): (0,0) (0, 1/2) (1/2,0) (1/2, 1/2)
3 10° 23.2' 0.032787 2 61 11 1 (a)
4 11 ° 25.2' 0.039604 4 101 10 1 (b)
5 12° 40.8' 0.048781 2 41 9 1 (a)
6 14°15.0' 0.061539 4 65 8 1 (b)
7 16°15.6' 0.080000 2 25 7 1 (a)
8 17° 29.2' 0.092486 16 173 13 2 (c): (0,0) (0, 1/4) (0, 2/4) (0, 3/4);
(1/4,0) (1/4,1/4) ... ; (2/4,0) ...
9 18° 55.4' 0.108108 4 37 6 1 (b)
10 20° 36.6' 0.128000 16 125 . 11 2 (c)
11 22° 37.2' 0.153846 2 13 5 1 (a)
12 25° 3.4' 0.188235 16 85 9 2 (c)
13 25° 59.4' 0.202247 18 89 13 3 (d): (0, 0) (0, 2/6) (0, 4/6);
(1/6,1/6) (1/6,3/6) (1/6, 5/6);
(2/6,0) ...
14 28° 4.4' 0.235290 4 17 4 1 (b)
15 30° 30.6' 0.276923 18 65 11 3 (d)
16 31 ° 53.4' 0.301887 16 53 7 2 (c)
17 33° 24.0' 0.330275 36 109 10 3 (c): (0,0) (0, 1/6) (0, 2/6) ...
(0, 5/6); (1/6,0) ... (1/6, 5/6);
(2/6,0) ...
xl 0)
-2
!!l

20 20

10 10

10 10

- 20 20

b - JI~IIIIII_I;~IIIIII:III~IIIIIII~Fllllllli~IIIIIII;)~1111''';10
Fig. 13.3/3. a) Pattern with N=1, {110}-rotation, () = 38° 56.6' (13 .3/T1,
No. 33), cell walls calculated according to Eq. (12.12-13). b) Corresponding
pattern from moire model No.1
 13.3. The Periodicity of the Pattern of Lattice Points 203

20 20

10 10

- 10 10

- 20 20

a
3~1 11'111;~III'II:III~II I'III~I~'I'IIII,I~IIII'I'i~I"1'1 ' ;10

1-

-t
20
L 20
10

· 10 ~ 10
I
!
-20 20

Fig. 13.3/4a and b. Two further examples of periodic patterns on the {112}-
plane. a) Table 13.3/ T 1, No. 81 and b) No. 89
 204 13. Applications of the O-Lattice Theory

Table 13.3/T 1 (Continued)

No.1 e I II-A-11 IN I N'

I ml I m21 Equivalence classes of o-points
(reduced O-lattice)

18 34 ° 12.4' 0.345946 64 185 13 4 (t): (0, 0) (0, 1/8) ... (0, 7/8);
(1/8,0) ... (1/8,7/8); (2/8,0) ...
19 39° 52.2' 0.400000 2 5 3 1 (a)
20 39° 58.0' 0.467153 64 137 11 4 (f)
21 41° 6.8' 0.493151 36 73 8 3 (c)
22 43° 36.2' 0.551724 16 29 5 2 (c)
(V2/2
Rotation on {110}-plane,
U=I
S =
1/2 ~) G=e/
1/2
4 1/~)
23 10° 6.0' 0.031008 1 4 129 1 16 -7 (b)
24 11 ° 32.4' 0.040404 4 99 14 -6 (g) : (0,0) (0, 1/4) (0,2/4) (0, 3/4)
25 13° 26.6' 0.054794 4 73 12 -5 (b)
26 16° 6.0' 0.078431 4 51 10 -4 (g)
27 20° 3.0' 0.121212 4 33 8 -3 (b)
28 22° 50.4' 0.156863 8 51 14 -5 (h) : (0, 0) (0, 4/8) ; (1/4, 3/8)
(1/4, 7/8) ; (2/4, 2/8) (2/4, 6/8);
(3/4, 1/8) (3/4, 1/8)
29 26°31.6' 0.210526 4 1 19 6 1-2 (g)
l! = (
-1
1
~)
30 29° 42.0' I 0.255474 35 137 16 -5 (0,0) (1/35, 1/7) (2/35,2/7) ..
(6/35,6/7); (7/35, 0)
(8/35, 1/7) ...
31 31 ° 35.6' 0.222222 6 27 10 -3 (0, 0) (1/6, 0) (2/6, 0) ...
32 33° 43.2' 0.186916 20 107 14 -4 (0,0); (1/20,2/5) (3/20,1/5)
(5/20, 0); (2/20,4/5)
(4/20, 3/5) ...
33 38° 56.6' 0.111111 1 9 4 -1 (0,0)
34 44° 0.2' 0.052632 3 57 14 -3 (i): (0, 0) (1/3, 0) (2/3, 0)
35 45° 58.6' 0.033898 2 59 10 -2 (0,0) (1/2,0)
36 47° 41.2' 0.019608 3 153 16 -3 (0, 0) (1/3, 0) (2/3, 0)
37 50° 28.6' 0.0 <Xl (11) 6 -1
38 53° 35.6' -0.016260 2 123 14 -2 (0,0) (1/2, 0)
39 55° 52.6' -0.024390 1 41 8 -1 (0,0)
40 58° 59.6' -0.030303 1 33 10 -1 (0,0)
41 61° 1.0' -0.030928 3 97 12 -1 (0, 0) (1/3, 2/3) (2/3, 1/3)
42 62° 26.4' -0.029851 2 67 14 -1 (0,0) (1/2, 0)
43 83° 29.4' -0.028249 5 177 16 -1 (0,0) (1/5,3/5) (2/5,1/5)
(3/5, 4/5) (4/5, 2/5)
44 70° 31.6' 0 <Xl (3) 2 0
45 79° 2.6' 0.074380 9 121 12 +1 (0, 0) (1/9, 7/9) (2/9, 5/9) ... ;
(5/9,8/9) (6/9,6/9) ...
46 80° 37.8' 0.093023 4 43 10 +1 (0, 0) (1/4, 1/2) (2/4, 0)
(3/4, 1/2)
47 82° 56.8' 0.122807 7 57 8 +1 (0,0) (1/7, 5/7) (2/7,3/7)
(3/7,1/7) (4/7,6/7) (5/7,4/7)
I I I I (6/7, 2/7)
 13.3. The Periodicity of the Pattern of Lattice Points 205

Table 13.3/T1 (Continued)

No·1 () II-A-11 I N I N' I ml I m 2 \ Equivalence classes of O-points

(reduced O-Lattice)

48 86° 37.6' 0.176471 3 17 6 +1 (0, 0) (1/3, 1/3) (2/3, 2/3)

49 89° 25.2' 0.222222 22 99 10 +2 (0,0); (1/22, 10/11)
(2/22,9/11) ... (11/22, 0);
(12/22,10/11) ... (21/22, 1/11)

Rotation on {111}-plane,
U=I

50 5° 5.2' 0.007874 127 7 6 (0,0)

51 6° 0.6' 0.010989 91 6 5 (0,0)
52 7° 20.4' 0.016393 61 5 4 (0,0)
53 9° 25.6' 0.027027 1 37 4 3 (0,0)
54 10°25.0' 0.032967 3 91 9 1 (j): (0,0) (1/3, 1/3) (2/3,2/3)
55 10° 59.4' 0.036697 4 109 7 5 (b)
56 11 ° 38.0' 0.041096 3 73 8 1 (j)
57 13° 10.4' 0.052632 1 19 3 2 (0,0)
58 15° 10.6' 0.069767 3 43 6 1 (j)
59 16° 25.6' 0.081633 4 49 5 3 (b)
60 17° 53.8' 0.096774 3 31 5 (j)
61 19° 39.2' 0.116505 12 103 9 2 (k): (0, 0) (0, 3/6); (1/6, 1/6)
(1/6,4/6); (2/6,2/6) (2/6, 5/6);
(3/6, 0)
62 21°47.2' 0.142857 7 2 (0,0)
63 24° 26.0' 0.179104 12 67 7 2 (k)
64 25° 2.4' 0.187970 25 133 9 4 (I): (0,0) (0, 1/5) (0,2/5) (0, 3/5)
(0,4/5); (1/5,0) (1/5, 1/5) .,.
65 0.194245 27 139 10 3 (0, 0) (0, 3/9) (0, 6/9);
(1/9,1/9) (1/9,4/9) (1/9, 7/9);
(2/9, 2/9) ...
66 0.202532 16 79 7 3 (0,0) (0, 1/4) (0, 2/4) (0, 3/4);
(1/4.0) (1/4, 1/4) '"
67 27° 47.8' 0.230769 3 13 1 (j)
68 23° 24.6' 0.257732 25 97 3 (I)
V-
Rotation on {112}-plane, s= ( ~ G= (~ ~/4)
U=I

69 13° 18.4' 1 0.053691 1 48 1 (0,0) (1/8, 1/2) (2/8,0)

(3/8, 1/2) (4/8, 0) ...
70 15° 30.0' 0.072727 (0,0)(1/4,1/2) (2/4,0) (3/4, 1/2)

u= (_~ ~)
71 (0,0) (0,2/12) (0,4/12) ...
17024.4'1-0.091603112113118121
(0,10/12); (1/2, 1/12)
(1/2,3/12) ...
 206 13. Applications of the O-Lattice Theory

Table 13.3/T1 (Continued)

No.1 e
I ml I m21 Equivalence classes of O-points
(reduced O-Lattice)

72 18° 33.0' 1-0.090909 7 I 77 1 1 10 (0,0) (1/7,4/7) (2/7, 1/7)

(3/7, 5/7) (4/7, 2/7) ...
73 19° 50.8' -0.089109 9 101 7 2 (0,0) (0, 1/3) (0,2/3); (1/3, 0)
(1/3,1/3) (1/3,2/3); ...
74 23° 4.4' -0.800000 2 25 1 8 (0, 0) (0, 1/2)
75 27° 31.6' -0.056604 3 53 5 2 (0, 0) (0; 1/3) (0, 2/3)
76 30° 27.0' -0.034483 29 1 6 (0,0)
77 34° 2.8' 0.0 CXl (35) 4 2
78 36° 10.4' 0.024096 2 83 2 10 (0, 0) (0, 1/2)
79 36° 29.0' 0.028016
80 38° 36.8' 0.054545 3 55 7 2 (0,0) (0, 1, 1/3) (0, 2/3)
81 44 ° 25.0' 0.142857 7 4 (0,0)

u=(_~ ~)
82 52° 15.4' -0.193548 6 31 5 4 (0,0) (0, 1/3) (0,2/3); (1/2,0)
(1/2,1/3) (1/2,2/3)
83 57° 7.2' -0.114286 4 35 2 6 (0,0) (0,1/2) (1/2,0) (1/2, 1/2)
84 62° 56.0' O. CXl (33)

u= (_~ ~)
85 66° 6.0' -0.052829 1~3 ~57
86 66° 18.2' -0.056075 6 107 4 10 (0,0) (1/6, 5/6) (2/6,4/6) ...
(5/6),1/6)
87 68° 27.8' -0.088608 7 79 5 12 (0,0) (1/7,6/7) (2/7, 5/7) ...
(6/7, 1/7)
88 72° 35.6' -0.142857 11 77 5 6 (0,0) (1/11, 10/11)
(2/11,9/11) ... (10/11, 1/11)
89 78° 27.8' -0.200000 1 5 2 (0,0)
90 85° 8.4' -0.237288 14 59 2 6 (0,0) (1/14, 13/14)
(2/14,12/14) ...
(13/14,1/14)
91 87° 13.2' -0.242718 25 103 7 12 (0,0) (0, 1/5) (0,2/5) ...
(0,4/5); (1/5,0) (1/5, 1/5) ...
(1/5,4/5); ... (4/5,4/5)

13.4. Displacement of the Pattern

Let us assume a minimum energy pattern as given. The question that
now arises is how the crystal will behave if the relative orientation of
the two lattices is slightly modified. In Section 13.3, it was shown that
the periodicity of the pattern is lost. It may be energetically more favor-
able for the crystal to conserve the minimum energy pattern and to
correct for the deviation by creating dislocations. As we know from
 13.4. Displacement of the Pattern 207

Section 5.4, dislocations limit areas which are displaced by translation

with respect to each other. In order to be able to determine the Burgers
vectors of such correcting dislocations, we have to investigate the trans-
lations of lattice 2 (and consequently those of the O-lattice) through
which the pattern as a whole is conserved.
We assume for convenience that the primary origin is a lattice
coincidence site, and first restate the relations between the different
translations (Sections 12.10):
d(b) = -A-I d(2) (13.4-1)
and
(13.4-2)

A pattern may be conserved in the following different ways:

1. Lattice 2 can be shifted by one of its own translation vectors.
Then the arrangement of lattice points after the translation is the same
as before. The b-lattice, and consequently the O-lattice, are also shifted
by one of their translation vectors. If rank (T) is smaller than the rank
of the problem, then, in the b-subspace, one b-lattice point is replaced
by another, such that also in this case the point configuration remains
unaffected by the translation. The lattice of these displacements is
termed the" pattern repeat lattice" or the DR-lattice (D for displace-
ment and R for repetition). The displacements are given by:

(13.4-3)

(13.4-4)
with T = (1 _A-I) and:
(13.4-5)
2. Another possible way to conserve the pattern consists in the dis-
placement of the primary O-point (which is a coincidence site, and thus
coincides with a lattice point of lattice 1) into another lattice point of
lattice 1. The coincidence site thus appears at a new lattice point of
lattice 1. The pattern as a whole is conserved but is displaced by a
translation vector of lattice 1. We term the lattice of all the displace-
ments of this type as the "elementary pattern shift lattice" or DSE-
lattice, indicated by the shifts d(2-SE), d(b-SE) and d(O-SE). The trans-
lation vectors of lattice 1 are, by definition, the b(L)-vectors. Hence:

(13.4-6)
with
d(b-SE) =T b(L) (13.4-7)
and
d(2-SE) = _Ad(b-SE) = -ATb(L). (13.4-8)
 208 13. Applications of the O-Lattice Theory

3. It was shown in Section 13.3 that the reduced O-lattice contains

all the internal coordinates of the O-points, where every point corresponds
to a distinct pattern element. We may now displace the O-point at the
primary origin into every position, with internal coordinates being
represented in the reduced O-lattice. Such a displacement always leads
to one of the elements of the pattern which has to be conserved, and this,
in turn, leads to the complete pattern. We term the lattice of all such
displacements: the "complete pattern shift lattice" or DSC-Iattice.

0) b) c)

d) (0 sn

x /. :~.~'/I. ~o· '

/,

..... ..... 8 "

........ ·X . ..X ..
. ................. ....... ,

Fig. 13.4/1. a) Pattern {100}-rotation, fJ = 36° 52.2', N = 2 pattern elements

(13.3 /T 1, No. 19); b) arising from translation of pattern (a) by d(O) = <1/2, 0)
[corresponds to points A in d). e) and f) J; c) translation d(O) = <1/ 2, 1/2)
(points B); d). e), f) pattern displacement lattices (enlarged 2.5 x ). Dashed
lines : DR-lattice, full lines: DSE-lattice, dotted lines: DSC-lattice
 13.5. Dislocation Networks in General Crystalline Interfaces 209

Thus, the d(O -scqattice is a configuration consisting of adjoining

reduced O-lattices (see Fig. 12.3/1). It contains, as a superlattice, the
d(O -Rl_ as well as the d(O -sEl-lattice. The d(O -SCLlattice now contains
all the translations of the O-lattice which conserve the pattern as a
whole. It is determined by the procedure of Section 13.3. The other
displacements are:
Td(O-SCl =d(b-SCl (13.4-9)
and
d(2 -SCl = _ A d(b-SCl = _ AT d(O -SC) . (13.4-10)
All the displacements which deviate from the DS C-Iattice modify
the pattern in such a way that the reduced O-lattice is translated within
the unit cell of lattice 1. Since the reduced O-lattice contains all the
internal coordinates 0 ~ Xi < 1 of the O-points, the number of points in
the reduced O-lattice remains constant upon such a translation. This
means that the number of pattern elements, and consequently the
total period, remains unchanged on arbitrary translations although the
point arrangement of the pattern is changed, and the pattern may, for
example, no longer contain coincidence sites. These features can be
observed on the moire models.
Fig. 13.4/1 presents an example of a pattern, together with the at-
tributed DSC-lattice, as well as two patterns created by other trans-
lations. It can be reproduced by the moire model No.3 (0 =36 52') 0

(Fig. 13.4/2).

13.5. Dislocation Networks

in General Crystalline Interfaces
A dislocation network in a high angle grain boundary or a general
crystalline interface will be formed if, for energetic reasons, the crystal
tends to conserve a certain pattern, and if the relative orientation of the
two crystals deviates only slightly from the optimum orientation. In
contrast to the primary dislocation networks of the subgrain boundary
type (intersections of the boundary with the O-lattice cell walls), we
denote the new networks as secondary dislocation networks. In order to
calculate such a secondary dislocation network, an analogous calculation
to that for the primary networks (Sections 12.9 and 12.12) is carried
out. First we determine an "0 2-lattice", then the cell walls between
the new" 0 2-elements" ; this time, however, the 0 2-lattice is an O-lattice
between two O-lattices instead of between two crystal lattices.
The first O-lattice shall have the optimum pattern given by the trans-
formation A:
(13·5-1 )
14 Bollmann, Crystal Defects
210 13. Applications of the O-Lattice Theory

20 20

10 10

- 10 · 10

- 20 · 20

-3~" 'I' ~';I~ "'I' I:',1~ "'1" I~f~ "I"",'~"'I'" i~IIIIII ';io
a
Fig. 13.4/2a-c. Moire pattern corresponding to Fig. 13.4/1

20 20

10 10

- 10 - 10

- 20 -20

-31~ '''1' ~ ;I~ "'I' I:',1~ '''1'' ~f~'1111 ",I~'1111 II ;I~"'I' II ;10
b
 13.5. Dislocation Networks in General Crystalline Interfaces 211

20 20

10 • 10

o
o -+-· ~

- 10 10

20 20

while the other O-lattice shall be that which deviates slightly from the
optimum pattern given by A':
(13·5-2)
x(O) and x' (0) are images of the same b(L)! Eliminating b(L), we obtain:
x' (0) = (T' -1T) x(O) : = Bx(O). (13·5-3)
Here the transformation B between two O-lattices is analogous to the
transformation A between two crystal lattices.
In analogy to the original O-lattice, the new O-lattice is termed the
"O-lattice of second order" or "02-1attice". In order to formulate the
new basic equation we have to know which vectors now correspond to
the original b(L)-vectors. They have to be those translations of the 0-
lattice which conserve the pattern as a whole i.e. the d(O -SC)-vectors.
Thus, the basic equation becomes
(13·5-4)
It is seen that, if A' =A then B =1, and (I-B-l) acquires rank 0, so
that here no secondary dislocation network can exist.
Before we calculate the cell walls explicitely, we have to determine
the Burgers vectors which are attributed to the cell walls. In a primary
14'
212 13. Applications of the O-Lattice Theory

O-lattice these are the differences between the b(L)-vectors, i.e. the
translation vectors of lattice 1, which are identical with the d(b-R)_
vectors. By analogy, the new Burgers vectors are chosen as the d(b-SC)_
vectors, which contain the d(b-R) (= b(L))-vectors as a subset. The
conceptual problems involved with the Burgers vectors in boundaries
(displacement of lattice 2 or lattice 1) will be discussed in Section 14.2.
Thus, considering the d(b-SC)-vectors attributed to the ;1:(02)-vectors,
we obtain, according to Eq. (13.4-2):
Td(O-SC) =d(b-SC)
(13·5-5)
which, together with Eq. (13.5-4) gives:
T (1 -B-1) ;1:(02) = Td(O -SC) =d(b-SC) (13·5-6)
with
(13·5-7)
Hence, the new basic equation is:
T(02) ;1:(02) =d(b-SC). (13·5-8)
Every vector d(O-SC) is related on the one hand to an ;1:(02), and on the
other to d(b-SC), in such a way that, to every cell wall, a Burgers vector
can be attributed according to the duality relations of Section 9.1.
The cell wall between the element ;1:(02) = [OOOJ and a neighbouring
;1:(02)-element is analogous to Eq. (12.12-13)
[d(b-SC)T GT(02)J ;1:-(1/2) [d(b-SC)T Gd(b-SC)J =0. (13.5-9)

G is again the metric tensor of the coordinate system considered.

The Burgers vectors of the secondary dislocations in the boundary
are markedly smaller than those of the primary dislocations in the single
crystal. Therefore, the strain field coupled to these dislocations is also
smaller.
The existence of distinct secondary dislocations compared to wavy
moin~s, is a proof of the existence of a minimum energy pattern. This
proof was provided by Schober and Baluffi (1970). These authors soldered
together gold foils, grown epitaxially on rock-salt, in a specific orienta-
tion. Fig. 13-5/1 shows an electron micrograph of foils in the {1 OO}-plane
rotated by about 36.9°. The secondary dislocations are clearly visible.
Fig. 13.5/2 shows the corresponding calculated network.
Fig. 13.5/3 gives a second example of a calculated network and
Fig. 13.5/4 an electron micrograph of stainless steel, but the orientation
here is not known.
 Fig. 13.5/1. Secondary disloca-
tion network in gold correspond-
ing to Fig. 13.5/1. (Photo:
T. Schober, R. W. Balluffi)

Fig. 13.5/2a-c. Formation of a

secondary dislocation network.
a) Distorted pattern ({100},
0=34° 12.8'). b) Secondary dis-
location network conserving the
minimum energy pattern Fig.
13.4/1. c) Transformation B
[Eq. (13 .5-3) ]. d) Pattern con-
serving displacements of: the
O-lattice 1 and lattice 2

I --+-+-+-+-< C)I I_
.,'

;/r
Ir '--Y-+-+-I- H
 ,/

100 1.0

Fig. 13.5/ 3. Calculated secondary dislocation network, mllllmum energy

pattern Fig. 13.3/3, L, () = -14.6', to the right: d(2-SC)-lattice

Fig. 13.5 /4 . Example of a secondary dislocation network.

(Photo: G . Dupouy, F. Perrier)
 14. Completion of the Linear O-Lattice Theory
and Extension to Non-Linear Problems
14.1. Outline
In this chapter, it will be shown how the definitions of dislocations
and Burgers vectors can be extended to general crystalline interfaces.
The conventional definitions of dislocations in single crystals will appear
as special cases of the more general definitions. It will further be shown
that practically every crystalline interface can be interpreted as a
"dislocation network", and a systematics of these networks shall be
presented. The three grain problem will also be outlined. With this
chapter, the linear geometrical theory of crystalline interfaces is essen-
tially closed.
Section 14.3 presents an extension of the theory to non-linear
problems which arise when a dislocation from the interior of the crystal
ends at a boundary or penetrates it.

14.2. Dislocations and Burgers Vectors

in Crystalline Interfaces
In Section 5.2 the dislocation line was defined by a Volterra cut.
Another possibile definition followed from the movement rule (Sec-
tion 5.4) where the dislocation line was understood as being a boundary
line between a displaced and an undisplaced region.
The definition and sign of the Burgers vector followed accordingly
from the Burgers loop or the displacement during the movement. In
connection with the Burgers loop, it was emphasized that the Burgers
vector has to be conceived as a coordinate difference in the (eventually
elastically distorted) coordinate system of the crystal, in order that its
invariance along the line and its continuity on dislocation branching
are preserved. This fact was already taken account of by Frank by
introduction of the reference lattice. A clear distinction is given in this
way between plastic and elastic deformation.
The dislocations within a crystal boundary are defined as the inter-
sections of the boundary surface with the O-lattice cell walls (Section 13.1).
The Burgers vectors are attributed to the dislocations by means of the
duality relations between the dislocation network and its b-net (Sec-
tion 9.1 and 11.2).
216 14. Completion of the Linear O-Lattice Theory

We shall now investigate how far the original definitions (Burgers

loop, movement rule) remain valid for dislocations in crystal boundaries,
and what aspects of the notion of a dislocation still have a meaning in
relation to general crystalline interfaces. For this purpose we outline
the analytical representation of a Burgers loop.
We start, for example, in crystal 1 in the cell of the O-point a:(0) at
a:(lLS) (5 = start) (Fig. 14.2/1 a) and proceed inside that cell by trans-
lation vector steps to the crystal boundary a:(lL). At the boundary we
jump to the related point in lattice 2, a:(2L), where:

(14.2-1)

a L -_ _~_ _ _ _~_ _ _ _~___~L-_ _~_ _ _ _~_ _ _ _~_ _ _ _L -_ _~

Fig. 14.2/1 a and b. Burgers loop across a high-angle grain boundary.

a) Start and end of the loop in crystal 1, b) in crystal 2

From the point a:(2L) we continue in translation vector steps, now in

lattice 2, into the cell of the neighbouring O-points a:' (0), and from there
we return over the boundary to the related point in lattice 1, which is
now defined with respect to a:' (0) :

(14.2-2)

We then continue inside lattice 1 to the endpoint a:(lLF) (F = finish). We

may use" pro memoria" the reference lattice inside which the loop has
to be closed (Section 5.2).
 14.2. Dislocations and Burgers Vectors in Crystalline Interfaces 217

If the translation vectors in lattice 2 are considered as being equivalent

to those in lattice 1 through
(14.2-3)
it can be shown that every single loop, however complicated it may be,
can be reduced to the following steps (Fig. 14.2/1a):
~(lLS) ~ ~(2L) _----? ~' (2L) ~ ~(lLF). (14.2-4)
In general,
~'(2L)_~(2L)=0 i.e. ~(2L)=~'(2L). (14.2-5)
The exceptions are those cases discussed in Section 12.11, where
~'(2L) _~(2L) = A (b' (LJJ _ b(LJ.)). (14.2-6)
Now, the sequence (14.2-4) will be formulated in detail:
~(2L) _~(O) =A(~(lLS) _~(O)) (14.2-7)
~(2L) _~' (0) = A (~(lLF) _~' (0)). (14.2-8)
When ~(2L) is eliminated, we obtain:
(1 _A-I) (~'(O) _~(O)) =~(lLF) _~(lLS). (14.2-9)
Taken together with the basic Eq. (12.3-6), this gives:
~(l LF) _ ~(l LS) = b' (L) _ b(L) . (14.2-10)
This means that the closure failure of the loop is equal to the difference
of the b-Iattice vectors, the images of which are ~(O) and ~' (0). It should
be noted that, in the disturbed crystal, the Burgers vector has to point
from the beginning to the end of the loop.
If the loop starts in crystal 2, then the steps are (Fig. 14.2/1b):

(14.2-11 )
(Attention has to be paid to the sequence of 0' and 0, in order that the
sense of the loop is preserved!)
The result here is:
A -1 (~(2LF) _ ~(2LS)) = b' (L) _ b(L) . (14.2-12)
The steps in lattice 2 are equivalent to those in lattice 1, insofar as both
lattices are coupled through the transformation A.
In the case of rank (T) < 3, the following holds:
(14.2-13)
which has to be taken into account in the loop steps (14.2-11).
218 14. Completion of the Linear O-Lattice Theory

Fig. 14.2/1 b

It has been shown above that the definition of the Burgers vectors
obtained from the duality relation between the 0- and the b-lattices is
equivalent to the original definition by the Burgers loop, if translation
vectors related by A in both lattices are taken as being" identical".
The kinematic definition of the Burgers vector (movement rule,
Section 5A) can also be extended to general crystalline interfaces. For
the case of a single dislocation in a perfect crystal it was assumed that
the dislocation could be displaced by arbitrary amounts in the direc-
tion v . In the present context, we make restrictions to this assumption.
The dislocation shall be displaced by one O-lattice cell period, such
that an O-element after the displacement takes the position of a neigh-
bouring element. The total pattern before and after the displacement is
identical with respect to structure and position.
This is achieved either by keeping lattice 1 fixed and displacing
lattice 2by one of its minimum translation vectors [d(2-R) Eq. (13.2-3)
to (13 .2-5) J, or vice versa by keeping lattice 2 fixed and shifting lattice 1
by its corresponding translation vector d(l -R) = d(b -R). If m [of the
movement rule Eq. (5A-1)J points into lattice 2, then d(2-R) is the
translation vector. If the direction and size of v are left unchanged
then m can only change its sign, and thus point into crystal 1 (which
now becomes the shifted part) , if band l simultaneously change their
sign. In addition, b (i.e. d(R)) changes its orientation, and eventually its
 14.2. Dislocations and Burgers Vectors in Crystalline Interfaces 219

size, on the transition from lattice 2 into lattice 1, i.e.

(14.2-14)
where
d(l-R) = b' (L) _ b(L) . (14.2-15)

Thus, also here the Burgers vector is the coordinate difference of minimum
translation vectors in both lattices which are coupled through A (mini-
mum, because it was assumed that the O-element is shifted into the
neighbouring position).
After having seen that the notion of the dislocation can be extended
to general crystalline interfaces, these interfaces can also be interpreted
as "dislocation networks". (An exception is the perfect twin boundary
which is formed by an O-plane.) We distinguish on the one hand physical-
and geometrical (or mathematical) dislocations and on the other hand
primary and secondary dislocations.
We understand here as a dislocation line the intersection of the crystal
boundary with a lattice cell wall. It is a physical dislocation if the O-lattice
cell is large compared to the unit cell of the crystal lattice. In this case,
the relaxation which starts from the O-elements acts such that con-
ventional dislocations, with cores and strain fields, are formed at the
intersection of the boundary with the cell walls.
However, if the O-lattice cell is of similar size to the crystal unit cell,
then no physical dislocation can be formed any more, but, by definition,
the intersection of the boundary with a cell wall is also here a "dis-
location" with the invariance and continuity properties of the Burgers
vector (the latter follows from the existence of a connected b-net.) The
" dislocation" is now a geometrical delimitation of the different O-element
regions, and as such it is called a geometrical dislocation.
The distinction between primary and secondary dislocations can be
seen best from the behaviour of the pattern on translation of one crystal
lattice by a moving dislocation. Is the pattern preserved with respect
to structure and position, the translation is caused by a (perfect) primary
dislocation, be it physical or geometrical. If, however, the pattern is
preserved only with respect to its structure but is allowed to change its
position, then the translation is caused by a secondary dislocatioll. As
was shown in Section 13.4, the Burgers vectors of the primary dislocations
form a subset of the Burgers vectors of the secondary dislocations (the
d(b -R) correspond to the primary Burgers vectors and these are included
in the d(b-SC) of the secondary Burgers vectors).
Secondary dislocations are formed if a minimum energy pattern has
to be preserved. Each one of these patterns has its own DSC-Iattice
(Section 13.4) and thus its own set of secondary Burgers vectors. The
 220 14. Completion of the Linear O-Lattice Theory

physics of the crystal will finally determine what secondary dislocations

are formed.
Fig. 14.2/2 presents schematically the ranges of the different types
of dislocation networks. The path of the energy curve corresponds
formally to that of the subgrain boundary (Section 11.7). The primary
dislocations extend over the whole range. With the exception of the

Boundary energy

Relative orientation
Type of dislocations:
of crystals
Geom. primary
, I
Phys. primary
Phys. secondary ,
5
1 ,
5
2
I--X-+--X---t
r--X~ S
, 1 '
3 ,
S
4
I--X--+-X---t
I

(X Periodic substnatum)
Perfect crystal

Fig. 14.2/2. Schematic representation of the ranges of different types of

dislocations

surroundings of the perfect crystal, they have to be considered as geo-

metrical dislocations. If the transformation A is transferred to a new
relation (Section 12.8), then the structure of the primary dislocation
network is also changed. The secondary dislocation networks are at-
tributed to their corresponding minimum energy patterns.
The situation of the primary and the secondary dislocations can also
be characterized in the following way: We assume an arrangement of
two crystals with a boundary. From the standpoint of primary dis-
locations this arrangement corresponds to one crystal containing a
dislocation network; from the standpoint of secondary dislocations,
however, it can be regarded as a sandwich of three" crystals" : crystal 1,
the periodic pattern of the boundary, and crystal 2. The periodic pattern
in the boundary acts as a periodic substrate for its own set of possible
dislocations. It will be shown in the next section that the dislocation
balance is also fulfilled, in most cases, if a dislocation from a crystal
enters or penetrates a boundary.
 14.3. Non-Linear Problems 221

The O-lattice procedure can be extended to three grains as follows:

We assume that three crystals grains are given in known orientations.
Lattices 1 and 2 are then related by
(14.2-16)
(where A(21) relates the nearest neighbours), and lattices 1 and 3 by
;1:(3) = A (31) ;1:(1) •
(14.2-17)
On eliminating ;1:(1) we obtain:

(14.2-18)

A(32)has to be checked and eventually corrected by a unimodular trans-

formation by the procedure given in Section 12.8 in order to again
correlate the nearest neighbours. On the basis of A(32) the O-lattice and
the boundary between crystal 2 and 3 can be determined.

14.3. Non-Linear Problems

Up till now we have regarded lattice 1 and lattice 2 as translation
lattices and developed the O-lattice theory on that basis. Now we shall
extend the procedure to non-linear problems, e.g. to distorted lattices or
lattices containing dislocations. For this purpose we represent the two
"lattices" as functions of a vector parameter z (;1:(1) (z) and ;1:(2) (z)).
We assume the functions to be essentially continuous (curvilinear
coordinates) and the " lattice points" shall be determined by integral
values of z. Here we denote the" period" of both lattices by the integral
vectors b(L). The lattices, partitioned into equivalence classes, are now
represented by the "lattice" of the z-vectors. The condition for the
coincidence of elements of related equivalence classes becomes:

(14·3-1)

From this equation the intermediate solution:

(14·3-2)
is determined. ;1:(0) is then given either by
;1:(0) (b(L)) = ;1:(2) (z (b(L))) (14·3-3)
or by
;1:(0) (b(L)) =;1:(1) (z(b(L)) + b(L)) . (14·3-4)

We can show that this procedure is a generalization of the basic

Eq. (12.3-6).
222 14. Extension to Non-Linear Problems

We set: (z)
X(l) =X(l) (14·3-5)
X(l) (z + biLl) = X(l) + biLl (14·3-6)
and
X(2) (z) = AX(I) (14·3 -7)
which again leads to the basic equation.
An example will now be discussed of a dislocation which ends at a
grain boundary. The boundary is assumed to be the one formed by a
rotation of 36° 52.2' in the cubic {100}-plane. The pattern is given in
Fig. 13.4/1 a with the number of pattern elements N=2 (coincidence
site and "cross"). Its period is N' = 5 crystal units . Again we let two
lattices interpenetrate, one containing an edge dislocation, the other
being perfect.

2I
xl') 0
0

./ ,
Xl21

,
xl 11

o
o

o·

o
o

Fig. 14.3/ 1. Calculated pattern for the case of a dislocation ending at a grain
boundary (original pattern Fig. 13.4/ 1 a)
 14.3. Non-Linear Problems 223

The strain field of an edge dislocation is, according to Section 6.6,

given by:
Ul ( Xl' X 2) = b rarctan (-.1'2 )'
"2'--
:n l Xl
+ -2(11 - ) - V
2
Xl
+ 1
Xl .1'2
2
.1'2
-b [ Xf-X~l (14·3-8)
+ xf +xr
2 2
U 2 (Xl , x 2) = 8:n(1 -v) (1-2v) . In (Xl +X2) .
The crystal containing the edge dislocation (here crystal 2) is represented
by;r +u(;r), where ;r corresponds to the parameter z. Lattice 2 shall be
rotated by R while the dislocation-free lattice 1 determines the co-
ordinate system. In this manner the generalized basic equation becomes:
R(;r+u(;r)) -(;r+b(L)) =0. (14·3-9)
Written in detail the equations are:
r ll [Xl +Ul (Xl x 2)J + r12 [X2 +U2 (X I x 2)J -(Xl + b~L)) = 0,
(14·3-10)
r 21 [Xl +Ul (Xl x 2)J +r22 [X2 + U 2 (X I x 2)J - (X2 +b~L)) = O.
;r(b(L)) and then ;r(O) =;r(b(L)) +b(L) were calculated and the result is
given in Fig. 14.3/1. Here the O-lattice itself contains a "dislocation".
Thus, we have calculated the configuration which appears when a dis-
location branches out of a network and for which in Section 11.9 the
duality relations failed. It is to be stressed that, if the intersection lines
of the crystal boundary with cell walls are strictly understood as "pri-
mary" dislocations, the Burgers vector balance (node condition, Sec-
tion 7.4) is closed over the whole network.

a) b)
t DISLOCATION

W LATTICE 1

c) d) e)

Fig. 14.3/2. Interpretation of Fig. 14.3/1 in terms of secondary dislocations

(Hatching: see Fig. 13.4/1)
224 14. Extension to Non-Linear Problems

The b-lattice can be formally looked on as a cubic lattice, but the

relation between the b- and the O-lattices is no longer given by a linear
transformation.
The situation can also be interpreted from the standpoint of the
secondary dislocations (Fig. 14.3/2): the incoming dislocation is first
split into two secondaries and these again are split into secondary
"partials". This second splitting follows from purely geometrical
considerations. The energy situation will decide for each individual case
whether a dislocation will split or not.
Fig. 14.3/3 shows the case where an edge dislocation penetrates a
boundary. Here both lattices contain an edge dislocation and the basic

o x(ZJ
/1

o o_xl1J
o 0 1

o
o

Fig. 14.3/3. An edge dislocation penetrating the grain boundary

 14.3. Non-Linear Problems 225

equation becomes:
R(~+u(~)) - [~+b(L) +U(~+b(L»)J =0. (14.3-11 )

From the standpoint of the primary dislocations the boundary is a

somewhat deformed square network. No dislocation branches off. The
dislocations entering and leaving the boundary in the two different
crystals are equivalent, and their Burgers vectors (coordinate differ-
ences!) cancel at the boundary although the corresponding translation
vectors in both lattices are rotated with respect to each other by about
37°. From the standpoint of the secondary dislocations, however, a
dislocation branches off in the boundary and appears again as split into
partials (Fig. 14.3/4); hence, from this view point the two crystal dis-
locations are no longer equal.

a) b)
f DISLOCATION

47
~LATTICE 2

I" LATTICE 1

c) e)

Fig. 14.3(4. Interpretation of Fig. 14.3/3 in terms of secondary dislocations

In the case given in Fig. 14.3/3 an edge dislocation penetrates a

perfect "minimum energy boundary" (optimum boundary). The dif-
ference between the outgoing and the incoming translation vectors is
exactly equal to the Burgers vector of a secondary dislocation (Fig.
14.3/4a). If, however, the rotation between the crystal lattices does not
exactly correspond to an optimum boundary, then a network of secondary
dislocations is formed in the boundary (Section 13.5). The secondary
dislocation which branched off will be woven into that network. Now
the difference vector between the corresponding translation vectors of
15 Bollmann, Crystal Defects
226 Chapters 13-14. Summary and Discussion

lattice 2 and lattice 1 will not be an exact Burgers vector of a secondary

dislocation. The nearest secondary - and eventually primary Bur-
gers vector is still a certain difference away. This difference does not
fit into the Burgers vector balance, but it exerts a strain similar to that
of a dislocation with a rest-Burgers vector. Marcinkowski (1970) terms
such a rest-dislocation a virtual grain boundary dislocation (abbreviated
to VGBD). A virtual dislocation also appears in a low angle boundary
when it is penetrated by a dislocation. In such a situation, the minimum
energy pattern is the pattern of the perfect crystal, and the possible
Burgers vectors are its translation vectors. Similar problems have been
discussed by Sleeswyk (1967). He introduced the notion of an unstable
dislocation, which is somehow related to the virtual dislocation. To
summarize, a virtual dislocation can appear if a boundary is penetrated
by a dislocation. In the case where the dislocation ends at the boundary,
the balance is always closed although the b-net is no longer topologically
planar.
Until now we have assumed that an edge dislocation ends at the
boundary or penetrates it. The same can naturally happen with a screw
dislocation. If the dislocation does not penetrate the boundary it con-
tinues inside the boundary as an edge dislocation and produces a step.
Such a step can end up as a spiral whereby the boundary is displaced.
Spirals in boundaries were observed by Gleiter (1969a). Gleiter (1969b)
developed a theory of grain boundary motion based on a model where
the two grains limited by steps are separated by some interlayer. The
grain boundary motion takes place by diffusion of atoms in different
stages through the interlayer.
If on the basis of the O-lattice theory, a model of the grain boundary
motion were developed, the atoms sitting in lattice 1 would have to
shift into a position in lattice 2. It is most probable that an atom would
pass from its position in lattice 1 into the nearest position in lattice 2
(i.e. into that position which is related to position 1 by A, with respect
to the nearest O-point taken as origin). These two positions are usually
less than one atomic diameter apart, so that, on the average, the actual
displacement of the individual atoms would be of the same order of
magnitude. No interlayer would be needed in this description. The atoms
would not have to pass through the boundary, the boundary would pass
through the atoms.

Chapters 13--14. Summary and Discussion

Some applications of the O-lattice theory are given in chapters 13
and 14. Two different cases have to be distinguished:
 Chapters 13-14. Summary and Discussion 227

1. the unit cell of the a-lattice is large compared to that of the

crystal lattice, and
2. the unit cell of the a-lattice and of the crystal lattice are of similar
dimensions.
The first case leads to boundaries which consist of dislocation net-
works of the subgrain boundary type. An important problem is the
determination of "optimum" boundaries where the orientation of the
crystals, as well as the orientation of the boundary, have to be deter-
mined in such a way that the value of the boundary energy attains a
minimum. No optimum subgrain boundaries exist because the "opti-
mum" state is represented by the perfect crystal. It is possible, however,
to find optima for phase boundaries. An example is discussed in Sec-
tion 13.2. If the problem is to locate an energy minimum without expecting
quantitative indications on the energy values, it is sufficient to replace
the energy by a function which can be anticipated to vary monotonically
with the energy.
The dislocation lines can be defined as the lines of intersection of the
crystal boundary with the a-lattice cell walls. Starting from the different
o-points as origins, the atoms in the crystals will align themselves by
relaxation until, at the cell wall, the coordination changes; this change
is the dislocation. For" physical" dislocation networks it is not essential
to locate exactly the a-points, i.e. to know, for example, their internal
coordinates.
However, for high-angle boundaries or boundaries between widely
different phases the situation is modified (Section 13.3). In this case
optimum boundaries are possible too, but on a completely different
basis. The decisive factor is the periodicity of the pattern of lattice
points. Here, the total pattern is subdivided by the o-lattice cells into
pattern elements, and for a given transformation A the pattern element
is determined by the internal coordinates (the equivalence class) of its
o-point. The essential data are the number of equivalence classes with
coinciding elements, the internal coordinates of the O-points, and the
size of the a-lattice cell. The smallest values of Eq. (13.3-3) should
correspond to an optimum boundary; the period of the pattern is
independent of translation and so is this equation. Yet, it might be
expected that, out of all the patterns which differ only by translation,
the most stable pattern would be that in which the o-points are occupied
by atoms.
Once again the difference between the o-lattice and the coincidence
site lattice should be emphasized: the coincidence site lattice is deter-
mined exclusively by the point pattern and is independent of the relation
A between the two lattices. It indicates the period of the pattern and is
15*
228 Chapters 13-14. Summary and Discussion

discontinuous with respect to any change in A. Hence, "coincidence site

lattice" is synonymous with" periodic pattern". The O-lattice, however,
depends strictly on A. It varies continuously with A and, if a coincidence
site lattice is present, it is contained in the O-lattice. There are cases
where the O-lattice and the coincidence site lattice are identical. Ho wever,
the information content of the coincidence site lattice is only a small
fraction of that of the O-lattice.
Slight deviations from optimum patterns can lead to secondary
dislocation networks. When secondary dislocations move through the
boundary the structure of the optimum pattern is conserved, but not
necessarily its position (Section 13.5).
The whole linear geometrical theory of crystalline interfaces is, in a
sense, condensed in Section 14.2. Practically every boundary can be
understood as a "dislocation network" where, however, only a geo-
metrical frame remains of the dislocation notion which still conserves
the invariance and continuity properties of the Burgers vectors. In the
case of an optimum pattern, this pattern itself can constitute a periodic
basis for secondary dislocations. The physical nature of these disloca-
tions will have to be reconsidered since the relaxation in high-angle
boundaries might be different from that in subgrain boundaries.
The extension of the O-lattice theory to non-linear problems is
discussed in Section 14.3, applied to examples where crystal dislocations
end at a boundary or penetrate it.
In this manner the geometrical basis for arbitrary problems of
crystalline interfaces is given. Future physical investigations may develop
from here.
Suggestions for Problems
Problems of the type determining optimum phase boundaries usually
go beyond the framework of exercises. It might be possible to construct
such problems on simplified two-dimensional models which can be
calculated by ordinary means (e.g. to find the optimum orientation of
two lattices with slightly different lattice constants, determine the
minimum geometrical parameter in order to calculate O-lattice and cell
walls for the optimum orientation).
Many exercises can be found for periodic patterns from Table 13.3/T1.
The different periodic patterns can be constructed and checked on the
moire models. Also, the DSC-Iattice and the secondary dislocation
networks for slight deviations from the optimum orientation can be
calculated.
Non-linear problems can hardly be solved without a computer.
 Appendix
A 1. Matrix Calculation Procedures
Since the O-lattice theory is based on linear algebra, i.e. the matrix
calculation, we present below the procedures involved in this technique.
It is not our aim to develop and prove the mathematics, but rather to
show how it is done. The proofs can be found in textbooks. A useful
practical introduction is given by Wayman (1964).
The matrix calculation can be formally presented in an elegant
manner, but it is often tedious for the practical calculation (e.g. by
means of a desk calculator). On the other hand it is suitable for program-
ming in computers, so that, without too much difficulty, problems can
be treated which would be unthinkable on working by hand.
We restrict ourselves to the three- and the two-dimensional space.
Unless otherwise mentioned, the three-dimensional space is assumed.

1. Coordinate Systems
All quantitative indications of positions are referred to a coordinate
system, which is determined by an origin and three linearly independent
basic vectors u 1 , U 2 , U:J, such that every point in the space is determined
by the position vector ;x:
(A 1-1)
The vectors are indicated by small, bold face letters, the coordinates in
ordinary small letters. As is done in tensor calculations, we distinguish
between lower and upper indices and apply Einstein's summation
convention
3
x' u i : = L xi u •. (A 1-2)
.=1
If, in a term or a product, the same index appears above and below, it
represents the summation over this index. The meaning of the upper and
the lower indices will be discussed below.
In principle, any arbitrary coordinate system may be chosen. In
our case, however, two types are especially useful:
a) The first is the orthogonal coordinate system with three mutually
perpendicular basic vectors of the same unit length (of 1 A, or the length
of the cube edge in a cubic crystal system). The unit vectors are arranged
230 Appendix

in a right-handed sequence ([Ut XU 2 ] =Ua). To facilitate plotting on

graph paper it is advantageous to express the results in orthogonal
coordinates.
b) The second is the crystal coordinate system with the unit vectors
of the crystal lattice as basic vectors. The advantage is here that the
crystal period is expressed as a "period" of integral numbers, so that the
coordinates directly represent the position of the point with respect to
the crystal unit cell.
Every crystal lattice is coupled with its" reciprocal lattice ", which is
often used in X-ray- and electron diffraction and also in solid state
physics (Brillouin zones). The unit vectors of the reciprocal lattice are
1 [UI X Us]
u= V (A 1-3)
with
(A 1-4)
the volume of the crystal unit cell. The other unit vectors UII and u 3
follow by cyclic permutation.
The following orthogonality relation is fulfilled:
(u.u k) =~: (A 1-5)
1 i=k
{ (A 1-6)
~i= 0 i=l=k·

A vector ;,:, referred to the reciprocal basis is:

(A 1-7)
In the unit vectors the lower indices u i denote unit vectors of the crystal
lattice, the upper indices u i the unit vectors of the reciprocal lattice.
In the coordinates, however, the upper indices Xi indicate that the
vector is referred to the crystal coordinate system, and the lower indices
Xi refer to the reciprocal coordinate system. Xi are often called the
contravariant and Xi the covariant components of a vector. We shall not
use this highly confusing notation here.
In the orthogonal coordinate system, the "crystal" system and the
reciprocal coordinate system can be considered as identical (with the
exception of the unit length A and A-I). Hence there is no need to
distinguish between lower and upper indices.

2. The Transition from the Orthogonal

to the Crystal Coordinate System
A general crystal system is determined by the lattice constants
a, b, c, oc, p, y, where a, b, and c are the lengths of the unit vectors (in A)
 A 1. Matrix Calculation Procedures 231

and IX, p, Y the angles between the unit vectors.

IX = -t (b, c) (A 1-8)

and correspondingly for p and y (Section 2.2).

Starting from an orthogonal coordinate system (x, y, z), we have
two degrees of freedom in the choice of the orientation of the crystal
coordinate system within the orthogonal system. We may, for example,
place the c-axis on the z-axis (as is conventional in mineralogy) and the
a-axis in the xz-plane. [The notion (x, y, z) is used synonymously with
(UtU 2 Ua) of the orthogonal coordinate system.] Now, we express the new
crystal unit vectors u~ u~ u; (i.e. a, b, c) in terms of the old orthogonal
coordinates:
u~ =Ut s~ +u2 s~ +Ua s~
u~ =Ut s~ +u2 s~ +Ua s~ (A 1-9)
u; =Ut s~ +u2 s: +Ua s:.
Written in abbreviated form:

(A 1-10)

(Ai-H)

and in abbreviated matrix notation:

(A 1-12)

(IT means transposed = row vector).

In this special case, w is a sort of a "super vector", the components
of which are the unit vectors. Here we deal with pure vector calculation
and do not yet decompose the vector into coordinates.
The general convention for matrix multiplication is:

C=AB. (A 1-13)
In index notation:
(A 1-14)
Sometimes it is convenient to denote the summing indices by Greek
letters as in Eq. (A1-14). For example the element:

(Ai-iS)
232 Appendix

i.e.
Eq. (A 1-16)

(A1-16)

This means that the scalar product of the row- or line-vector of the first
matrix and the column vector of the second matrix is the element at the
crossing point in the resultant matrix. The matrix product is usually not
commutative (An =l=BA). If the first matrix consists of only one row
(row vector), then the resulting matrix consists also of one row. Cor-
respondingly, if the second matrix is a column vector, the result is also
a column vector.

Remark
We combine here matrix- and tensor notation. In the matrix notation the
position of the elements in the matrix is determined by the sequence of the
indices and the matrix multiplication is non-commutative. In the tensor
notation, however, the coefficients form a set, where every element is iden-
tified by its indices but has no fixed position in a matrix. Hence, in the tensor
notation the indices can be placed on top of each others and the multiplica-
tion of the coefficients is communtative.
In the combined matrix-tensor notation Eq. (AI-H) should be written:

(A1-14a)
Then the transpose of C: C T = BT AT is:
(A1-14 b)
In tensor notation
(A1-14 c)

is the same as Eq. (AI-14) as only coefficients are multiplied and added
according to the summation convention and not rows an columns of matrices.
In this book, where the indices are placed vertically above each others the
upper index has to be considered as the first.
Our task now is to determine the coefficients of the S-matrix
[Eq. (A 1-11)]. If the c-axis is placed on the z-axis, then

sl=c (Ai-H)
and
(A1-i8)
 A 1. Matrix Calculation Procedures 233

and, if the a-axis is placed in the (xz)-plane:

s~=o. (A1-19)

The remainder of the coefficients are determined from Eq. (A 1-9) by

the scalar products of the vector u~ by using the orthogonality of the u.:

(A 1-20)
and the definition (A 1-8).
For example:
(u ,l U3') = a . c . cos fJ = 3 = SI
3 • S3
SI 3 . C (A 1-21)
i.e.
s~=a. cosfJ. (A 1-22)

Further equations are given by the scalar products of the vectors u~

with themselves, e.g.

(A 1-23)
i.e.
(A 1-24)
and
1
SI . fJ •
=a· sIn (A 1-25)

In this way, the whole S-matrix is obtained. Due to the lack of space, we
write the column vectors below, instead of beside one another.

a s~=a.sinfJ
s~=o
s~ =a· cosfJ
b s~ = (b/sin fJ) (cos y -cos ex cos fJ)
s~ = (b/sin fJ) [2 cos ex cos fJ cos y +sin2 ex sin 2 fJ
(A 1-26)
- cos2 ex cos 2 fJ - cos 2 y]1
s~=b· cos ex

c sl=o
s:=o
s:=c
The column vectors of the S-matrix are thus the unit vectors of the
crystal coordinate system expressed in orthogonal coordinates.
The basic vectors of any crystal system can be expressed in this
form by specifying the lattice constants. Summarizing: The S-matrix
234 Appendix

is determined by expressing the new basic vectors in terms of the old

ones, without decomposing the vectors into coordinates.

3. Coordinate- and vector transformations

We assume now that both the coordinate systems: the orthogonal and
the crystal system, are given, and consider an arbitrary vector;,: with the
orthogonal coordinates ~. and the crystal coordinates ~'.
(A 1-27)

By using the Greek letter ~', ~'. we stress the fact that they represent
the same vector in different coordinate systems. The transition between
the coordinates is given by:
{A 1-28)

Therefore, the 8-matrix transforms the crystal coordinates of a vector

(e.g. f (1) = [100]) into its orthogonal coordinates (e(l) = a . sin p,
0, a . cos P). We term this transition, in which the vector is fixed but
the coordinate system changes, a coordinate transformation.
Eq. (A1-28) takes another meaning when written in the form:

(A 1-29)

Here we choose the orthogonal coordinate system and consider it as

fixed, but now;,: and ;,:' represent two different vectors. For example,
the vector [1 OOJ in the orthogonal system (x-edge of the unit cube) is
transformed into the vector a = [a . sin P, 0, a . cos PJ in the same co-
ordinate system. The vector has changed its length from 1 A to a A
and is rotated by n/2 -P in the positive sense in the xz-plane. The
complete orthogonal unit cube is transformed by 8 into the crystal unit
cell. We call this transition between two different vectors a vector trans-
formation.
The inverse coordinate transformation is given by

e'=8-1 e (A 1-30)

and the inverse vector transformation by

;,: = 8-1 ;,:' • (A1-31)

[The calculation of the inverse matrix S-l will be given in Eq. (A1-36),
(A1-37).J
It is important to know, in every case, whether one is dealing with a
coordinate transformation or with a vector transformation.
 A 1. Matrix Calculation Procedures 235

4. The Determinant and the Inverse Matrix

The determinant of a matrix is the following sum of diagonal products:

,+,
3+,
"'-
"'-
~ /' / / ' / / / 6-
4-

""
'- ~ /' --'
2+ " ~ /' ~ /' /' 5-
/' '---- /' '---- /' /.

, /' "'-,,- / ' "-

/' a31 , / , 032
""
./' a33
/'---- (A 1-32)
"'-
/' >" " ",,- "'-
/'-. /"

~ all an a33 + a 12 a 23 a 31 + a 13 a Z1 a 3Z
'----v----' '----v---'
(1) (2) (3)
- a13 aZZ all - a1z aZl all - all 0Zl an
'---v------' '---v----' ~.~~~

(4) (5) (6)

For two-row matrices, the determinant is:

Eq. (A1-33):

(A 1-33)

As we shall see, the determinant is a number, a ratio.

In order to calculate what is called the subdeterminant Aik of an
element a ik of the general matrix A, the vertical and horizontal rows
containing the element aik are crossed out. The determinant of the
remaining two-row matrix is calculated and multiplied by (_1)Hk in
order to fix the sign, i.e.
Eq. (A1-34):

(A 1-34)

The subdeterminant of a two-row matrix An is:

An = (_1)2+1. a12 = -a12 • (A 1-35)
236 Appendix

An important concept is the rank of a matrix. The rank is the number of

rows (columns respectively) of the largest subdeterminant (determinant
resp.) with non-zero value. If for a three-row matrix the determinant
IAI ,*,0, then its rank is 3. If, however, IAI =0, and at least one two-row
subdeterminant Aik ,*,0, then rank (A) =2, etc.
The elements of the inverse matrix are determined by:
(-1) _ Aki
aik - IAI . (A 1-36)

In order to calculate the element a~31) of the inverse matrix, the sub-
determinant of the element a32 (the mirror image across the main dia-
gonal) is calculated and divided by the determinant of A.

-1
"'. . .
( •" ' ., a23(-1) )

. .X., (A1-37)
"

The definition of the inverse matrix implies that

AA-l=A-IA=I (A1-3 8)
with 1=151 = identity.

1= (°10 0°1°0) .
1 (A 1-39)

As a check, it is recommended to calculate the product (A 1-38) every

'*'
time an inverse matrix has been determined.
An inverse matrix only exists if the determinant IAI 0, as shown
by Eq. (A1-36).
By means of the inverse matrix, for example, Eq. (A1-29) may be
solved for ;r:
;r' =S;r. (A 1-40)
Both sides are multiplied from the left by S-I:
S-I;r' = S-1 S;r =I;r =;r (A1-41)
which is the same as Eq. (A 1-31).
Another application of the inverse matrix is the determination of the
basic vectors of the reciprocal lattice U'i, expressed in orthogonal co-
 A 1. Matrix Calculation Procedures 237

ordinates. On the one hand the U'i and u; are related by Eq. (A 1-5):
(A 1-42)
(u; = general unit vectors, u i = orthogonal unit vectors). On the other
hand:
(A1-43)
According to the matrix multiplication convention a row vector of the
first matrix is multiplied as a scalar product with a column vector of the
second matrix. Here, the column vectors of 8 are the vectors u~, hence
the row vectors of 8-1 must be the U'i. Thus, the vector products of
Eq. (A1-3) can be replaced by the matrix inversion in order to deter-
mine the vectors u' i .
5. Metrics
The metric tensor is defined in the following way:
G =gik = (u; u~). (A 1-44)
The terms of G are thus the scalar products of the basic vectors. Since
the scalar products are commutative, the metric tensor is always sym-
metrical.
(A1-45)
In the orthogonal system, G is, according to Eq. (A 1-20), given by
G =gik = (Ui Uk) =~ik =1. (A 1-46)
G can be determined from the 8-matrix by:
G=8 T 8 (A 1-47)
with 8T the transposed matrix of 8. (The transposed matrix is the mirror
image with respect to the main diagonal.)
(A 1-48)
The metric tensor is a characteristic of the coordinate system. It helps to
determine metric properties such as absolute length and angles in general
coordinate systems. The scalar product of two vectors is:
I;}:I 'Iyl cos ¢ =;}:T Gy = X· gik yk (A 1-49)
and, as a special case, the square of the absolute length is
1;}:12 =;}:T G;}: = Xi gik Xk (A 1-50)
and the angle ¢ between the vectors;}: and y is given by:
xTGy xi gik yk
cos ¢ = ~ = [(xi gik xk)'-'-;(~y'o--·g-ik-y---'k-:-:)]"! . (Ai-51)
238 Appendix

Another way would be to transform the vectors re and y first to the

orthogonal coordinate system and then to determine the length and
angle by the standard orthogonal scalar products (Xi y.).
The metric tensor is also required for the coordinate transformation
from the crystal coordinate system to the reciprocal coordinate system.
(A 1-52)
The metric tensor of the reciprocal coordinate system is:
gi1~=G-l (A 1-53)
and thus the inverse coordinate transformation is
~. =gi" ~". (A 1-54)
(In the orthogonal coordinate system G = I, therefore ~. = ~ i' Hence,
disregarding the dimension A or A-I, the distinction between upper and
lower indices is meaningless in this case.) The scalar product can also
be written as:
(rey) = (xi Yi) (A 1-55)
(rere) = lrepa = (x' x.) (A i-56)
and
X· Yi
cos cjJ = [(xi Xi) (yi Y.)]! . (Ai-57)

Summarizing the coordinate transformations, we obtain:

orthogonal crystal reciprocal
8- 1 o
--+ --+
e(orth) etc) e(') •
8 0-1
~ ~

6. General Linear Transformations

A linear transformation A describes an affine deformation (leaving
parallel lines parallel) of the whole space with all its vectors. Such trans-
formations are, for example, expansion, rotation, and shear. A trans-
formation is said to be homogeneous if the point at the origin remains
fixed (no constant terms in the equations). If the transformation
contains a translation, it is termed nonhomogeneous. We shall only
discuss homogeneous transformations.
If by the transformation A every vector re is related to a distinct
vector re' and vice versa
x'i=a~ x" (Ai-58)
the transformation is characterized as regular or non degenerate. In this
case the determinant IAI =l= 0 and the inverse matrix A-I can be calculated
 A 1. Matrix Calculation Procedures 239

[Eq. (A1-36)J. (In a non regular case, for example, a three-dimensional

space would be imaged onto a two-dimensional one.)
The coefficients of the matrix A depend, of course, on the coordinate
system in which the transformation is represented. Many transforma-
tions acquire an especially convenient form in the orthogonal coordinate
system.
Examples in orthogonal coordinates are:
a) An expansion (or contraction) by the factor a in the direction of the
x-axis, by b in the y-axis and by c in the z-axis is described by:

a
Expansion E = ( 0 b 0 .
0 0) (A 1-59)
o 0 c

I
b) A convenient representation of a "otation in the right-handed
screw sense by the angle () around an axis given by the unit vector
[Ct, clI ' caJ (c. = direction cosines) is the following:

"ik = flik cos () +c. c" (i-cos ()) + C'k sin ()

c.,,= ( ~3 -; -::).
(A 1-60)
-Cll '1. 0

A special advantage of the rotations (orthogonal transformation) in the

orthogonal coordinate system is that:
R-I=R T • (A 1-61)
In order to determine the inverse of R, the mirror image of the matrix
has to be written with respect to the main diagonal.
An important application is the rotation of an arbitrary vector v
into a special position, e.g. into the z-axis. Such a vector can be the
normal to the b-subspace or the invariant eigenvector. [Standard form
Eq. (12.9-22}.J The given vector v is assumed to be normalized (Ivl =1).
Its components are [VI' V 2 , Va]. At the same time they are the direction
cosines, i.e. for example, va = cos ~ (v, Us) = cos (). In order to rotate
the vector v into the z-axis, we have to rotate it around an axis which
is perpendicular to both, v and Us. This is obtained from the vector
product [v X UsJ :
,, "-, "" , "" yz
..!:': VI Vz
)C"
V3
'y"
.vI,
.x
'0 (A 1-62)
.? : 0 0/ '1" '0/
'--v---'~~
/, ", "-
Vz -VI 0
240 Appendix

This new vector is also nonnalized:

Va
c= { (vf+vl)§' a}. (A1-63)

In this way, the vector c required for the rotation Eq. (A 1-60) is deter-
mined. Further: cos 0 = va and sin 0 = (1 - v=)§, so that now the data
needed for the formulation of the rotation matrix are all known.
c) A pure shear in the direction of the y-axis, with the (xy)-plane
invariant, is described by:

Shear C = (1 0 0)
0 1 (J
001
• (A 1-64)

The determinant of a matrix A of a linear transfonnation is the ratio

between the new and the old unit volumes. A is here understood as the
transfonnation of the basic vectors
(A 1-65)
Then
V'
IAI=v· (A 1-66)

A transfonnation for which IAI = 1 leaves the unit volume unchanged.

It is called a unimodular transfonnation (e.g. pure shear). If, in addition,
the distances between all the points stay unchanged the transfonnation
is tenned orthogonal (general rotation + mirror image).
7. Combination of Transformations
Several consecutive transfonnations can be app1ied to a vector,
where the later transfonnation acts on the result of the earlier trans-
fonnation. For example, a space can be sheared in the direction of the
y-axis [transfonnation CEq. (A 1-64)] and then be expanded by 1 +8. in
the direction of the z-axis [Eq. (Ai-59), a =b =1, c =1 +8.]. The result
is given by:
aJ"=ECaJ. (A 1-67)
Arbitrary transfonnations may be constructed in this way. The con-
vention of the matrix calculation is such that the sequence of the operations
goes from right to left, opposite to the sense of writing.

8. Transformation of Transformations
It may occur, for example, that a pure shear has to be represented
in such a way that the orientation of the shear direction and the shear
plane are given in space. We may start from a special representation
of the type of Eq. (A 1-64), and then rotate the whole transfonnation.
 A 1. Matrix Calculation Procedures 241

We have to decide here whether we leave the coordinate system

fixed and transform the vector space or whether we rotate the coordinate
system. This is essentially a question of the sign. It is important for the
representation whether, for example, the vector space or the coordinate
system is rotated by + 30° around the positive z-axis. The analytical
representation, however, is the same for a rotation of the vector space
by + 30° or of the coordinate system by - 30°.
Let us consider for the moment the rotation of the vector space.
The rotation R shall be defined in the same way in which 8 is defined
[Eq. (A 1-11), (A 1-60), (A 1-61)J such that, in the fixed coordinate
system, the new vector follows from the old by:
(A 1-68)
We now apply this rotation on the general transformation A (e.g. shear).
The rotated transformation together with its characteristic elements is
then given by (vector transformation) :
A'=RAR-l. (A 1-69)
If, however, the coordinate system is rotated by R, then the trans-
formation A" represented in the new coordinate system (coordinate
transformation) is:
A"=R-IAR. (A 1-70)
It might, for example, happen that the simple shear in the special
orientation C in a fixed orthogonal coordinate system has to be rotated
into another orientation (vector transformation) :
C'=RCR-l (A1-71)
then the new general shear has to be represented in terms of crystal
coordinates [coordinate transformation, Eq. (A1-29)J
C{c) =8-1 C'8 =8-1 RCR-1 8. (A 1-72)
The same transformation now represented in the reciprocal coordinate
system [,'= G, ~ aJ =GaJ' ~ aJ'= G-1 aJ i.e. G-l f':::jR in Eq. (A 1-70)] is:
(A1-73)
In this way, arbitrary transformations in arbitrary coordinate systems
can be constructed. (A control of tedious matrix calculations is given by
the determinants as IABCI =IAI·IBI·ICI and lA-II =1!1AI·)
We have to distinguish clearly between the combination of trans-
formations and the transformation of transformations. When a shear
in the y-direction [Eq. (A1-64)J is combined with a rotation by () =90°
around the positive z-axis, the vector space is sheared beforehand and
16 Bollmann, Crystal Defects
242 Appendix

then rotated such that, as a result of the combined transformation BC,

the following vectors are related: 00[100J -+ 00' [01 OJ and oo[001J -+
00'[ -Cl, 0, 1J. If, however, the shearing is transformed by a rotation
(BCB-I) then the following vectors are related: 00[1 OOJ -+ 00' [1 OOJ and
00 [001J -+ 00' [ -Cl, 0, 1J. The vector space is only sheared but in a new
direction. This example may be calculated as an exercise. In any case
it is to be recommended in doubtful cases to calculate model examples.

9. Addition of Transformations
The addition of two transformations A +B means the addition of
transformed vectors 00' originating from the same vector 00
Coo=(A+B) 00 =Aoo+Boo. (A 1-74)
The matrices are added element by element
C'k =a'k +bik · (A1-75)
10. Invariant Elements of a Transformation
By a non-degenerate linear transformation, the vector space as a
whole is transformed into itself. In the case of a homogeneous trans-
formation the origin stays unchanged oo[OOOJ -+ 00' [OOOJ. We may ask
whether further unchanged elements exist. Vectors with invariant
orientation (while their length may change), are called eigenvectors. They
are determined by the following condition:
(A 1-76)
The numerical parameter A is called the eigenvalue. Its meaning is the
ratio of the length of the eigenvector after and before the transformation
~ (A 1-77)
A= 1:Ili .
The equation for the eigenvectors is:
(A-U)oo=O. (A 1-7S)
where 0 means the zero vector [000]. This equation has non-trivial
solutions only when:
IA-UI=O. (A 1-79)
Eq. (A 1-79) is an equation of the third degree for A. The (real) solutions A
are introduced into Eq. (A1-7S), from which the eigenvectors 00 can be
calculated.
If we refer to the end vector 00' instead of to the starting vector 00,
and denote this end vector by 00(0), Eq. (A 1-7S) becomes:
(l-AA-I) 00(0) =0. (Ai-SO)
 A2. Moire-Models 243

The eigenvalues Aare determined in an analogous manner to Eq. (A 1-79).

If we abbreviate
(1 - AA -1) :1)(0) : = T:I)(O) (A 1-81)
this equation becomes, in index notation:
4X(0)k=0. (A 1-82)
The eigenvectors can be determined by the procedure given in Section 12.9
[Eq. (12.9-9)] etc.
For A= 1, not only the direction but also the length of the eigen-
vector is conserved. In this case, it is called an invariant eigenvector. If
we compare (A 1-80) with the basic equation ofthe O-lattice [Eq. (12.3-6)]:
(A 1-83)
we see that the O-element passing through the origin (b(L) = 0) contains
the invariant eigenvector.
This collection of procedures of matrix calculation cannot naturally
be complete. The best way to get acquainted with the technique is to
calculate examples with directly evident results.

A2. Moire-Models
Four pairs of moire models and a scale M are joined to the book.
The model 1 corresponds to a {11 0}-face in the body-centred cubic (bcc)
structure. The point arrangement is approximately hexagonal. The
length of the cube edge is chosen as 2 mm. Model 2 represents a {11 2}-
face and model 3 a {1 OO}-face in the same structure. Model 4 consists
of parallel lines also with 2 mm spacing. Fig. A2/1 shows the dimensional
situation.
Every model consists of a sheet a and a sheet b. Sheet a may be
placed below the scale and its position fixed by means of the straight
scales. The zero position on all four linear scales corresponds to the
exact centering. Sheet b can be placed on top of the scales, and its
position is determined with respect to rotation. As moire figures are
extremely sensitive to changes in relative position, a high precision is
needed, especially if calculated patterns are to be compared quantitatively
with the models.
It was shown in Section 12.10 that, with respect to translation a
shift d(S) is equivalent to one by d(l) ( = d(b») , where d(l) = - A -1 d(lI) •
Instead of executing the rotation and the translation with sheet b, the
translation can be done with sheet a in coordinates of crystal 1 and the
rotation alone with sheet b. The angle of rotation is best fixed with all
four marks simultaneously on the circular segments. The coordinate
244

.-.-.-....
••••••••••- •••••
•••••
Appendix

-1.·.·.·.·.·
-·~r·!·!·! .....
•••••••••• •••••
••••• 1312· • • • •
-
I fil

•••••••
•••••••
- .......
•••••••
-1• • • • • • •

11 I
Fig. A2/1. Dimensions of the moire models (the unit corresponds to 2 milli-
meters)

origin in this case is the centre of the scale sheet, i.e. also the centre
of sheet b.
The upper left-hand comer of the given lattices are cut such that the
sheets can always be placed on top of each other in the same orientation.
In this way eventual irregularities in the drawing or distortions due to
the photographic reproduction may cancel out optimally.
In order to add an expansion to the rotation, the aspect of perspective
can be applied by keeping the sheets apart when viewing them. The
movement rule [Eq. (5.4-1)] for edge dislocations can be checked in this
way with model 4. Expansion combined with rotation and translation
can also be controlled quantitatively by mounting the sheets on glass
plates and photographing them above a luminous table by means of a
reflex camera.
The models are transpcuent in view of their possIble use as demon-
stration material in lectures, where they can be projected. Here a view-
graph is recommended, in which the sheets M and a can be placed
under- and the sheet b on top of the transparent plastic foil used for
writing. In this way the displacements of sheet b do not affect the lower
sheets. An epidiascope should not be used for projection, as the models
are likely to be burnt.
 A3. The Direct Observation of Crystal Defects 245

Many calculations can be controlled by means of the models, especially

in connection with Table 13.3/T1. Also, the behaviour of the pattern
on translation of one lattice with respect to the other can be studied.
Point lattices have been treated exclusively until now. In order to
calculate, for example, the rotation of a line lattice (moire model No.4)
the line lattice can first be treated as a point lattice, which leads to
Eq. (12.3-7) with (l-A-ltl in the form of Eq. (12.3-12). Then, in the
final result, the b~Ltcomponent is assumed to be continuous while b~L)
stays discrete; i.e. biLl =E and b~L) =n. Afterwards, Eis eliminated, from
which follows the equation of the moire lines:
x~O) = _x~O) cotan((J/2) +n/(2 sin 2 ((J/2)). (A2-1)
A wide range of moire problems can be solved by the general non-linear
method given in Section 14.3; for example, the superposition of circular
waves. The calculation shall be briefly described here:
Origins: wave 1 (d/2, 0)
wave 2 (-d/2, 0)
x(l)(z+b(L») = (x-(d/2))2+y2-(z+n)2=0 (A 2-2)
x(Il)(z) = (x+(d/2)}2+y2-Z2=0. (A 2-3)
From
X(2) (z) - X(I) (z + bILl) = 0 (A 2-4)
follows
xd n
z=-
n ---
2
(A2-5)
in Eq. (A2-2):
x· yl
---ns- dl-nl =1. (A 2-6)
4 4
For Inl < Idl, the equation represents hyperbola, and for Inl > Idl ellipses
(which arise if one wave is moving outwards and the other inwards).
(For further details on moire patterns, see G. Oster.)

A3. The Direct Observation of Crystal Defects

We give here a few indications on the literature dealing with the
direct observation of crystal defects. The method which shows directly
the atomic arrangement is field ion microscopy, which was developed by
E. W. Miiller. It is described among others by D. G. Brandon (1966).
The method which is mostly used today for the study of dislocations and
stacking faults is transmission electron microscopy which, based on
work of R. D. Heidenreich and also R. Castaing, was developed by the
246 Appendix

group P. B. Hirsch, M. J. Whelan, R. W. Home and independently by

the author. Books are indicated in the bibliography. X-ray topography
also permits the direct imaging of crystal defects in an analogous manner
to transmission electron microscopy, on much thicker specimens with,
however, lower magnification (see e.g. Brandon, 1966). Somewhat less
direct methods are on the one hand etch pits, which form, on etching a
crystal, at locations where dislocations end at the crystal surface, and
on the other hand the surface decoration (developed by Basset and
further by Bethge). Here, gold is lightly evaporated on rock salt, whereby
the gold particles diffuse to atomic surface steps which become visible
in this way. (See e.g. Picht and Heydenreich, 1966.)
 References
Dislocation Theory
Bueren, H. G. van: Imperfections in crystals. Amsterdam: North Holland
Publ. Co. 1961.
Cottrell, A. H.: Dislocations and plastic flow in crystals. Oxford: Clarendon
Press 1953.
Fiedel, J;: Dislocations. Oxford: Pergamon Press 1964.
Hirth, J. P., Lothe, J.: Theory of dislocations. New York: McGraw-Hill
1968.
Nabarro, F. R. N.: Theory of crystal dislocations. Oxford: Clarendon Press
1967.
Read, W. T.: Dislocations in crystals. New York: McGraw-Hill 1953.
Seeger, A.: Moderne Probleme der Metallphysik. Berlin-Heidelberg-New
York: Springer 1966.
- Handbuch der Physik, Bd.7/1, S.383. Berlin-Heidelberg-New York:
Springer 1953.
Weerlman, J., Weerlman, J. R.: Elementary dislocation theory. New York:
McMillan 1964.
Symposia on Dislocation Theory
Dislocation dynamics. Edit. by: Rosenfield, A. R., Hahn, G. T., Bement,
A. L., u. Jaffee, R. I. New York: McGraw-Hill 1968.
Dislocations in solids. Discussions Faraday Soc. No. 38 (1964).
Fundamental aspects of dislocation theory. Edit. by: J. A. Simmons, R.
,deWit, R. Bullough. National Bureau of Standards. Special Publication
317. Washington, D.C. 1970.
Physique des dislocations. Paris: Presses Universitaires de France 1967.
Grain Boundaries and Phase Transformations
Amelinckx, S., Dekeyser, W.: Solid state physics, vol. 8, p. 325. London:
Academic Press 1959.
Christian, J. W.: The theory of transformations in metals and alloys. Oxford:
Pergamon Press 1965.
McLean, D.: Grain boundaries in metals. Oxford: Clarendon Press 1957.
Wayman, C. M.: Introduction to the crystallography of marlensitic trans-
formations. New York: McMillan 1964.

Special References
Section:
2.2-2.3 International tables for X-ray crystallography. Birmingham, Eng-
land: K ynoch Press 1962.
4.1 Billington, D. S.: Radiation damage in ·solids. Proceedings of the
Internat. School of Physics. "Enrico Fermi" XVIII Course,
Italian Physical Society. New York: Academic Press 1962.
- Crawford, J. H., Jr.: Radiation damage in solids. New Jersey:
Princeton University Press 1961.
248 References

Damask, A. C., Dienes, G. J.: Point defects in metals. New York:

Gordon and Breach 1963.
Dienes, G. J., Vineyard, G. H.: Radiation effects in solids. New
York: Interscience Publishers Inc. 1957.
Lattice defects in quenched metals. Edit. by: Cotterill, R M. J.,
Doyama, M., Jackson, J. J., Meshii, M.: London-New York:
Academic Press 1965.
Strumane, R, Nihoul, J., Gevers, R, Amelinc1a:, S.: The interaction
of radiation with solids. Proceedings of the Intemat. Summer
School on Solid State Physics held at Mol/Belgium. Amsterdam:
North Holland Publishing Co. 1964.
Vacancies and interstitials in metals. Edit. by: Seeger, A., Schu-
macher, D., Schilling, W., Diehl, J., Amsterdam: North
Holland 1970.
5.2 Bilby, B. A., Bullough, R, Smith, E.: Proc. Roy. Soc. (London),
Ser. A 231,263 (1955).
6.3 Weertman, J.: Phil. Mag. 11, 1217 (1965).
6.5 Brown, L. M.: Phil. Mag. 10,441 (1964).
7.5 Hale, F. K., Brown, H. M.: Fourth European Regional Conference
on Electron Microscopy, Rom 1968, vol. 1, p. 409.
8.3 Bollmann, W., Perry, A. J.: Phys. stat. sol. 24, K141 (1967).
Thompson, N.: Proc. Phys. Soc. (London) B 66, 481 (1953).
8.4 Bollmann, W.: Acta Met. 9, 972 (1961).
Christian, J. W., Swann, P. R.: Stacking faults in metals and alloys
(AIME-Conf. on "Alloying Effects on Concentrated Solid
Solutions"). Cleveland, Ohio 1963.
Gedwill, M. A., Altstetter, C. J., Wayman, C. M.: Trans. AIME
230,453 (1964).
Hirth, J. P. (Symposia, ref. 3).
Seeger, A.: Z. Metallkunde 44, 247 (1953); 47, 653 (1961).
9.1 Bollmann, W.: Phil. Mag. 7, 1513 (1962); - Discussions Faraday
Soc. No. 38, 26 (1964).
9.3 Hale, F. K., Brown, H. M. (see ref. 7.5).
9.5 Thompson, N. (see ref. 8.3).
to.1 Damask, A. C., Dienes, G. J.: (see ref. 4.1).
Hirsch, P. B., Silcox, J. Smallman, R E., Westmacott, K. H.:
Phil. Mag. 3, 897 (1958).
Homstra, J.: Acta Met. 10, 987 (1962).
Yoshida, S.: Kiritani, M., Shimomura, Y.: J. Electronmicroscopy
12,146 (1963).
11.8 Bailey, J. E.: Phil. Mag. 5, 833 (1960).
Bollmann, W.: J. Inst. Metals 87, 439 (1959).
Merklen, P., Furubayashi, E.: C. R Acad. Sci. Paris 268, 2159
(1969).
12.2 Ranganathan, S.: Acta Cryst. 21, 197 (1966).
12.3 Bollmann, W.: Phil. Mag. 16, 363, 383 (1967).
13.1 Fletcher, N. H.: Phil. Mag. 16, 159 (1967).
Gleiter, H. (see ref. Section 14.2).
Ishida, Y., Hasegawa, T., Nagata, F.: J. Appl. Phys. 40, 2182
(1969).
Levy, J.: Phys. stat. sol. 31, 193 (1969).
13.2 Bollmann, W., Nissen, H.-U.: Acta Cryst. A 24, 546 (1968).
 References 249

13.3 Bollmann, W., Perry, A. J.: Phil. Mag. 20, 33 (1969).

13.5 Schober, T., Balluffi, R. W.: Phil. Mag. 20, 511 (1969).
- Phil. Mag. 21, 109 (1970).
14.2 Bollmann, W.: (Symposia, ref. 3).
Gleiter, H.: Acta Met. 17, 565 (1969a); 17,858 (1969b).
Marcinkowski, M. J.: (Symposia, ref. 3).
Sleeswyk, A. W.: Physique des dislocations (Ref. Symposia),
P·78.
14.3 Bollmann, W.: Dislocation dynamics (Ref. Symposia), p. 275.
Ai Z. B. International tables (Ref. 2.2).
A2 Oster, G.: The science of moire patterns. Barrington, New Jersey
08007, U.S.A.: Edmund Scientific Co. (with two sets of moire-
models).
A3 Amelinc1a, S.: The direct observation of dislocations. Solid state
physics 6. New York: Academic Press 1964.
Brandon, D. G.: Modem techniques in metallography. London:
Butterworths 1966.
Heidenreich, R. D.: Fundamentals of transmission electron-
microscopy. New York: Interscience Publishers 1964.
Hirsch, P. B., Howie, A., Nicholson, R. B., Pashley, D. W., Whelan,
M. J.: Electronmicroscopy of thin crystals. London: Butter-
worths 1965.
Muller, E. W., Tsong, T. T.: Field ion microscopy. New York-
London-Amsterdam: Elsevier 1969.
Picht, J., Heydenreich, J.: Einfuhrung in die Elektronenmikro-
skopie. Berlin: VEB Verlag Technik 1966.
Saada, G.: Microscopie electronique des Lames minces cristal1ines.
Paris: Masson 1966.
 Sources of Illustrations
Thanks are due for permission of reproduction of figures to the following
publishers and Journals:
Books
New York: McGraw-Hill: Dislocation dynamics. Edit. by A. R. Rosenfield
et al., 1968, Figs. 13.4/1; 14.3/1, 2, 3,4.
Utrecht: A Oosthoek's Uitgeversmaatschappij NV: Symmetry aspects of
M. C. Escher's periodic drawings by Caroline H. Mac Gillavry, 1965
(© International Union of Crystallography). Fig. 2.2/1.
New York: Reinhold Publishing Corporation: The encyclopedia of micro-
scopy. Edit. by A. L. Clark, 1961, Figs. 3.2/2, 3, 4; 3.9/1, 2; 3.10/1;
5.3/1; 5.4/1, 2.
Berlin-Heidelberg-New York: Springer: Einfiihrung in die Mineralogie by
C. W. Correns, 1968, Fig. 132/2.
Washington D.C.: National Bureau of Standards, special Publication 317.
Fundamental Aspects of Dislocation Theory. Edit by J. A. Simmons,
R. deWit, R. Bullough. 1970. Figs. 12.3/4; 12.8/2.
Journals
Acta Crystallographica: 13.2/1, 3, 5,6, 7, 8.
Acta Metallurgica: Figs. 8.4/3, 4, 5, 6, 7, 8.
The Faraday Society: Figs. 11.1/1; 11.4/1; 11.6/1; 11.9/1; 11.10/1,2.
Journal of the Institute of Metals: Figs. 11/8/1, 2, 3,4, 5,6.
The Philosophical Magazine: Figs. 9.1/1; 9.3/1, 2, 3,4; 10.1/1; 11.4/2; 11.5/1;
12.5/1; 12.6/1, 2; 12.8/1, 4, 5; 12.10/1; 12.11/1; 13.3/2; 13.5/1, 3.
Zeitschrift fiir Kristallographie: Fig. 3.11 /1.
 Subject Index
The page numbers in italics indicate principal references

ABA-sequence 14f. close packed layer 13

ABC-sequence 14f. - - structures 13
activation energy 40 cobalt 87
adaptation, monoclinic-monoclinic coincidence-site lattice 143 f., 197
194 column-vector 232
- , monoclinic-triclinic 191 combination of transformations
addition of transformations 242 240
albite law 193 complete pattern shift lattice 208
alkali feldspar 188 condensation of point defects 113
amorphous state 3 condensed interstitials 30
anisotropic crystals 37 - vacancies 29
Arrhenius plot 39 configuration 37
contravariant components 230
b-Iattice 154 coordinate system, crystal 7, 230
b-net 101 - - , orthogonal 7, 229
b-node 100 - - , reciprocal 230
b-polygon 99 coordinate transformation 234
b-space 121 coordinates, external 148
b-subspace 170f. - , internal 148, 198
b2 -criterion 73 f. coset 148
bad material 41, 56 covalent bonds 3
basic equation 150 covariant components 230
- - , solutions 169 Cremona plan 98
- vectors 7 critical radius 137
body-centred cubic (bcc) structure crossing of split dislocations 108
22f. cryptoperthite 189
Boltzmann constant 39 crystal 1, 3
boundary 186 - , anisotropic 37
Bravais lattices 11 - coordinates 8f.
Burgers loop 43, 216 - defects 1
- vector 41 I. - - , direct observation 245
- - , continuity 98 - , ideal 1,3
- - as coordinate difference 44 - , mathematical description 5
- - , edge orientation 45f. - , one-dimensional 3
- - in interfaces 215 - , real 1,4
- - , invariance 44 - space 121
- - , screw orientation 45f. - structure, classification 11
- systems 11
cell walls 121 f., 212 - , three-dimensional 3
chain molecules 4, 35 - , two-dimensional 3
climb 47f. crystalline interfaces, general theory
close packed hexagonal (cph) 1431·
structure 16, 25 f. - state 3
252 Subject Index

cubic 11 energy barrier 40

curl vector 123 - of a dislocation 55f.
- of stacking fault 33
deformation stacking fault 28f. - of su bgrain boundary 131 f.
determinant 235, 240 equilibrum concentration 38
diamond structure 20f. equivalence classes 148 f.
diffusion 38 expansion 239
dipole mechanism 95 extrinsic fault 30
dislocation, breaking away from a
network 138 face-centred cubic (fcc) structure
- core 41, 56 15,17,37
- , curvature 60 fibrous structure 34
- dynamics 117 forces on a body, external 51
- , ending at boundary 222 - - - - , internal S1
- line 411. - on dislocation 53
- , foreign 129 framework of stacking faults 94
- , geometrical 219 Frank's formula 1191., 155
- loops 113 - - , quantized 121
- networks 1181. - node condition 73, 98
- - , primary 209 Frank-Read mechanism 61
- - , secondary 209 FS/RH-convention 43f., 46
- , penetrating boundary 224
- , physical 219 generating function k 145
- , primary 219 geometrical fit 119
- reactions 72f., 102 - parameter 192
- - , in bcc structure 77 glide 47
- - , dualistic representation 98 - reflection planes 12
- - , in fcc structure 75 good material 42f., S6
- - , in hexagonal structure 76 grain boundaries 1, 188
- - , partial dislocations 107 graphite 3 f.
- - , stability 74 - structure 27f.
- , resistance to bowing out S9 growth stacking fault 28
-,secondary 219
dislocations, intersection 115 H-sequence 1 Sf.
- and moire figures 49 hardness 133
- and point defects 113 helical dislocation 113
divergency 123 heterogenious system 38
double vacancies 37, 40 hexagonal structure 11
DR-lattice 207 holes 1
DSC-lattice 208 honeycomb-like cells 123
DSE-lattice 207 hydrostatic pressure 54
dualism in three dimensions 139
dualities 99f. identity I 236
dumbbell 38 impurity atoms 1, 37
inclusions 1
eigenvalue 174, 194, 242 interaction of dislocations 70f.
eigenvector 174, 194, 242 interpenetrating lattices 122
Einstein's summation convention interstitial atoms = interstitials
7,229 1, 37
elastic behaviour 3 intrinsic fault 30
elementary cell 6 invariant elements of transfor-
- pattern shift lattice 207 mations 242
 Subject Index 253

inverse matrix 236 O-line 170f.

inversion 11 O-plane 173 f.
irradiation 38 O-point as origin 157
O-points 149, 155
jog 113 f. o 2-elements 209 f.
K-sequence 15f. 02-lattice 209f.
kink 113f. -, cells 212
octahedral holes 37
L-field 99 optimum phase boundary 188
L-node 99 orthogonal coordinates 8f., 229
L-sequence 15 f. orthogonality relation 230
L-space 98 orthorhombic structure 11
lattice coincidence site 143
- constants 7, 158, 230 parallel displacement = translation
line defects = dislocations 1 f. 5
- sense 42 parallelepiped 5
- tension 58 partial dislocations = partials 78
--vector 232 partials in the fcc-structure 78f.
linear transformations 238 partition 148
- - , nonhomogenious 12 pattern, displacement 207
lower indices 230 - elements 196
- of lattice points 196
matrix 8f. - repeat lattice 207
- calculation procedures 229 f. Peach and Koehler formula 54, 71
- inversion 236 Peierls-Nabarro force 115
- multiplication 231 pericline law 193
- , transposed 9, 231 periodic arrangement 5
mechanical deformation 38 - patterns, search for 198
metric tensor G 9f., 237 - substrate 41, 220
mica 3 periodicity of the pattern 196
minimum strain points = MS-points phase transformations 36, 8if.
123, 147, 156 physical fit 119
moire-models 243f. plastic behaviour 3, 117
- pattern 49 point of best fit 143, 155
molecular chains 3 - defects 1 f., 37f.
monoclinic structure 11 - - , formation energies 39f.
moonstone 189 - groups 11
motion of dislocations 47 - symmetry operations 11
movement rule 48, 50, 101, 218 Poisson's ratio v 57
pole mechanism 90f.
NaCl-structure 19 polygonization 133 f.
nearest nei~hbours 165 polyhedron 11
non-linear problems 221 precipitate 1, 38
nucleation of crystal grains 136£. projection onto b-subspace 180
nylon 3f. pseudo-vector 41, 45
O-elements 174 radius of curvature 137
O-lattice 148 f. rank of a matrix 236
- , applications 1861. - (T) 169f.
-,cells 182 reciprocal lattice 236
- and Frank's formula 153 recovery stage 133f.
- , reduced 196 recrystallization 1331.
- , translation 175 f. reference crystal 42
254 Subject Index

reflection 11 supersaturation 38, 113

- mechanism 90 symmetry operation lOf.
- at stacking faults 91 f.
related classes 148 T-matrix 169f.
remnants of stacking faults 94 tensor notation 232
rhombic structure 11 tetragonal structure 11
rhombohedral cell of NaCl-structure tetrahederal holes 37
20 thermal equilibrum 38
- cell of fcc structure 18 Thompson's notation 109
- structure 11 - tetrahedron 82f.
rotation R 11, 151, 239 three-dimensional defects
row-vector 232 three grain problem 221
tobermorite 34
S-matrix 8f., 158f., 201 f., 223 transformation A 147f, 158/.,
screw axes 12 160£.
sequence of partials 81 f. - of cobalt whiskers 95f.
shear 240 - , coordinate 9, 234
- modulus G 56f. - of fcc into cph structure 87f.
- strain 55 - , linear 9f.
silicate fibers 3 - of transformations 240
solid state matter 3 - , vector 9, 234
space groups 12 translation lattice 6, 10
specific surface energy 137 - group 148
- volume energy 137 - vector 6, 1 54
splitting of dislocation 80 triclinic structure 11
stabilisation of stacking sequence trigonal structure 11
30 tungstene 3, 5
stackig faults 1, 28 f. turbostratic arrangement 33
stacking fault energy 33f., 80 twin boundary 28, 219
- faults in diamond structure - law 167, 193
32f. two-dimensional defects
- sequence 14f., 78
- - , irregularities 28 f. unimodular transformation U
standard form 172f. 10, 162, 201 f.
strain field of dislocation 61 f. unit cell 6, 10, 165
stress, compressive 52 - vectors 10, 230
- field 51 upper indices 230
- - of dislocation 63 f.
-,shear 52 vacancies 1, 37
- , tensile 52 - , double 37,40
structure of crystals 3, 11 Van der Waals bonds 3,27
subdeterminant 235 vector transformation 234
subgrain boundaries 1181. virtual grain boundary dislocation
- - , classification 124/. 226
- - , energy 131 Volterra cut 41
- - , general 130 Wigner-Seitz cell 10f., 121, 167
- - , partial twist 127
- - , pure twist 125 yield strength 60
- - , special 125f.
- - , tilt 128 zig-zag reaction 104