High Resolution X-Ray Diffractometry and Topography 1998

High Resolution X-ray Diffractometry and
Topography
High Resolution X-ray
Diffractometry and
Topography
D.KEITH BOWEN
University of Warwick
and
BRIAN K.TANNER
University of Durham
This edition published in the Taylor & Francis e-Library, 2005.
“To purchase your own copy of this or any of Taylor & Francis or Routledge’s collection
of thousands of eBooks please go to www.eBookstore.tandf.co.uk.”
UK Taylor & Francis Ltd, 1 Gunpowder Square, London EC4A 3DE

USA Taylor & Francis Inc., 1900 Frost Road, Suite 101, Bristol, PA 19007
Copyright © D.Keith Bowen and Brian K.Tanner 1998

All rights reserved. No part of this publication may be reproduced, stored in a
retrieval system, or transmitted, in any form or by any means, electronic,
electrostatic, magnetic tape, mechanical, photocopying, recording or otherwise,
without the prior permission of the copyright owner.
British Library Cataloguing-in-Publication Data
A catalogue record for this book is available from the British Library
ISBN Master e-book ISBN
ISBN (Adobe eReader Format)

ISBN 0-8506-6758-5 (Print Edition)
Library of Congress Cataloging-Publication-Data are available

Cover image by Kevin M.Matney, Bede Scientific Incorporated
Cover design by Youngs Design in Production
Contents
Preface viii
1 Introduction: Diffraction Studies of Crystal Perfection 1

1.1 The response of a crystal to a plane wave 2
1.2 Comparison of radiations for diffraction 6
1.3 Comparison of surface analytic techniques 7
1.4 High resolution X-ray diffractometry 8
1.5 Triple-axis diffractometry 10
1.6 X-ray topography 10
1.7 Summary 12
References 13
2 High Resolution X-ray Diffraction Techniques 15
2.1 Limitations of single-axis diffraction 15
2.2 X-ray sources 16
2.3 High resolution systems 18
2.4 The duMond diagram 20
2.5 Beam conditioners 24
2.6 Aberrations in high resolution diffractometry 36
2.7 Detectors 43
2.8 Setting up a high resolution experiment 47
2.9 Summary 49
References 50
3 Analysis of Epitaxial Layers 51
3.1 Introduction 51
v
3.2 Plane wave diffraction from heteroepilayers 52

3.3 Selection of experimental conditions 54
3.4 Derivation of layer parameters 56
3.5 Summary 69
References 0
4 X-ray Scattering Theory 71
4.1 Introduction 71
4.2 Kinematical theory 72
4.3 Dynamical theory 82
4.4 Scattering in imperfect crystals 98
4.5 Formulae for rocking curve widths, profiles and intensities 102
4.6 Summary 108
References 108
5 Simulation of X-ray Diffraction Rocking Curves 111
5.1 Introduction 111
5.2 Numerical solution procedures 113
5.3 Matching an experimental measurement 118
5.4 Limitations of the simulation approach 120
5.5 Examples of simulations 122
5.6 Summary 130
References 0
6 Analysis of Thin Films and Multiple Layers 133
6.1 Review of rocking curves from thick layers 133
6.2 Thin single layers 135
6.3 Graded thick layers 137
6.4 Bragg case interferometers 139
6.5 Superlattices 140
6.6 General characteristics of large repeat superlattices 145
6.7 Summary 152
vi
References 0
7 Triple-axis X-ray Diffractometry 157
7.2 Instrumentation 158
7.3 Setting up a triple-axis experiment 160
7.4 Separation of lattice tilts and strains 161
7.5 Measurement of kinematical scattering from a defective surface 162
7.6 Measurement of distorted specimens 163
7.7 Full reciprocal space mapping 165
7.8 Applications of reciprocal space mapping 171
7.9 Summary 176
References 177
8 Single-crystal X-ray Topography 181
8.2 Contrast mechanisms 182
8.3 The Berg-Barrett technique 183
8.4 Lang topography 185
8.5 Experimental procedures for taking Lang topographs 187
8.6 Topographic resolution 189
8.7 Contrast on X-ray topographs 194
8.8 Theoretical tools for the interpretation of X-ray topograph 199
contrast
8.9 Dislocation and stacking fault images in X-ray topographs 206
8.10 Summary 215
References 216
9 Double-crystal X-ray Topography 219
9.2 Principles and experimental geometries 220
9.3 Defect contrast in double-crystal topographs 222
9.4 Spatial resolution 224
vii
9.5 Synchrotron radiation double-crystal topography 226

9.6 Applications of double-axis topography 227
9.7 Summary 231
References 232
10 Synchrotron Radiation Topography 233
10.2 Synchrotron radiation sources 233
10.3 Source requirements for synchrotron X-ray topography 236
10.4 Applicability of synchrotron radiation for X-ray topography 240
10.5 Techniques of synchrotron radiation topography 243
10.6 Summary 260
References 261
Index 263
Preface
The purpose of this book is to provide the theoretical and practical background
necessary to the study of single-crystal materials by means of high resolution X-
ray diffractometry and topography. Whilst some of these techniques have been
available for over fifty years, and the basic theory for even longer, it is in the last
decade that they have grown enormously in importance, essentially for two
technological reasons. On the one hand there has been the development of the
powerful sources for scientific and industrial research: dedicated electron storage
rings for synchrotron radiation, which have enabled the ideas and techniques
developed in the earlier years to be applied and extended. On the other hand has
been the industrial need for characterisation and control of the high-quality
crystals that now form the heart of so many of the devices used for the
electronics, communications and information engineering industries: integrated
circuits, sensors, optoelectronic and electroacoustic devices and
microprocessors. Techniques such as high resolution X-ray diffractometry, that
for half a century were rather obscure research tools, are now in daily use for
industrial quality control.
Educational establishments have not yet fully responded to this growth, and
the availability of trained ‘characterisation’ scientists and engineers is a limiting
factor in many industries that manufacture devices based upon high-quality
crystals. In consequence, many research or quality control laboratories are
staffed by people trained in different fields, who may be unaware of both the
potentialities and the pitfalls of the X-ray techniques. In this book we attempt to
redress these problems. We assume only a knowledge of the materials structures
and defects whose characterisation is required. We provide the elementary and
advanced theory of X-ray diffraction that is required to utilise the techniques;
besides covering the mathematical formalism we have taken pains to present the
conclusions of the theory qualitatively and visually, to make its consequences
more accessible. We discuss in depth the techniques of X-ray diffractometry and
topography, including both the practical details and the application of the theory
in the interpretation of the data. The limits of the techniques are explored
throughout, an attitude that takes particular technological importance in the light
of the present concern with very thin films and surf ace layers.
ix
We are pleased to acknowledge the contributions and the stimulation of our

graduate students and research associates over many years of research into
materials characterisation. The content and approach of this book emerged over
almost as many years of teaching advanced courses, beginning with a NATO
Advanced Study Institute project in 1979 and including workshops and courses
for the Denver X-ray Conference and the European Union COMETT and
TEMPUS schemes. We owe special gratitude to our colleagues at Bede
Scientific Instruments, where we are both directors, for funding the development
of these courses and, especially, for posing interesting and difficult problems for
us arising from the real needs of industrial scientists working with advanced
materials. Finally, we thank the European Synchrotron Radiation Laboratory and
the University of Denver for being our hosts during periods of intensive writing
on sabbatical leave.
D.Keith Bowen
Department of Engineering
University of Warwick
Brian K.Tanner
Department of Physics
University of Durham
x
1
Introduction: Diffraction Studies of Crystal
Perfection
In this chapter we introduce high resolution diffraction studies of

materials, beginning from the response of a perfect crystal to a plane
wave, namely the Bragg law and rocking curves. We compare X-
rays with electrons and neutrons for materials characterisation, and
we compare X-rays with other surface analytic techniques. We
discuss the definition and purpose of high resolution X-ray
diffraction and topographic methods. We also give the basic theory
required for initial use of the techniques.
The experimental beginning of the study of materials by diffraction methods took

place in 1911 when von Laue, Friedrich and Knipping 1 obtained the first X-ray
diffraction patterns, on rock salt. The theoretical foundation was laid between
then and 1917 by Bragg, 2 von Laue, 3 Darwin 4 and Ewald. 5 Many workers
then contributed to the steady development of the subject, and we highlight in
particular the work of Lang, Authier, Takagi, Taupin and Kato in developing the
experimental and theoretical tools for the study of distorted and defective
crystals by X-ray methods which form the main topic of this book.
In its first sixty or seventy years, two great methods dominated X-ray diffraction
research. Single-crystal diffraction for structure analysis revealed the crystal and
molecular structure of inorganic, organic and more recently biological
compounds, and is now an essential tool in the understanding of complex
processes such as enzyme action and in drug design. The powder diffraction
method has become indispensable for the practical analysis of materials ranging
from corrosion products to contraband. However, two developments in the early
1980s stimulated the explosive growth of high resolution X-ray diffraction,
which began in the 1920s but had remained a small, specialised method. The
push was the ample intensity available from synchrotron radiation sources, which
allowed the rapid exploration of many new X-ray scattering techniques. The pull
was the industrial requirement for rapid, non-destructive analysis of defects in
highly perfect materials for the electronics industry, which was beyond the
capability of the lower-resolution techniques. Interestingly, developments in X-
ray optics which were enabled by the availability of large, highly perfect crystals
2 HIGH RESOLUTION X-RAY DIFFRACTOMETRY AND TOPOGRAPHY
of silicon from the electronics industry have permitted the development of

laboratory-based instruments for rapid industrial materials characterisation.
The development of electron diffraction and microscopy was parallel but
separate. Although there are many obvious similarities between the theories, they
are usually expressed in different notations and it is common that workers trained
in the one have little knowledge of the other. This is particularly unfortunate in
that the techniques are almost always complementary, and many investigations
will benefit by the use of both electron and X-ray (or neutron) methods. The
emphasis in this book is on the X-ray methods. Electron techniques have been
amply covered in a number of excellent publications; 6 , 7 we have not
attempted to include any detailed description or assessment of them, but we shall
take the opportunity to draw parallels and complementarities between the
electron, X-ray and neutron beam methods.
1.1
The response of a crystal to a plane wave
Any radiation striking a material is both scattered and absorbed. Scattering is
most easily approached by thinking of a plane wave. This is formally defined as
one whose phase is constant over any plane normal to its direction of travel, its
wavefront is a plane. It is more easily thought of as a point source of waves at an
infinite distance; a perfectly collimated wave. When such a wave strikes a three-
dimensional atomic lattice, each scattering point (electron or nuclear particle)
acts as a source of spherical waves, whose wavefronts lie on spheres centred on
the scattering points. The addition of the amplitudes of all these waves in given
directions results in almost zero intensity in most directions but strong beams in
some directions if the wave-lengths of the wave are comparable with the spacing
of the scattering centres. This is the phenomenon of diffraction.
The simplest and most useful description of crystal diffraction is still that
obtained by Bragg. 2 Strong diffraction occurs when all the wavelets add up in
phase. By considering an entire crystal plane as the scattering entity, rather than
each individual electron, it is easy to see from Figure 1.1 that strong diffraction
results when
nλ =2dsinλ (1.1)
where n is an integer representing the order of diffraction, λ is the wavelength, d
the interplanar spacing of the reflecting (diffracting) plane and λ the angle of
incidence and of diffraction of the radiation relative to the reflecting plane. The
requirement for the angle of incidence to equal that of diffraction is not seen
directly from Figure 1.1, but arises from the incorporation of scattering from
many planes normal to the surface. A small number of planes give a very broad
peak, and large numbers of planes a narrow peak, converging to a value
characteristic of a thick crystal. Thus, diffraction for a given plane and
wavelength does not take place over the zero angular range defined by the Bragg
law, but over a small finite range. This range, called the rocking curve width,
DIFFRACTION STUDIES OF CRYSTAL PERFECTION 3
Figure 1.1 Diffraction of a plane wave off successive crystal planes. Strong diffraction
results when the angles of incidence and diffraction, λ , are equal and the path difference
AOB between the two beams is equal to nλ , an integral number of wavelengths. Hence
the Bragg law, nλ =2dsinλ
varies tremendously and it governs the strain sensitivity of the technique.
Examples of X-ray rocking curves for perfect crystals are shown in Figure 1.2.
Any radiation is also absorbed to a greater or lesser extent by a material. This
is well described by the usual absorption equation:
I/I 0=exp{− μ t} (1.2)
where I is the transmitted and I 0 the incident intensity, μ is the linear absorption
coefficient and t the specimen thickness in the direction of the X-ray beam.
The intensities of diffracted beams, or reflections as they are commonly
called, depend upon the strength of the scattering that the material inflicts upon
the radiation. Electrons are scattered strongly, neutrons weakly and X-rays
moderately. The basic scattering unit of a crystal is its unit cell, and we may
calculate the scattering at any angle by multiplying
■ the scattering strength of an electron or nucleus

■ the scattering strength of an atom
■ the scattering strength of a unit cell
■ the scattering strength of the total number of unit cells
all with regard to the direction of scattering and the relative phase of the
scattered waves.
Electron, nuclear and atomic scattering factors are either calculated or
measured and tabulated for X-rays. 8 The angular dependence of the intensity is
most usefully represented by calculating the scattering strength of the unit cell
for a particular lattice plane reflection, hkl, wavelength and, of course, crystal
structure. This most important parameter is the structure factor F hkl . This is a
fundamental quantity, which appears in all the expressions for diffracted
intensity, penetration depth and rocking curve width, and is calculated as follows.
Waves of vector k scattered from two points in the unit cell which are separated
by a vector r (see Figure 1.3), will have a phase difference
Figure 1.2 Calculated plane wave X-ray rocking curves. (a) Si 004 with CuK λ 1 (0.154
nm), FWHM=3.83 arcsec, (b) Si 333 with MoK λ 1 (0.071 nm), FWHM=0.73 arcsec, (c)
Ge 111 with CuK λ 1, FWHM=16.69 arcsec, (d) GaAs 004 with CuK λ 1, FWHM=8.55
arcsec
exp{−2λ ik.r}=exp{−2λ i(hu + kv + lw)} (1.3)
for the hkl reflection, where ( uvw ) are the fractional coordinates of the vector r.
We now need to add up the waves scattered from each atom with regard to the
atomic scattering factor of each atom and the phase of the wave from each atom
and sum this over the unit cell:
(1.4)
Figure 1.2 (cont.)

The f i are the atomic scattering factors of the atoms of type i. These depend
upon both λ and λ . The phases may cause the waves to add, as in for example Si
004, or cancel, as in for example Si 002. The larger the structure factor, the
broader is the rocking curve.
There is a relationship between scattering strength and absorption strength, in
that strong scatterers are also in principle strong absorbers. The parameters are
often combined, as will be seen in Chapter 4, into a complex structure factor in
which the real part represents the scattering factor (equation (1.4)) and the
imaginary part the absorption coefficient (equation (1.2)).
Figure 1.3 The addition of waves scattered by an angle 2λ from an atom at the origin and
one at a vector r from the origin. The wavevectors k 0 and k h are in the directions of the
incident and diffracted beams, respectively, and |k 0|=|k h|=1/λ
1.2
Comparison of radiations for diffraction
Later chapters will deal with a more complete description of the diffraction
process, but we now have enough to discuss the selection of radiations and
techniques. If the structure factor and scattering strength of the radiation are high,
the penetration is low and the rocking curve is broad. This is the case with
electron radiation. For X-rays and even more for neutrons, the structure and
absorption factors are small, penetration is high and rocking curves are narrow.
These factors have three main consequences for X-rays and also for neutrons:
1 The techniques are normally non-destructive, since adequate penetration and

coverage of industrial-scale samples is possible.
2 The strain sensitivity is high, since narrow rocking curves are strongly
influenced by small rotations caused by strains in the crystal.
3 The spatial resolution of diffraction imaging is poor, compared with
electrons, since small areas are not scattering enough radiation for good
imaging, and the high strain sensitivity means that defect images are very
broad.
Table 1.1 summarises the characteristics. Electron microscopy and diffraction

and X-ray topography and diffraction are complementary techniques in almost
every respect. The neutron techniques have applications similar to X-rays but
decisive advantages in some cases, such as the study of magnetic materials and
of very thick samples. The theory is well understood for all three. Two great
advantages of X-ray techniques are their convenience and non-destructive

nature; they
Table 1.1 Characteristics of electron, neutron and X-radiations

Characteristic Electrons Neutrons X-rays
Absorption high very low low
Penetration <1 μ m ~cm ~mm
Rocking curve width degree <arc second arc seconds
Strain sensitivity 10−3 10−7 10−7
Spatial resolution 1nm 30 μ m 1μm
Destructive? yes no no
Cost high very high medium
Convenience good poor good
may thus be performed before, say, sectioning for electron microscopy and the
results for both techniques combined, or they may be used for on-line quality
control.
1.3
Comparison of surface analytic techniques
The penetrating power of X-rays is a major factor in their applicability to
problems such as crystal growth. However, there is strong and increasing
demand for methods of characterisation of very thin layers of material. Devices
are now being marketed that depend for their operation on the existence of a
controlled layer of material a few nanometres thick. If such layers are near the
surface, as they often are for reasons of manufacture, they may also be
characterised by X-ray methods using the new grazing incidence methods. This
term is used for angles of incidence close to the critical angle for total external
reflection, in practice, up to a few degrees. As seen in the previous section, the X-
ray methods are particularly sensitive to strain (as are many modern devices) and
it is relevant to compare the applicability of strain-sensitive surface techniques.
Figure 1.4 shows a comparison of the application range of the major strain-
measuring techniques that are sensitive to strains in the top micrometre of
crystalline material. The important role of X-ray methods is clear, even though
the methods lack the spatial resolution that can be achieved with charged particle
probes.
Figure 1.4 Comparison of the application ranges of techniques that are sensitive to near-
surface strains. Minimum detection limits are plotted against depth resolution of the
measurement. XRD: X-ray diffraction; DOR: differential optical reflectometry. RBS:
Rutherford back scattering; MEIS: medium energy ion scattering; TEM: transmission
electron microscopy
1.4
High resolution X-ray diffractometry
The term ‘diffractometry’ here means the measurement of the rocking curve of a
sample. It is always necessary to define the incident radiation, both in
wavelength and divergence, and a clear understanding of the latter is particularly
important. The reference is always the plane wave rocking curve, such as would
be measured with a perfectly parallel, monochromatic incident beam, and
theoretical calculations are based upon this imagined radiation. Of course, any
real radiation has both a wavelength spread and a divergence, and a good
approximation to a plane wave curve is only found if the broadening effects of
these are small compared with the width of the theoretical plane wave rocking
curve. The examples in Figure 1.4 make this clear. The rocking curve widths
range from 0.6■ to 12■ . The divergence of a good synchrotron radiation beam
Figure 1.5 The symmetrical double crystal (+, −) setting for measuring rocking curves
could be about 1■ , and that of a sealed tube X-ray source at 1m is about 80■ .
Synchrotron radiation is a continuous spectrum and single-crystal rocking curves
map the spectrum. The rocking curve obtained in all other cases will be
dominated by the source profile, little influenced by the specimen. In other
words, the instrument function for conventional powder and single-crystal
diffractometers is far too great to measure subtle changes in rocking curves of
nearly perfect crystals.
Better conventional collimation will not do, except for the largest synchrotron
radiation installations; to obtain sub-arc-second collimation in the laboratory
would require a collimator some 100m long with a sealed-tube source, and at this
distance the intensity would be impracticably low. The problem is solved by the
use of a beam conditioner, which is a further diffracting system before the
specimen. The measured rocking curve is then the correlation of the plane wave
rocking curves of the beam conditioner and the specimen crystals, from which
most of the diffracting characteristics of the specimen crystal may be deduced.
The simplest conditioner is a perfect crystal of the same type as the specimen,
using the same reflecting planes, with the deviation of the diffracted beam in the
opposite sense to that at the specimen. This is the classic ‘+, − symmetrical
double crystal method’, as shown in Figure 1.5, which gives excellent and easily
interpreted results. Many variations are, however, possible, for example to
maximise the sensitivity to strain, or to emphasise the contribution of near-
surface layers to the diffraction, and we shall treat these in detail in this book.
The great utility of rocking curve measurements is based upon two
fundamental properties:
1 The details of the rocking curve are extremely sensitive to the strains and
strain gradients in the specimen.
2 For a given structural model, the rocking curve may be computed to high
accuracy using fundamental X-ray scattering theory.
In addition, the measurements are rapid and simple, and are now even used in
100% inspection for quality control of multiple-layer semiconductors. An
example is shown in Figure 1.6. This is a GaAs substrate with a ternary layer and
a thin cap. The mismatch between the layer and the substrate is obtained
immediately from the separation between the peaks, and more subtle details may
be interpreted with the aid of computer simulation of the rocking curve. This
curve can be obtained in a matter of minutes. Routine analysis of such curves
gives the composition of ternary epilayers, periods of superlattices and
thicknesses of layers, whilst more advanced analysis can give a complete strain
and composition profile as a function of depth.
1.5
Triple-axis diffractometry
Triple-axis diffractometry is an extension of high resolution diffractometry in
which an analyser crystal (or crystals) is placed before the detector in order to
restrict its angular acceptance. This has the effect of separating the effects of
strains and tilts on the measurement, and permitting the measurement of diffuse
scatter. In reciprocal space notation, which will be explained in Chapter 5, the
intersection of the angular collimation of the beam conditioner crystal(s) and the
analyser crystal(s) in reciprocal space defines a small volume of reciprocal space
that is sampled by the detector at a given setting. Thus, complete two-
dimensional reciprocal-space maps may be obtained, giving very detailed
information on thin layers and surface conditions.
In later chapters we shall give both the fundamental theoretical treatment
necessary to understand the complex rocking curves that arise from complex
structures, and the practical experimental details required to measure them
reliably and unambiguously.
1.6
X-ray topography
The essence of the topographic methods is that they map the intensity of the
diffracted beam over the surface of the crystal. Defects affect the diffracted
intensity, so give contrast in the image. The methods are quite sensitive enough
to reveal individual dislocations, precipitates, magnetic domains and other long-
range strain fields but cannot reveal point defects except in dense clusters.
1.6.1
X-ray topography with conventional sources
There are many methods of X-ray topography, though the most popular are the
Lang method with slit-collimated radiation and the double-crystal methods,
which may be thought of as high resolution diffractometry with an imaging
Figure 1.6 (a) An example of a measured rocking curve, for an epitaxial layer 17 nm
thick of In0.18Ga0.82As, grown on GaAs, with a 0.1 μ m layer of GaAs on the top. CuK λ 1
radiation, GaAs single-crystal beam conditioner. (b) A theoretical simulation of this
structure, by means of which the composition of the InGaAs layer was determined to <1%
and its thickness to 0.1 nm
detector. The principles of image formation and diffracted intensity are common,
and may be applied to any method. The two examples show opposite extremes.
Figure 1.7 shows a high resolution image of a silicon wafer, taken on an ultra-
high resolution photographic plate, in which details of the strain fields of
individual dislocations may be observed. Figure 1.8 shows a high-speed image
taken on a television detector in which the broad features of the defect density
and distribution are quickly visualised.
Figure 1.7 A Lang topograph (in transmission) of a silicon wafer. MoK λ 1 radiation (0.07
nm), conventional X-ray tube, IIford L4 ultra-high resolution plate. Field size 3 mm by 3.
3 mm 022 reflection. Diffraction vector in direction up the page. The black lines are
images of individual dislocations
1.6.2
Synchrotron radiation topography
The very high intensities and good collimations available make synchrotron
radiation very suitable for topography. Against this must be set the
inconvenience and slow turn-round necessitated by transport to a central
synchrotron radiation laboratory. Hence, the technique is most appropriately used
in the following cases:
1 For dynamic experiments, such as modelling of crystal growth or circuit

processing.
2 For extensive survey topography, where the time taken on conventional
machines would be prohibitive.
3 For obtaining information inaccessible to conventional methods, e.g.
distorted crystals, stroboscopic topography, very thin layers, anomalous
dispersion, simultaneous topography and fluorescence.
4 For rapid research into new methodologies and techniques, which may in
some cases be transferred later to laboratory systems.
1.7
Summary
Crystals diffract radiation of comparable wavelength to the atomic spacing, as
described by Bragg’s law. The range of diffraction, or reflection, is of the order
Figure 1.8 A double-crystal topograph of a GaP crystal, taken with CuK λ 1 radiation
from a conventional 1 kW X-ray tube. Direct imaging X-ray detector (Bede Scientific), 1
minute exposure. Cracks, sub-grain boundaries and variations in dislocation density are
seen
of seconds of arc for X-rays, and depends on the structure factor; this in turn
depends upon the atomic species and their arrangement and indicates the strength
of reflection from a given crystal plane. X-rays may probe millimetres,
sometimes centimetres, into a material, or, using grazing incidence methods,
penetration can be restricted to nanometres to make the techniques surface-
sensitive. High (angular) resolution methods measure the reflection curve of a
material to a pseudo-plane wave. This curve may be simulated using diffraction
theory and a model of the material structure to determine this structure.
Topographic methods give a direct image of many types of crystal defect and
may be used to assess crystal perfection with high (strain) resolution.
References
. W.FRIEDRICH, P.KNIPPING & M.VON LAUE, Proc. Bavarian Acad. Sci., 303
(1912).
. W.L.BRAGG, Proc. Roy. Soc. A, 89 , 248 (1913).
. M.VON LAUE, Ann. der Physik, 56 , 497 (1918).
. C.G.DARWIN, Phil. Mag, 27 , 315 (1914).
. P.P.EWALD, Ann. der Physik, 54 , 519 (1917).
. P.B.HIRSCH, A.HOWIE, R.B.NICHOLSON, D.W.PASHLEY & M.J.WHELAN,
Electron microscopy of thin crystals (Butterworths, London, 1965).
. M.H.LORETTO & R.E.SMALLMAN, Defect analysis in electron microscopy

(Interscience, New York, 1975).
. International tables for X-ray crystallography, Vol. 3 (Reidel, Dordrecht, 1985).
2
High Resolution X-ray Diffraction
Techniques
We first define the geometry and instrumental parameters common

to high resolution diffractometry. As a reference, we then develop
the duMond diagram for visualisation of X-ray optics and use it to
discuss practical beam conditioners. Next we treat the principal
aberrations of high resolution diffractometry: tilt, curvature and
dispersion. We discuss the requirements on X-ray detectors, and
finally show how to set up a high resolution measurement in
practice.
2.1
Limitations of single - axis diffraction
As we have seen in Chapter 1, we need something near a plane wave in order to
see the finest details of the specimen structure. A single-axis diffractometer
utilises a beam that is very far from a plane wave. Thus, single-crystal rocking
curves are broadened due to the beam divergence, and the spectral width of the
characteristic X-ray lines.
The divergence is a function of the source size h and slit size s and the source-
specimen distance a. As seen in Figure 2.1,
(2.1)
Thus, for a typical case of h=0.4 mm, s=1 mm, a=500 mm, λ λ ~500 arc seconds,
far above the width of the rocking curve for highly perfect crystals, which is
typically a few arc seconds.
The spectral width of the X-ray characteristic lines is approximately λ λ /λ ~10−
4. This rises to 10− 3 if both the K
λ 1 and K λ 2 ines are diffracted by the
specimen. The effect this has upon the rocking curve depends on the dispersion
of the whole system of beam conditioner and specimen, and ranges from zero to
very large. This will be discussed below, in section 2.6.
We therefore need to limit the divergence and wavelength spread of the beam
incident upon the specimen. Beam conditioners are used to collimate and to
16 HIGH RESOLUTION X-RAY DIFFRACTION TECHNIQUES
Figure 2.1 The divergence of a single-axis system
monochromate the beam. We see that the requirement is for a system with high
angular resolution and sufficient monochromatisation. We shall first discuss the
characteristics of sources and the general features of high resolution diffraction
systems, then analyse the requirements for beam conditioning.
2.2
X-ray sources
Electromagnetic radiation is generated by the acceleration or deceleration of
charged particles. If the acceleration is sufficiently high and the particles
sufficiently energetic, the radiation is in the X-ray regime, forming a continuous
spectrum. In addition, transitions of electrons between energy levels in atoms
cause emission of radiation of wavelength λ =hc/■ E, where h is Planck’s
constant, c the velocity of light, and ■ E the energy difference between the levels.
This gives rise to a line spectrum, whose most energetic components are in the
hard X-ray regime for most elements. Both laboratory and synchrotron radiation
sources are in common use for high resolution diffractometry.
2.2.1
Laboratory generators
In a laboratory generator, electrons are accelerated by a potential around 30 kV
towards a solid target, where they are stopped by impact. The output contains the
line spectrum superimposed upon a continuous spectrum. The line, or
characteristic spectrum is characteristic of the element and is used in X-ray
fluorescent analysis to identify the type and amount of an element present in a
sample. The continuous radiation is also called the Bremsstrahlung, from the
HIGH RESOLUTION X-RAY DIFFRACTION TECHNIQUES 17
Figure 2.2 Spectrum of X-ray emission from a copper target at various excitation
voltages
German for ‘braking radiation’, as it is emitted when the electrons are ‘braked’
by the solid. The complete spectrum from a copper target as a function of
accelerating voltage is shown in Figure 2.2. It is clear that the characteristic
radiation is far more intense than the continuous, and it is used almost
exclusively in high resolution diffractometry.
The characteristic lines are labelled K, L, M, etc., according to the label of the
electron shell to which the transition occurs, with the subscript λ , λ , etc.
indicating where the transition started. Thus the K λ lines result from transitions,
the K λ line from transitions, and so on. By far the most popular choice for high
resolution diffractometry is the CuK λ 1 line, because of its convenient
wavelength (0.154 nm), which is useful for most interplanar spacings, and
because of its high intensity. The copper target is the most easily cooled of all
metallic targets, thanks to its high thermal conductivity, and the output of an X-
ray target is limited only by heat dissipation. The cooling may be by air, water or
a combination of these with a rotating target to spread the heat more effectively.
X-ray tubes run from fractions of a milliamp for a microfocus source up to about
an amp for a very high power rotating target tube.
The K λ line is a doublet with separation about 10− 3. It is often important to
remove the K λ 2 line, which has approximately half the intensity of the K λ 1, by
the use of a beam conditioner as shown later.
2.2.2
Synchrotron radiation sources
Synchrotron radiation sources remove the limitation of the thermal properties of
solids by confining the electron beam in an ultra-high vacuum. Magnetic fields
are used to bend the beam into a closed ‘ring’ (in fact, a polygon with rounded
corners), and the installations are called electron storage rings. Since the
phenomenon was first created on earth in synchrotrons (although it had been first
postulated for explaining astronomical X-ray emission), the name synchrotron
radiation has stuck. A storage ring is simply a synchrotron run at stable rather
than cyclic energy and magnetic field conditions. The installations are large, with
diameters up to ~100 m, and are built on a national or international scale. The
radiation spectrum is shown in Figure 2.3. All synchrotron radiation sources from
simple bending magnets show this spectrum, with changes only in the vertical
and horizontal position of the curve on the axes.
Since the electrons in storage rings are travelling at relativistic speeds, the
emission of electromagnetic radiation is foreshortened into a cone whose axis is
the instantaneous direction of motion of the electron. The radiation is therefore
intrinsically collimated and is a good match to the subsequent beam conditioner.
This contrasts favourably with a laboratory source, in which very little of the
more-or-less isotropic emission reaches the specimen. The principal
characteristics of synchrotron radiation are:
■ continuous radiation spectrum

■ high intensity and brightness
■ high degree of polarisation
■ pulsed time structure, typically at megahertz frequencies.
The sources are invaluable for the tunability of the radiation, that is where
spectroscopic as well as scattering properties are important, and for experiments
requiring the polarisation and time structure. However, with recent advances in
X-ray tubes, beam conditioners and detectors, many scattering experiments are
just as well performed in the convenience of the laboratory. Although it is difficult
to attain the same intensities in the laboratory, it is in fact easier to achieve good
signal-to-noise ratios. If CuK λ 1 is suitable for the experiment, it is likely that
better productivity will be obtained with a laboratory source.
2.3
High resolution systems
Figure 2.4 shows the elements of a high resolution diffractometer. The beam
conditioner controls the divergence and wavelength spread of the beam by a
combination of diffracting elements and angular-limiting apertures. The latter
may also control the spatial width of the beam. This falls upon a specimen,
Figure 2.3 Synchrotron radiation spectrum: spectra for various beamlines at the Photon
Factory
which must be adjustable in tilt so that the normal to the reflecting planes lies
accurately in the plane of the diagram (assuming that the beam conditioner has
been similarly set). The specimen is carried on an axis that may be adjusted to a
precision of at least one arc second. The diffracted beam enters a detector, which
accepts all X-rays scattered off the specimen.
The instrument shown in Figure 2.4 is a double-axis instrument. The first axis
is the adjustment of the beam conditioner, the second is the scan of the specimen
through the Bragg angle. It is irrelevant to this definition that a practical
diffractometer may contain a dozen or more controlled ‘axes’, for example, to
tune and to align the beam conditioner, to locate the specimen in the beam, to
align and to scan the specimen and to control slits. It is the differential
movement of the two main axes that make the measurement and determine the
precision and accuracy of the instrument. This is the basic high resolution
diffractometer, which is now widely used for measurements of crystal
perfection, epilayer composition and thickness.
In the double-axis, high resolution diffractometer, the detector integrates the
intensity scattered by the specimen over its acceptance angle. More detailed
information can be obtained by collimating the detector with another X-ray
optical element, at the expense of intensity. This results in the triple-axis
instrument (Figure 2.5) in which the differential movements of the beam
conditioner, specimen and analyser make the measurement and determine the
precision. The triple-axis instrument is growing in popularity for research
applications because of the detail that it provides, but characterisation for quality
control is usually better performed on an optically simple double-axis high
resolution diffractometer because of its simpler operation, higher intensity and
more rapid measurement.
X-ray topographic systems are simply high resolution diffractometers in which
the beam is made large enough to cover the area of crystal required, or in which
Figure 2.4 Schematic of high resolution double-axis instrument
Figure 2.5 Schematic of high resolution triple-axis instrument
the specimen is scanned through the beam. An imaging detector is used to record
the topograph. In this case sufficient collimation may be achieved by a small
source and slits, as in the Lang method, but higher strain sensitivity is achieved
by X-ray diffracting optical elements. In all the topographic methods, the
diffracting power of the crystal is mapped across its surface.
2.4
The duMond diagram
Many types of beam conditioners have been designed, since the original double-
crystal experiments of Ehrenberg, Ewald and Mark 1 in the early 1920s. These
Figure 2.6 Si 220 reflection. (a) duMond diagram showing the wavelength-angle coupling
imposed by the Bragg law, (b) the corresponding real-space geometry
may have between one and twelve diffracting elements, different geometries,
planar, circular, elliptical, paraboloidal or logarithmic surfaces. With such a
bewildering variety it is essential to have some simple means of analysing their
performance in order to decide between the available choices. It is logical to
consider each element in turn and to see how its output is related to its input.
This and the next section are primarily for reference, but are valuable in
determining the choice of system and when understanding the details of
performance is important.
A straightforward graphical representation invented by duMond 2 in 1937
remains extremely useful. The wavelengths and angles diffracted by a crystal must
obviously satisfy Bragg’s law, and duMond realised that a plot of wavelength
versus incident angle would show the diffraction geometry of the crystal
reflection. This is shown in Figure 2.6(a), and is obviously part of a sine curve.
In effect, the crystal couples the wavelengths and the incident and scattered
angles, as seen in Figure 2.6(b).
Next we consider the input to the crystal. A plane monochromatic wave is

represented by a single point on the wavelength axis, and will diffract at a single
angle, as shown in Figure 2.7. The horizontal angular axis may also give
information, since an aperture before the crystal will impose a certain range of
angles that may be accepted by the crystal. In Figure 2.7 a wide aperture is
centred on the incident beam and the crystal is at the exact Bragg angle.
However, in practice we work with a source that is polychromatic and
divergent. In Figure 2.8(a) we show, superimposed on a section of the duMond
diagram, the K spectral lines for copper radiation. The diagram shows the angles
at which the CuK λ 1 and CuK λ 2 lines will diffract, and the angular aperture
settings (controlled by equation (2.1)) that would be used to exclude CuK λ 2
from the subsequent beam. This aperture is now only 0.06°. Figure 2.8(b) shows
the real-space correspondence.
The integrated intensity of the reflection will depend also on its angular width.
Therefore, the plot can be thought of as made of two lines, separated by the
rocking curve width. This is a function of wavelength, Bragg angle and structure
factor. On the scale of Figure 2.6 these lines are too close together to be seen, but
the figure suffices to discuss a single reflection. It is now seen quite easily why
single-crystal diffractometers do not have enough angular resolution to measure
the fine structure of the rocking curve.
We now add another reflection. The sense of a reflection in relation to
the previous reflection is important. The notation used for multiple reflections is
as follows:
■ The first reflection is given the notation + n.

■ The next reflection is given the symbol n if the planar spacing is identical to
the previous reflection (same material, same plane), and m if it is different.
■ This reflection is given the positive sign if it deflects the beam in the same
sense as does the first crystal, the negative sign if the opposite sense.
Thus the commonly used channel-cut crystal with two reflections has the
notation (+ n, − n ). The formal duMond diagram notation is as follows:
■ The λ (λ ) plot for the first crystal is drawn starting from the origin. This
defines the direction of the incident beam and hence the zero for angular
rotations of the crystals.
■ With the beam entering from the left and the specimen beneath the beam, the
crystal diffracts when rotated anticlockwise through the Bragg angle.
■ The λ (λ ) plot for the second crystal is also drawn starting from the origin.
However, there are two possibilities for sign.
— If the second crystal has the opposite sense of deflection from the first (a+n
followed by a–n or – m) then the plot goes in the same direction as that of the
first crystal. The origin is the same.
Figure 2.7 Si 220 reflection with a plane incident monochromatic wave. (a) duMond
diagram showing the angle at which the wave will diffract, and a 2° angular aperture that
easily allows it to pass, (b) the corresponding real-space geometry
• If the second reflection is – n then the plots superimpose and no rotation is

necessary for diffraction to occur from the second crystal, i.e. the crystals
remain parallel. A small rotation causes the curves to separate and diffraction
stops (Figure 2.9). This is also called the non-dispersive setting.
• If the second reflection is −m then the plots diverge and a rotation is
necessary for the second crystal to diffract (Figure 2.10). The reflecting range
of the system comprising source, aperture and first crystal is defined as
discussed above, and diffraction from the second crystal occurs when its
curve on the duMond diagram overlaps with some part of this range.
— If the second crystal has the same sense of deflection from the first (a+n
followed by a+n or + m) then the plot goes in the opposite direction as
that of the first crystal. The origin is again the same. However, the second
crystal will not diffract until it has been rotated by 2λ B; again, the possible
operating ranges of the second crystal must overlap the actual operating
range of the first crystal (Figure 2.11).
■ The intensity of the combined reflection is proportional to the area of overlap

of the rocking curves in the duMond diagram if the source intensity is uniform
across the relevant wavelength region.
■ Figure 2.12 shows the relevant region of the duMond diagram for the (+ n, −
n ) and (+ n ,+ n ) systems at much greater magnification. The shaded area is
the overlap. This takes the plane wave rocking curve to be a square function,
which, as will be seen in Chapter 4, is not a bad approximation.
■ To allow for the profiles of both the reflecting curve and the source spectrum
then we integrate the product of area element taken at constant wavelenth A (
λ ) with the intensity I ( λ ) at that wavelength and the reflectivity R ( λ , λ ) of
each crystal, i.e.
(2.2)
The rocking curve resulting from scanning the second crystal in angle is the plot
of the above intensity against angle. Mathematically this is a cross-correlation as
discussed below.
The duMond diagram notation does not purport to describe quantitatively the
complete intensity-wavelength-divergence transfer function of a beam
conditioner. It is, rather, a highly useful qualitative and semiquantitative tool
which greatly aids the visualisation and understanding of multiple diffracting
elements. It has been extended to a full quantitative description by Matsushita 3
in his phase-space approach. This is of particular interest to engineers designing
beam conditioners, but we do not require such detail in order to understand and
to interpret high resolution diffractometry.
2.5
Beam conditioners
We shall discuss beam conditioners in some detail, since no general review
applicable to high resolution diffractometry contains all the recent developments,
though earlier reviews by Beaumont and Hart, 4 Kohra et al . 5 and Wilkins and
Stevenson 6 are useful. Beam conditioners may in principle contain the
following elements:
Figure 2.11 The (+n, +n) setting. (a) Real-space geometry, (b) duMond diagram when
the second crystal is parallel to the first and does not diffract, (c) duMond diagram when
the second crystal is rotated anticlockwise by 2λ B and the second crystal diffracts. Si 220
with CuK λ
Figure 2.12 An enlarged view of duMond diagrams near the overlapping regions bounded
by the aperture. (a) (+ n,−n) setting (the curves overlap completely), (b) (+n,−m ) for InP
004 and Si 004 with CuK λ 1 (see also Figure 2.10), (c) (+n,+n ) setting aligned for
maximum intensity. Si 220 with CuK λ 1. The hatched areas show the band passed. Note
that case (a) gives no monochromatisation, (b) gives poor monochromatisation, (c) gives
excellent monochromatisation
1 apertures
2 Bragg reflections off crystals or multilayers
3 specular reflections off plane or curved surfaces.
(1) and (2) are of interest in high resolution work with laboratory sources, though
specular reflections form important components of beam conditioners at
synchrotron radiation sources, especially with softer radiations. Apertures are
simple limitations of the lower and upper bounds of angle accepted by the beam
conditioner, whose effect can be seen directly on the duMond diagram. The
complexity usually lies in the design of the diffracting elements.
2.5.1
The non-dispersive setting
For the (+ n, − n ) setting the duMond diagram is identical for both reflections,
since the crystals and reflecting planes are identical, as seen in Figure 2.9. The
region of overlap defines the wavelengths and angles for which diffraction by
both crystals is satisfied. Note that when the reference and specimen crystals are
parallel, all wave-lengths will diffract simultaneously. Each wavelength picks
out its own angle for Bragg reflection from the first crystal. If the crystals are
parallel, then the same Bragg angle will be satisfied for the specimen. Thus the
setting is non-dispersive in wavelength. A very small displacement of one crystal
with respect to the other will result in no wavelength satisfying the double
diffraction condition (Figure 2.9(c)).
The intensity at any angular setting resulting from the two reflections
corresponds approximately to the area under the overlap of the two perfect
crystal duMond diagrams, Figure 2.12. More precisely, the intensity of each
‘ray’ reflected by the first crystal is multiplied by the reflectivity of the second
crystal for the angle at which the ray strikes the second crystal; the overall
intensity is obtained by integrating over all rays (small elements of solid angle).
Mathematically, this is the correlation of the two perfect crystal reflecting curves
calculated from diffraction theory. If R 1 (λ ) and R 2 (λ ) are the reflectivities (in
intensity) of the first and second crystals as functions of the angle of incidence λ ,
then the total double crystal reflectivity R ( λ ) at any angle λ of the first crystal
relative to the second crystal is given by 7
(2.3)
The denominator (normalising constant) is the integrated reflectivity of the first

crystal. Figure 2.13 shows the plane wave and the double-crystal rocking curve,
again for Si 220 with CuK λ 1. We note the following:
■ The double-crystal rocking curve is only about 40% broader than the perfect
crystal reflecting curves. It can be as low as, or less than, an arc second for
high-order reflections, short wavelength radiation and highly perfect crystals.
■ The peak and integrated intensities of the double-crystal rocking curves are
comparable with those of the plane wave reflectivity curve. Multiple-
reflection beam conditioners can therefore be quite efficient if care is taken to
match the divergences and acceptance angles.
The double-crystal rocking curve is symmetric, though the plane wave

reflectivity curve is not. This is a consequence of the autocorrelation, since the
autocorrelation of any function is an even function.
Note that we use the characteristic lines only because they have high intensity.
We can work with the Bremsstrahlung but the intensity is low. However, it is a
way of exploiting wavelengths that do not correspond to characteristic lines.
If the second crystal is the specimen rather than a beam conditioner element,
we shall have got close to the aim of measuring the plane wave reflectivity of a
material. The narrow rocking curve peaks permit us to separate closely matched
layer and substrate reflections and complex interference details, as already seen
in Figure 1.6. The sensitivity limit depends on the thickness of the layer but for a
1 micrometre layer it is about 50 ppm in the 004 symmetric geometry with GaAs
and CuK radiation. This method has been used extensively to study narrow
crystal reflections since the invention of the technique.
Another way of looking at the traditional double-crystal method is that it
measures the difference in reflectivity between the specimen and a reference
perfect crystal. This is the original high resolution method and is still the best
choice when good reference crystals are available and the specimens to be
Figure 2.13 The plane wave and double-crystal rocking curves for Si 220 with CuK λ 1
examined are of the same material. However, it does have some practical
drawbacks:
1 If the specimen is changed to a diff erent material or reflection then the

reference crystal should also change to match. This is inconvenient if
required very frequently, for example in a research laboratory. In a
production-control environment, however, it is not important.
2 The wavelength reflected varies across the width of the beam. If the
specimen is curved, the rocking curve will show separate K λ 1 and K λ 2
peaks, even though the arrangement is non-dispersive with plane crystals. A
narrow slit is then required to eliminate K λ 2.
3 The rocking curve from a single reflection has rather broad tails, as seen in
the logarithmic plot of Figure 2.14. This can obscure subtle features of the
rocking curve of the specimen.
Accordingly, a number of beam conditioners have been developed which both

collimate and monochromate the beam, for general-purpose high resolution
diffractometers.
Figure 2.14 The plane wave rocking curve of Si 220 with CuK λ 1, showing the tails on a
logarithmic scale
2.5.2
CCC design
The problem of the broad tails of the rocking curve was attacked by Hart and
Bonse, who pointed out that successive (+ n, − n ) reflections narrow the rocking
curve without much affecting its height, since the peak reflectivity is near unity
(Figure 2.15). They also invented a simple method of making two parallel,
perfectly aligned crystals: cut a channel out of a large single crystal with a
diamond saw. This began the trend of monolithic X-ray optical elements, which
now include water-cooled ‘hot optics’ versions for extremely powerful
synchrotron radiation sources, 8 and scanning X-ray interferometers with one
reflection and three transmission crystals, one scanning in angle and one in
translation. 9
These devices only work if large single crystals of low X-ray absorbance can
be obtained. Silicon is almost ideal, and is economical thanks to the electronics
industry. It is preferable to use float-zone (FZ) material rather than Czochralski
(CZ) since the latter contains some inhomogeneities from its oxygen content.
Figure 2.15 shows the rocking curves produced from one, two, three and four
successive reflections in a channel-cut Si 220 crystal for CuK λ radiation, on a
logarithmic scale. The effect on the tails is dramatic. Although the integrated
Figure 2.15 The rocking curves produced from one, two, three and four successive
reflections in a channel-cut Si 220 crystal for CuK λ radiation. The inset shows that little
intensity is lost near the peak, but the tails are greatly reduced
intensity is reduced, most of the reduction takes place in the tails of the curve,
where it is adding to noise rather than signal. Germanium is also available in
good enough quality, though it is very much more expensive. It is more
absorbing, but its broader rocking curve gives higher intensity at the expense of
resolution. Diamond is used for some hot optics applications, because of its
unrivalled thermal properties. Note that these curves assume ideal crystallinity,
and the effects of thermal vibration and point defects mean that the curves will
not drop as rapidly below a reflectivity of about 10−6.
The channel-cut collimator (CCC) is not a monochromator. It is non-
dispersive in wavelength and all wavelengths that hit the first crystal at their
Bragg angle will reflect from the second crystal. The X-ray optics and the
duMond diagram are identical to the (+ n, − n ) setting of the double crystal
method, shown in Figure 2.9, except that the two crystals are cut out of one
monolith. The spectral bandpass is determined simply by the angular aperture,
controlled by a slit before or after the crystal or (for example, with some
synchrotron radiation beam lines) by the natural divergence of the source.
The addition of a single monochromating crystal gives the device great
flexibil ity, 10 since one has independent control over the collimation, by the
number of reflections, and the bandpass, controlled by the width of the reflection
of the monochromating crystal. Figure 2.16 shows one variant, a three-reflection
Figure 2.16 The CCC with a monochromating crystal, developed by Tanner and Bowen.
The four-reflection Si 220 with single Ge 200 monochromating crystal is shown, for CuK
λ 1
CCC to give very good angular resolution, plus a Ge 220 monochromating

crystal which has a high bandpass.
2.5.3
duMond-Hart-Bartels design
One of the earliest designs was proposed (but rejected!) by duMond himself but
only realised by Hart and popularised in diffractometry by Bartels. It comprises
four reflecting crystals, in the (+ n, − n, − n, + n ) arrangement, as shown in
Figure 2.17. If the gaps between the crystals are equal, the beam is undeviated in
height on exit from the conditioner. Slits between the crystal are not part of the
collimation, but serve to reduce any scatter. This arrangement is convenient for
diffractometer design but is difficult to adjust if four independent crystals are
used. Beaumont and Hart, 4 and Bartels 11 used Hart’s channel-cut crystals to
make a simple design. Using germanium, the 220 reflections may be selected for
high intensity, or the 440 for high resolution. Formally the duMond-Bartels
design is equivalent to a CCC with three parallel reflections and a
monochromating reflection, all of the same material and reflecting plane. It
would be possible to use different materials for the first and second monoliths to
make a different compromise between intensity and angular resolution, but this
loses the simplicity of the device.
2.5.4
Asymmetric CCC and duMond
The use of germanium rather than silicon gives higher intensity for the same
geometry but has the attendant practical disadvantages of higher cost (about ten
times), lower availability, greater fragility of the material and almost 50% losses
from polarisation effects when the Bragg angle is near 45° as it is for the 440
reflection. Loxley, Tanner and Bowen 12 have further developed the CCC and
duMond beam conditioner designs by using the principle of asymmetric
diffraction. As shown in Figure 2.18, the rocking curve is broadened by an
asymmetric reflection in comparison to the symmetric reflection. The reasons for
Figure 2.17 The duMond-Hart-Bartels ‘four crystal’ (+n, −n, −n, +n) beam conditioner
design
this will be seen when the theory is discussed in more detail in Chapter 4. For
now, we simply observe that the rocking curve width in the asymmetric case is
given by
(2.4)
where λ h and λ 0 are the cosines of the angles between the incident and
diffracted beams, respectively, with the inward-going surface normals. For the
Bragg (reflection) case, we have
(2.5)
for glancing or grazing incidence, and
for glancing or grazing exit, where is the angle between the Bragg plane and the
crystal surface.
The advantage of an increased input divergence is apparent when we plot the
CuK λ 1 line at high magnification (Figure 2.19). Indicated on the plot are the
angular widths of the symmetric and asymmetric Si 220 reflections as used for
the beam conditioner shown in Figure 2.20. The symmetric reflection clearly
does not take the full bandwidth of the source and thus compromises intensity
further than is in fact necessary.
The Bede design is shown in Figure 2.20. The geometry and channel widths
are chosen so that a simple translation of each monolith converts the device from
a very high resolution beam conditioner, with the equivalent of seven parallel
reflections, to a high intensity system. Moreover, the angle may be selected to
give the exact degree of compromise required, rather than having it dictated by
the material. It would be simple to add further slots at different values of The
value shown in Figure 2.20 is appropriate for compound semiconductors, whose
rocking curves are often around 12■ wide, but a powder diffraction experiment
could take the full width of the line at about 200■ .
Finally we remark that proper alignment of all diffractive-optic beam
conditioners normally requires that they be adjusted with a common dispersion
plane. This is defined as the plane containing the incident and diffracted beam
(and therefore also the normals of the diffracting plane) for any given reflection.
If the dispersion planes are not parallel then some intensity will be lost from the
restriction this causes to the angular acceptance in the plane perpendicular to the
dispersion plane.
2.5.5
X-ray mirrors
With developments in ultraprecision engineering it is now possible to produce
surfaces that are smooth enough to reflect X-rays efficiently at small grazing
inci dence angles (typically <1°), and accurate enough to focus them with small
aberrations. 13 These have particular applications at synchrotron radiation
sources, in which the divergence is intrinsically extremely low, but beamlines are
some tens of metres long and great increases in flux are achieved by collecting
radiation over a larger angle. This may be achieved by curved mirrors. Doubly
curved mirrors collect flux over a solid angle, and are particularly useful.
However, the engineering problems are acute, requiring <1 μ m figure accuracy
and <2 nm surface roughness over a surface 100×500 mm for a typical advanced
beamline.
The complexity of the geometry usually necessitates numerical analysis by ray
tracing in order to design such beam conditioners and to predict their aberrations.
The design principles are straightforward:
■ An elliptical mirror is used to focus radiation from a source at one focus to the
specimen at the other focus
■ An ellipsoidal mirror has the same focusing properties but collects over two
axes rather than one
■ A parabolic mirror is used to collect radiation from one focus and deliver a
parallel beam
■ A paraboloidal mirror will similarly collect over two axes
Figure 2.18 (a) The Si 220 rocking curve with different values of the offcut angle from 0
to 17.14° in comparison with the Ge symmetric 220. The widths of the Ge 220 symmetric
and the Si 022 1 7.14 offcut are each 12■ . (b) Rocking curves for ‘high resolution’ settings
of Si 220 (4-reflection) and Ge 440 (2-reflection). All calculations are for random
polarisation; this gives rise to the inflections on the left-hand sides of the curves, and to
higher polarisation losses in the Ge case
■ The above figures are often approximated by the simpler cylindrical or
toroidal mirrors, but these are much less efficient and give significantly
increased aberrations
Figure 2.19 A detailed plot of the spectrum from a Cu X-ray tube, in the vicinity of the K
λ lines, showing the area selected by the high resolution and high intensity settings of the
beam conditioner shown in Figure 2.20
■ Mirrors are combined with crystals to deliver monochromatic beams, but care
must be taken to match divergences at the crystal with its intrinsic acceptance
range (rocking curve)
High intensities can be achieved with conventional sources as well as with

synchrotrons, especially if the mirror can be placed very close to a microfocus
source as recently demonstrated by Bede Scientific. The same precisions and
accuracies are required as for the synchrotron radiation mirrors, but the optical
components are much smaller. Further improvement in efficiency with the same
geometries is achievable if Bragg diffraction can be used instead of specular
reflectivity. This requires that the interplanar spacing changes with position
along the mirror. This is impracticable with crystals, but possible with synthetic
multilayers and such beam concentrators are now becoming available in both
focusing (elliptical and ellipsoidal) and collimating (parabolic and parabo-loidal)
geometries. A review is given by Arndt 14 and an application of graded
multilayer mirrors by Göbel. 15
Figure 2.20 The Loxley-Tanner-Bowen combined high resolution and high intensity
beam conditioner based on the duMond principle, illustrated for Si 202 with CuK λ 1
radiation. (a) Geometric arrangement for high resolution, (b) geometric arrangement
for high intensity
2.6
Aberrations in high resolution diffractometry
As with most instruments, erroneous results may be obtained if the high
resolution diffractometer is not correctly set up. In modern instruments this is
greatly assisted by the manufacturers, by prealigned components, permanent
alignment on installation and by automated algorithms for specimen alignment.
The primary errors to be considered are
■ Tilt of the Bragg plane of the specimen, so that its dispersion plane is not parallel
to that of the beam conditioner
■ Curvature and mosaic structure of the specimen itself
■ Wavelength dispersion arising from the bandpass of the beam conditioner
2.6.1
Bragg plane tilt
The theory of this aberration was worked out in the 1920s by Schwarzchild. 16
For simplicity we shall discuss the case of a beam conditioner comprising a
single crystal and an aperture as in the classic double-crystal arrangement. If the
Bragg planes are tilted about an axis contained in the incidence plane and the
Bragg planes, then rays which are not contained in the incidence plane will not
see equal angles with respect to the specimen and the reference. If we set the
crystals so that the median ray (in the incidence plane) makes equal angles, then
an inclined ray may make the Bragg angle for the reference crystal but will not
be diffracted from the specimen (Figure 2.21). The result is that only a band of
rays satisfies the Bragg conditions for both crystals. The band moves up (or
down) as the crystals are rotated. The consequences are:
■ the rocking curve is broadened

■ the peak height is reduced
■ the area under the peaks remains approximately constant
■ the separation of epilayer and substrate peaks remains approximately constant
This is shown experimentally in Figure 2.22. The specimen comprises a GaAs

substrate with a GaAlAs layer, and it was measured using a diffractometer with a
single GaAs beam conditioner. The three-dimensional plot shows the diffracted
intensity as a function of specimen rotations both parallel and perpendicular to
the dispersion plane. The latter is the axis λ that we scan when measuring the
rocking curve, the former, λ , is the source of aberration. The larger peak is that
from the substrate, the smaller is that from the layer. In the next chapter we shall
see that the composition of the GaAlAs layer may be determined from the peak
separation, the substrate and layer perfections from the peak widths and the layer
thickness from the integrated intensity of the layer reflection. These must
therefore be measured accurately, and it is seen that aberrations are particularly
acute on the peak height, but that the peak splitting is not badly affected.
The broadening λ w of the rocking curve in the (+ n, − n ) non-dispersive
setting, is, to a good approximation
λ w=λ λ (2.7)
where λ is the beam divergence normal to the incidence plane and λ is the tilt
between the Bragg planes of the specimen and the dispersion plane of the beam
conditioner.
The broadening is purely geometric. Therefore it is more important for
intrinsically narrow rocking curves than for broad ones. For strained layer
systems, tilt optimisation is often wasted time. Also if you have a dispersive (+ n,
−m ) set-up, such as InP sample with a GaAs beam conditioner, the tilt
broadening is usually small compared with the dispersion broadening. Again it is
not then worth doing a tilt optimisation. For synchrotron radiation work, where
the divergence is very low, tilt optimisation is unnecessary except in extremely
high sensitivity work with silicon. However, tilt optimisation should be the norm
with laboratory sources on good quality samples. For a typical collimator, λ is of
the order of 10− 2 and thus misorientation of a specimen by 2° will result in peak
broadening of a large fraction of an arc minute. There are two methods for
adjusting the specimen tilt systematically, considered below.
Figure 2.21 Bragg plane tilt aberration. (a) Diffracting planes parallel, diffraction occurs
simultaneously over the whole height of the beam. (b) Diffracting planes skewed,
diffraction only takes place over a narrow band. (c) As the crystal is rotated to measure
the rocking curve, the band moves up or down the crystal. The integrated intensity
remains approximately the same as in case (a) but the peak intensity decreases and the
width increases
2.6.1.1
Tilting about axis in incidence plane and Bragg plane
From Figure 2.22 it is seen that we must alter the tilt until the rocking curve peak
height is a maximum. This is possible but tedious to do manually, but it may be
achieved under computer control with a suitable algorithm. The algorithm has to
take account of the fact that there are multiple peaks in the field if layers are
present, that the scales in Figure 2.22 are very different in the λ and λ directions,
that slight inaccuracies in the construction of the instrument will skew the axis of
symmetry of the (approximate) parabola in the λ − λ plane, that ultraprecision
axes may take an appreciable time to move and that even the best diffractometer
in the world has some backlash in its axis control! An idealised peak-finding
algorithm will not succeed.
Figure 2.22 A three-dimensional plot of the diffracted intensity as a function of specimen

rotations both parallel ( λ ) and perpendicular (λ ) to the dispersion plane. The larger peak
is that from the substrate GaAs and the smaller from the GaAIAs layer. CuK λ 1, slit-
limited from a single GaAs beam conditioner
Two successful automated procedures have been given. In each case one must
first find the peak, and a binary search with doubled limits on each reversal 17 is
the fastest search method if the peak position is unknown. One is then on either
the upper or the lower branch of either the substrate or the epilayer ‘parabola’ (it
is in fact a very shallow and distorted parabola). Fewster 18 first showed that the
branch could be detected by driving down in λ until the intensity is about halved
from that of the peak, then driving down in λ . If the intensity rises one is on the
upper branch of the parabola, if it falls one is on the lower branch. Fewster then
alternated steps in λ and steps in λ until the peak was found. The latter is
determined by the intensity falling instead of rising on beginning the λ step.
This algorithm works in most circumstances but it is slow, and the end
condition is not unique. The I–λ – λ field narrows sharply near the optimum
position, so that at certain intensity levels the intensity contour is double-valued
in λ ; that is, it is shaped like the Greek letter ■ rather than a simple parabola.
With some starting conditions the peak will be found on one of the two left-hand
points of the ■ rather than the correct right-hand point. An algorithm been
developed by Loxley et al . 17 which avoids this problem. Fewster’s condition is
used to determine the location of the branch of the parabola. The axis λ is moved
in the known direction until the intensity has dropped by about 10%, then a step
in λ is made until the intensity has peaked and then dropped by 10% from the
peak. Iteration of this procedure sends the specimen adjustment towards the
optimum, while ensuring that it stays on the ridge of the surface and does not
encounter the ambiguous conditions.
2.6.1.2
Rotation about the specimen surface normal
A different method, based upon the rotation λ of the specimen about the surface
normal, has been described by Tanner, Chu and Bowen; 19 this depends upon a
geometrical location of the end point and should in general be more rapid for a
comparable accuracy. It also lends itself to automation. However, it will only
work for symmetrical reflections. It uses the concept that the crystal’s reflecting
planes are parallel when the reflecting plane normal is rotated to lie in the
incidence plane, and the Bragg angle has been found. The principle is illustrated
in Figure 2.23, which shows a 360° λ -scan of a sample with a single epilayer,
indicating the position of ■ that gives perfect alignment of the diffracting planes.
The plot is symmetrical about the ideal position, irrespective of the complexity
of the rocking curve, thus giving a geometrical means of finding the ideal
position without having to find any maximum positions or even having to decide
which is the substrate peak.
The peak is first found by a positive sense search on the λ axis. Since wafers
are usually some fractions of a degree misoriented from the nominal
crystallographic plane—and sometimes deliberately offset by up to 4 degrees—
this provides a rapid scan over a relatively large angle. If this fails, a binary
search is made on the λ axis. When the Bragg position is located on the positive
sense ■ -scan, this axis is searched in the opposite direction to find the
‘conjugate’ Bragg peak. The correct setting of ■ is midway between these two
peaks.
All automated algorithms require extensive tests and traps to detect unusual
and fault conditions, such as failure of the X-ray source, grossly misaligned
samples, and even perfectly aligned samples!
2.6.2
Curvature and mosaic structure
If the specimen crystal is curved, there will be a range of positions where the
diffraction conditions are satisfied even for a plane wave. The rocking curve is
broadened. It is simple to reduce the effect of curvature by reducing the
collimator aperture. For semiconductor crystals it is good practice never to run
rocking curves with a collimator size above 1 mm, and 0.5 mm is preferable.
Curved specimens are common; if a mismatched epilayer forms coherently on a
substrate, then the substrate will bow to reduce the elastic strain. The effect is
geometric and independent of the diffraction geometry. Table 2.1 illustrates this
effect.
Figure 2.23 A 360° λ -scan of a sample with a single epilayer, indicating the value of λ
that gives perfect alignment of the diffracting planes
A mosaic spread in the crystal will give a similar broadening. If the beam can
be reduced to below the average mosaic cell size then the rocking curve will
narrow. For a broader beam, substrate quality may be assessed by measurement
of the full width at half height maximum (FWHM) of the high resolution Bragg
peak, provided that the tilt and curvature aberrations have been removed as
described above. The increased width of the rocking curve then arises principally
from the variation in tilts and dilations in the crystal from its dislocation
structure. It is worth noting, however, that when crystals contain sub-grains of
size comparable to the beam area, the rocking curve FWHM can be a strong
function of beam size. Measurements made with a different size or divergence
beam may well yield differ
Table 2.1 Broadening of rocking curve from 004 reflection from InP. CuK λ radiation
with a beam diameter of 0.5 mm
Radius (m) Substrate peak width (arc seconds)
Infinity 9.8
100 10.1
10 14.1
5 22.6
1 104
ent values for the FWHM if there is a sub-grain structure and this can be the
cause of protracted and frustrating arguments between substrate vendors and
customers! The triple-axis reciprocal space mapping methods are preferable in
such cases, since they separate the effects of tilt (mosaic spread) and strain.
2.6.3
Dispersion
If the beam conditioner delivers a significant spread of wavelengths, then the
rocking curve may be broadened. Whether it is or not depends upon the
diffraction geometries. The (+ n, − n ) geometry is non-dispersive and all
wavelengths diffract at the same relative setting of specimen and reference. In
the (+ n ,− m ) cases, in which the lattice plane spacings of reference and
specimen are not equal, the rocking curve is broadened by dispersion. The
broadening λ λ in these cases is given by
(2.8)
where λ 1 and λ 2 are the Bragg angles of reference and specimen crystals and λ λ /
λ is the bandwidth of the X-ray wavelengths passed by the beam conditioner. For
the (+ n, − m ) setting this can be taken as the separation between the
characteristic K λ lines as these lines contribute most to the intensity. The
separation is the difference in the K■ 1 and K■ 2 angular separations in the
reference and specimen reflections. For example, if InP is measured using a
single GaAs crystal as beam conditioner, there is a broadening of the order of 25
arc seconds. This is adequate for many purposes, including the study of poor-
quality materials and also strained layer systems, where the layer peaks are
typically 1000 seconds or more from the substrate. It would be inadequate for
quality analysis of good substrates.
The dispersive (+ n, − m ) mode has already been seen clearly with the
duMond diagrams, Figure 2.10. Here, the curves are no longer identical and the
crystals must be displaced from the parallel position in order to get simultaneous
diffraction. As the crystals are displaced, so the band of intersection moves up
and down the curve. When the curves become very different, the K λ 1 and K λ 2
intensities are traced out separately. Then the peaks are resolved in the rocking
curve, and if no better beam conditioner is available it is important in such cases
to remove the K λ 2 component with a slit placed after the beam conditioner. A
slit placed in front of the detector, with the detector driven at twice the angular
speed of the specimen, also works very well. This is in effect a low resolution
triple-axis measurement.
In the general multiple-crystal beam conditioner case, there is no universal
formula for broadening. Rather, the duMond diagram is constructed for the beam
conditioner and the shape of the passed band in λ λ and λ λ is determined. The
specimen crystal is then represented on the duMond diagram and scanned
through the beam conditioner bandpass. For accurate simulation, of course, the
true shape of the rocking curves must be taken into account, since they are not
simple rectangular functions. The duMond diagram representation is most useful
for obtaining a semi-quantitative understanding of the properties of a given beam
conditioner. For example, if the bandpass region is very much smaller in both
directions than the rocking curve width of the specimen, the latter will be
accurately measured though the intensity may be small. The assumption that the
broadening in a multiple-reflection monochromator is equal to the divergence of
the exit beam is usually adequate.
We can now answer the common question, is it better to use the double crystal
or the four-reflection monochromator for high resolution work? We see from the
above, that if the first and second crystals are well matched in Bragg angle, the
resolution loss in double-crystal will be simply that from the correlation of
equation (2.3). This is about 1.4 times the plane wave rocking curve width. If
they are not well matched, then the further loss of equation (2.8) arises. This is
plotted in comparison with the high resolution and high intensity settings of the
Loxley-Tanner-Bowen monochromator (Figure 2.20) in Figure 2.24. The
diagram is drawn for a Si 004 first crystal; for another material, the ‘V’ would be
a very similar shape, but centred on the Bragg angle of the material. The
minimum value of about 2arc seconds in the double-crystal arrangement is the
difference between the width of the resulting rocking curve and that which would
be obtained for a perfect plane wave. This value would increase to about 5
seconds for the heavier III–V materials such as GaAs. However, if the crystals
are not well matched the four-reflection monochromators are superior and hence
more valuable for general work.
2.7
Detectors
The detector must obviously have good sensitivity to X-rays, a good dynamic
range and a low background noise. The choice for high resolution measurements
has traditionally been the ionisation chamber, scintillation counter or proportional
counter with solid-state detectors used occasionally for low-noise applications. It
is necessary for good noise performance to count single photons rather than to
measure a current or voltage from the detector. This eliminates ionisation
chambers except for use on synchrotron radiation sources, where their very high
saturation level can be exploited and low-noise performance is less essential. We
summarise the behaviour of counting detectors here; a detailed treatment may be
found in Knoll. 20
Figure 2.24 The angular resolution of a high resolution system, comparing the
‘performance functions’ of the double-crystal method with a four-reflection
monochromator (Loxley-Tanner-Bowen, high resolution and high intensity settings). The
performance functions show the resolution (ordinate) and intensity (width) of the
conditioner, read off at the intersection with the vertical line at the Bragg angle of the
specimen crystal. The double-crystal setting (using Si 004 as an example) is superior in
both resolution and intensity when the reference and sample crystals are well matched,
but the monochromator systems are superior if a wide variety of samples are to be
measured
2.7.1
Proportional counters
A proportional counter consists of a tube filled with a gas such as xenon, with
positive and negative electrodes. The negative electrode is a thin wire maintained
at a potential around −2 kV. Incoming photons ionise gas molecules. These drift
towards the negative electrode, until the field enhancement around the thin wire
is sufficient to multiply them by the cascade effect, and cause a charge pulse on
the wire. The pulse is quenched by the addition of a quench gas, normally a
halogen or hydrocarbon which reacts with the ions and stops the cascade.
The proportional counter can give very good low-noise performance, well
below 1 cps. It also has the property that the charge pulse is proportional to the
incoming photon energy, with a resolution around 10% λ E / E, so a pulse height
analyser may be used to clean up the signal against spurious photons, for
example cosmic rays. Its main disadvantages are that it is difficult to quench the
pulse rapidly enough to get a count rate above about 10 6 cps, and that the quench
gas is used up, so the detector has a limited lifetime, typically two years or so.
The efficiency is rarely above 50%.
2.7.2
Scintillation counters
The scintillation counter works by first converting the X-ray photon to optical
photons via a scintillator. The optical pulse is then amplified by a
photomultiplier. Although these are only about 5% efficient, several hundred
optical photons are produced by one X-ray photon so the quantum efficiency
approaches 100%. Originally, scintillation counters had poorer noise and a lower
dynamic range than proportional counters, but recent developments have
reversed this position. Using fast scintillators with a 10ns pulse width,
photomultipliers with low-radioactivity glass and high stability electronics, a
dynamic range of 0.1 to 3×106 cps has been achieved at high quantum efficiency
21 (Figure 2.25). The energy resolution is not so good as the proportional
counter, at around 50%, but is still good enough to discriminate against the
cosmic ray background.
2.7.3
Solid-state detectors
In solid-state detectors the incoming photon illuminates a piece of silicon or
germanium, creating electron-hole pairs. These are separated by a biased p-n
junction and amplified with a charge amplifier. The pulses have to be measured
with a multi-channel analyser for spectroscopy, but for diffraction a single-
channel analyser (pulse-height analyser) suffices. The noise performance is very
good, below 0.01 cps, but the upper limit is rather low for diffraction at a few
times 104 cps. Solid-state detectors have the highest energy resolution, around
2%. They are therefore mainly used for experiments in which the count rate is very
low and good signal-to-noise is required. However, the preamplifier has to work
at low temperature, usually under liquid nitrogen, for such low noise to be
achieved.
Recent advances in semiconductor materials have made feasible room-
temperature solid-state detectors made from crystals such as mercuric iodide. At
present they are not competitive in noise and dynamic range with advanced
scintillation or proportional detectors, but may become so in future.
2.7.4
Position-sensitive detectors
For normal high-resolution studies, the detector has a fairly large aperture,
typically 5–25 mm. There are some applications for position-sensitive detectors
for low resolution triple-axis diffractometry. These are in essence gas-filled
Figure 2.25 The behaviour of an advanced scintillation detector at high count rates. After
Cockerton and Tanner 21
proportional counters, with detection of the position of the charge pulse along the
wire. This is achieved by comparing the times taken for the pulse to arrive at
either end of the wire. Again, these are limited to about 104 cps maximum count
rate, but the position resolution can be a few tens of micrometres.
2.7.5
Dead time correction
At the upper end of the range, counting detectors saturate owing to pile-up of the
pulses from the ion cascade, scintillator or silicon crystal. These pulses have
well- defined characteristics, and a statistical treatment allows the count rate to
be corrected for the ‘dead time’ during which a detector cannot handle a new
pulse because it is still processing the last one. There are two limits of behaviour:
the paralysable detector, in which the dead time switches the detector completely
off and the accumulation of successive pulses eventually reduces the quantum
efficiency to zero; and the non-paralysable detector which simply gets less and
less efficient as the intensity increases. The equations describing the limiting
cases are:
Paralysable:
R measured =R true exp(−R true λ ) (2.9)
Non-paralysable:
R measured =R true /(1+R true λ ) (2.10)
where R measured and R true are the measured and true count rates respectively
and λ is the dead time during which the detector cannot sense a new pulse. A
practical detector is likely to be somewhere in between. The most useful
approach may therefore be to measure curves such as that illustrated in
Figure 2.25 and fit them to a cubic spline polynomial rather than one of these
simple equations. An improvement factor of about five in the dynamic range can
thus be achieved.
2.8
Setting up a high resolution experiment
The requirements for full alignment and specimen set-up are given in this section.
The way in which these are achieved will differ between instruments and
manufacturers, and some of them may well be built into the instrument in the
factory. Even so, it is reasonable for the user to enquire how, and to what
precision, these are achieved.
The symmetric setting, illustrated in, for example, Figure 2.4, is the
commonest and easiest diffraction experiment, since the reflecting planes are
parallel to the specimen surface—normally 004 reflections in semiconductor
wafers. This gives information about the strains normal to the surface but it is
important to note that this tells us nothing about the deformation of the layer in
the plane of the substrate. No information on layer coherency or relaxation is
available in the symmetric geometry. By using Bragg planes which are not
parallel to the surface, we obtain a measure of the in-plane strains and hence may
measure relaxation. This is the asymmetric setting, for example the 224 or 113
reflections from an (001) wafer. Both symmetric and asymmetric settings must
be available for complete characterisation.
2.8.1
Instrument alignment
1 Determine the take-off angle from the X-ray tube. This controls the beam
size, and a resultant beam size of approximately 0.5×0.5 mm is about right.
Use the maximum take-off angle consistent with this size. Align the
instrument so that the beam conditioner points at the source at this take-off
angle.
2 The beam conditioner may have some angular, tilt and translation
adjustments. These should be set so that the beam delivered
■ passes along the axis of the instrument, to a precision, 0.1°

■ passes over the specimen (λ ) axis, to a precision ~0.1 mm
■ is parallel to the diffractometer body (perpendicular to the λ axis), to a
precision ~0.1°
■ is maximised in intensity
■ contains only the K λ 1 line
3 Clamp the diffractometer if appropriate so that these alignments are

preserved.
2.8.2
Specimen alignment and peak finding
1 Mount the specimen on the holder, so that it is securely held without

vibration or creep but is not strained by the mounting. This is achieved by
fixing the specimen at one point only, using PVC adhesive tape (as supplied
for crystal growth laboratories). Kinematic location (resting under gravity
with minimum constraints) may be used in some diffractometers. At all
costs avoid over-constraint of the specimen.
2 Ensure that the point to be measured on the specimen is that being hit by the
beam. This is best achieved by aligning the specimen so that its surf ace lies
on the λ axis. An adjustment perpendicular to the specimen surface is
required for this alignment, which should be made to 0.1 mm for normal
work, and 0.01 mm or better for grazing incidence work.
3 For symmetric reflections the peak search may now begin. For asymmetric
reflections, the specimen must be rotated about its normal until the desired
diffraction vector lies in the incidence plane of the beam conditioner. This is
normally the diffractometer surface. An accurate knowledge of the
orientation of the specimen in two axes is required to set asymmetric
reflections; this is usually taken from the position of the orientation flat or
groove.
4 Switch on X-rays and search for a peak by scanning the λ axis. The binary
peak search described in section 2.6.1.1 is the most efficient if the position is
unknown. An intense peak is easily seen, therefore the scan can be fast, with
short detector counting times.
5 When a peak is found, do a rapid scan in its region to find the narrowest
peak. This is the substrate peak and is normally also the most intense.
6 Use either of the methods described in sections 2.6.1.1 and 2.6.1.2 to
optimise the tilt setting of the specimen.
2.8.3
Measuring the rocking curve
Now the specimen is perfectly aligned and it remains to measure the rocking
curve. This is normally done by selecting a step size on the λ axis and a counting
time per step for the detector. Some systems will allow instead the selection of a
total number of counts at each step, so that constant statistical accuracy is
achieved. In either case the diffracted intensity is found as a function of the angle
on the λ axis.
The step size and counting time must be chosen so as to record all the
significant features of the specimen for the measurement. If only the peak
splitting is needed, for example for a composition map on a ternary wafer, the
measurement may be made very rapidly and data processing used to obtain good
locations of the peaks. If fine features are to be extracted, such as weak
interference fringes, the data must be collected more slowly. The most important
principle to grasp is that the emission of X-rays follows a Poisson distribution.
Therefore the standard deviation λ of a number of counts is equal to the square
root of the number of counts itself, to a very good approximation. Note that it is
the total number of counts, not the count rate, that is important. A good rule is
that the acceptable error in a reading is ±3λ . Thus the precision of a count rate of
100 counts per second, measured with a counting time of 5 seconds, is ±3 ■ ±67
counts or ±13 counts per second.
There is no point in measuring for longer than is required, nor is there ever any
point in collecting bad data. As will be discussed later, an important measurement
strategy in materials characterisation is to simulate the expected rocking curve
using a dynamical diffraction computer package (see Chapter 5), and, noting the
intensity of the main peak, decide what counting time is required to distinguish
the features of interest. The properties of the diffractometer, in particular the
intensity and bandpass of the beam conditioner, the resolution of the axis and the
saturation and noise levels of the detector, will determine the scale of the finest
features that can be resolved.
2.9
Summary
In this chapter we have discussed the choice of X-ray sources and shown the
appropriate regimes for sealed tubes, rotating anodes and synchrotron sources. We
have treated the valuable duMond diagram for qualitative design of X-ray optics,
and applied it to various high resolution beam conditioners. We have covered the
theory of all aspects of the high resolution diffractometer, its aberrations and its
adjustment in some detail, and given systematic methods for minimisation of the
aberrations. We have not attempted to describe the current technology, for
example of motor drives and angle encoders, since these will vary with
manufacturer and with time. It is not necessary to have an understanding at the
level of detail of this chapter for routine use of the instruments, but it is available
as a resource, and should be found particularly useful when new experiments or
new instruments are contemplated.
References
. H.MARK, in Fifty years of X-ray diffraction, ed. P.P.EWALD (Int. Union of

Crystallography, 1962) p. 603 .
. J.W.DUMOND, Phys. Rev., 52 , 872 (1937).
. T.MATSUSHITA & U.KAMINAGA, J.Appl.Cryst., 13 , 465 and 472 (1980).
. J.H.BEAUMONT & M.HART, J.Phys.E: Sci. Inst., 7 , 823 (1978).
. K.KOHRA, M.ANDO, T.MATSUSHITA & H.HASHIZUME, Nucl. Inst. Meth., 152 ,
161 (1978).
. S.W.WILKINS & A.W.STEVENSON, Nucl. Inst. Meth., A269 , 321 (1988).
. R.W.JAMES, The optical principles of the diffraction of X-rays (Ox Bow Press,
Connecticut, 1982) p. 308 .
. J.P.QUINTANA, M.HART, D.BILDERBACK, C.HENDERSON, D.RICHTER, T.
SETTERSTON, J.WHITE, D.HAUSERMANN, M.KRUMREY &
H.SCHULTESCHREPPING J. Synchrotron Radiation, 2 , 1 (1995); J.
P.QUINTANA & M.HART, J.Synchrotron Radiation, 2 , 119 (1995);
L.E.BERMAN & M.HART, Nucl. Inst. Meth., A334 , 617 (1993).
. D.K.BOWEN, D.CHETWYND, N.KRYLOVA & S.SMITH, to be published.
. N.LOXLEY, D.K.BOWEN & B.K.TANNER, Mat. Res. Soc. Symp. Proc., 208 , 107
(1991).
. W.BARTELS, J. Vacuum Science and Technology, B1 , 338 (1983).
. N.LOXLEY, B.K.TANNER & D.K.BOWEN, Advances in X-ray Analysis, 38 , 361
(1995).
. P.KIRKPATRICK, J. Opt. Soc. Am., 38 , 766 (1948); T.W.BARBEE, Opt. Eng., 25 ,
898 (1986).
. U.ARNDT, J. Appl. Cryst, 23 , 161 (1990).
. M.SCHUSTER & H.GöBEL, J. Phys. D: Appl. Phys., 28 , A270 (1995).
. M.SCHWARZCHILD, Phys. Rev., 32 , 162 (1928).
. N.LOXLEY , S.COCKERTON , B.K.TANNER , M.L.COOKE , T.GRAY &
D.K.BOWEN , Mater. Res. Soc. Symp. Proc., 324 , 451 (1993).
. P.FEWSTER, J. Appl. Cryst., 18 , 334 (1985).
. B.K.TANNER, XI CHU & D.K.BOWEN, Mat. Res. Soc. Res. Symp. Proc., 69 , 191
(1986).
. G.F.KNOLL, Radiation detection and measurement (Wiley, New York, 1979).
. S.COCKERTON & B.K.TANNER, Adv. X-ray Anal., 38 , 371 (1995);
S.COCKERTON, B.K. TANNER & G.DERBYSHIRE, Nuclear Instruments and
Methods in Physics Research, B97 , 561 (1995).
3
Analysis of Epitaxial Layers
In this chapter we discuss the measurement and analysis of simple

epitaxial structures. After showing how to select the experimental
conditions we show how to derive the basic layer parameters: the
composition of ternaries, mismatch of quaternaries, misorientation,
layer thickness, tilt, relaxation, indications of strain, curvature and
stress, and area homogeneity. We then discuss the limitations of the
simple interpretation.
3.1
Introduction
The greatest use of high resolution diffractometry in industry is without doubt
the characterisation of epitaxial structures on compound semiconductors. Such
epilayers are mismatched, misoriented, defective, non-uniform and bent. These
defects affect device performance and production yield. Residual strains in the
layer can be correlated with poor device performance or with degradation, and
dislocations in the interface are particularly damaging. III–V layers are nowhere
near as perfect as silicon, either in defect concentration or in composition
uniformity. The defects affect carrier lifetimes and may act as non-radiative
recombination centres, and the composition affects the bandgap as well as the
mismatch. It is therefore essential in the development phase of a device to
determine whether the defects generated in a particular production process will
permit adequate yield of good quality devices. Later on, in manufacture, it is
necessary to determine whether the process is under adequate quality control.
In this chapter we shall see how the basic parameters can be very easily
obtained and interpreted from X-ray rocking curves for relatively simple
structures. We define these as good quality substrates (giving rocking curves
near the theoretical width) and up to, say, three layers, each at least 0.5 μ m thick.
Though these are simple, they nevertheless comprise a large sector of the
optoelectronics device production and are certainly realistic cases. The
discussion will also illustrate many important principles. In later chapters we
shall develop the methods of interpretation of more complex cases.
3.2
Plane wave diffraction from heteroepilayers
Figure 3.1 shows schematically the defects that are common in epilayer
structures. Mismatch is the most vital parameter to determine, both in itself as a
measure of
Table 3.1 The effect of substrate and epilayer parameters upon the rocking curve
Material parameter Effect on rocking curve Distinguishing features
Mismatch Splitting of layer and Invariant with sample
substrate peak rotation
Misorientation Splitting of layer and Changes sign with sample
substrate peak rotation
Dislocation content Broadens peak Broadening invariant with
beam size
No shift of peak with beam
position on sample
Mosaic spread Broadens peak Broadening may increase
with beam size, up to
mosaic cell size
No shift of peak with beam
position on sample
Curvature Broadens peak Broadening increases
linearly with beam size
Peak shifts systematically
with beam position on
sample
Relaxation Changes splitting Different effect on
symmetrical and
asymmetrical reflections
Thickness Affects intensity of peak Integrated intensity
increases with layer
thickness, up to a limit
Introduces interference Fringe period controlled by
fringes thickness
Inhomogeneity Effects vary with position Individual characteristics
on sample may be mapped
strain in the epilayer but also because it allows us to determine composition for
ternary compounds and hence infer the bandgap. Layer thickness is also an
important characterisation parameter for device behaviour.
If we imagine the diffraction of a plane wave from epilayers we see that there
will in general be differences of diffraction angle between the layer and the
substrate, whether these are caused by tilt ( λ λ ) or mismatch ( λ d ). Double or
multiple peaks will therefore arise in the rocking curve. Peaks may be broadened
ANALYSIS OF EPITAXIAL LAYERS 53
Figure 3.1 A schematic representation of the defects common in epilayer structures. (a)
mismatch, (b) misorientation, (c) dislocation content, (d) mosaic spread, (e) curvature, (f)
relaxation, (g) inhomogeneity
by defects if these give additional rotations to the crystal lattice, and there will
also be small peaks arising from interference between waves scattered from the
interfaces, which will be controlled by the layer thicknesses. The material will
show different defects in different regions. Table 3.1 summarises the influence
on the rocking curve of the important parameters. The distinguishing features of
each effect will be discussed quantitatively in this chapter. However, we need to
ensure first of all that the experimental measurement is properly designed.
3.3
Selection of experimental conditions
Most high resolution diffractometry is done in reflection. III–V and II–VI
materials are too absorbing to use conveniently in transmission and, in any case,
reflection techniques provide information only from the relevant surface region.
In reflection, the intensity of the X-ray wavefield inside the crystal falls off
very rapidly away from the surface, due to transfer of energy to the diffracted
beam. Absorption also becomes important at low incident angles to the surface.
By choosing the radiation and the reflection (including its symmetry), the
penetration may be varied between about 0.05 and 10 micrometres. This is
ideally matched to device structures. This is quantified by the extinction distance
λ g, defined as the depth at which the incident intensity has decreased to 1/e of its
value at the surface. This may be calculated from diffraction theory, and some
examples, for GaAs with CuK λ radiation, are shown in Table 3.2. It is assumed
that the wafer surface is (001), hence the 004 reflection is symmetric and the
others asymmetric.
The effect of absorption will be, roughly, to halve the extinction distances at
the lower glancing angles. To get to really low extinction distances with
characteristic radiation it is possible to use skew reflections to get very low
glancing or grazing angles. The minimum penetration in diffraction is obtained
at the critical angle for total external reflection. There is then a very rapid
transition to external reflection; but, of course, there must be a little ‘sampling’
of the crystal by the beam and the penetration is a few atomic layers. This gives
two-dimensional diffraction, which is now being used to give information about
surface layers, but is rarely attempted with conventional X-ray sources. Even
with more usual arrangements, glancing angle diffraction can be an invaluable
tool for the study of thin films. Figure 3.2 shows the dramatic effect obtained for
a thin epilayer of InGaAs on InP. We see that if the symmetrical reflection is not
usable then the radiation and reflection should be chosen primarily to control
penetration depth. Strong reflections give shorter extinction distances.
Unfortunately, the strongest reflections of all in the zincblende structure, 111 and
220, are not accessible in reflection with (100) wafers with CuK λ radiation,
though they are ideal in transmission.
Most of the important parameters may be derived from symmetric reflections,
as shown below. The exception is the measurement of relaxation, that is, the
extent to which the interface is less than perfectly coherent with the substrate. A
large mismatch will cause a substantial strain in the layer which, at some critical
thickness, will be relieved by the generation of interface dislocations. The
average density of these may be accurately measured, by measuring the lattice
mismatch parallel as well as perpendicular to the interface; this requires an
asymmetric reflection with a substantial component of the reflection vector
parallel to the interface, in addition to the symmetric reflection and the
determination of layer tilt relative to the substrate.
The accuracy of all measurements will be improved if the background (away
from the peaks) is low. This should never be more than a few cps if beams of
1×1 mm or less are used, and <0.1–0.2 cps is attainable with standard
instruments and a good detector. The principles necessary to achieve this are as
follows:
Table 3.2 Extinction distances in GaAs with CuK λ radiation

Reflection Bragg angle (degrees) Rocking curve width (arc Extinction distance (µm)
seconds)
004 33.024 5.34 7.1
044 50.420 6.80 9.5
115 45.070 2.60 21.2
224 41.873 9.50 6.5
1 Ensure that there is no unshielded path between the end of the collimator
and the entrance of the detector.
2 Ensure that there is no unshielded path between the point at which the
primary beam strikes the first crystal and the entrance of the detector.
A scatter shield must therefore surround the first crystal/collimator assembly, and
the beam-defining slits should be placed in between the crystals. The necessity
for defining the optic axis of the instrument, and placing the first and second
crystal surfaces on this optic axis is now apparent. If high energy radiation is
being used, it may be necessary to shield the sides of the detector as well. If
problems remain, they can usually be overcome by experimenting with the
placement of scatter shields using the above principles. Proper design of the point
at which the collimator enters the tube shield is necessary as this can be a source
of considerable noise. Lead-loaded Perspex (Plexiglass) is a very useful shielding
material, since it is transparent and thus easy to align.
Figure 3.2 The effect of glancing incidence geometry, calculated for a 150 nm ln0.5177Ga0.
4823As film on a (001) InP substrate. The symmetric 004 and the grazing incidence 044
curves are shown
3.4
Derivation of layer parameters
We now assume that we have properly recorded rocking curves available, and
that a substrate with a single epitaxial layer >0.5 μ m thick (and less than, say, 5
μ m) is measured. This will result in two peaks, one each from the substrate and
layer. The same analysis will apply to multiple peaks (other than those from
superlattices) provided that they are well separated so that interference effects
are minimised. The basic parameters are derived as follows, with the symmetric
reflection used unless otherwise specified. The examples are given for (001)
substrates and layers, but are quite general (with the caution that if the Poisson ratio
is required, the value appropriate to the crystal orientation should strictly be used).
3.4.1
Mismatch
The peak separation in the non-dispersive setting between the substrate and layer
reflections is measured; call this λ λ . This is related to the change of interplanar
spacing normal to the substrate through the equation
λ d / d=−λ λ cotλ (3.1)
Figure 3.3 The tetragonal distortion in a coherent epilayer, (a) fully relaxed, and (b)
constrained to match the substrate
If the reflection is the usual symmetric 004 then the ‘experimental X-ray
mismatch’
m* = λ a / a = λ d / d (3.2)
assuming, for the moment, that the layer is not tilted with respect to the
substrate. This is derived directly from the measurements with no extra
parameters or assumptions. However, this is not the mismatch that would be
measured if the layer were removed from the substrate and allowed to relax to its
natural, unstressed state. The epilayer is elastically constrained to match the
substrate in the plane parallel to the substrate, in both x and y directions, and
there is a consequent tetragonal distortion of the epilayer, as shown in
Figure 3.3.
The true or relaxed mismatch m is defined with respect to the relaxed (i.e. the
normally tabulated) lattice parameters a l and a s of the epilayer and substrate
materials as
m=(a l −a s )/a s (3.3)
It can then be calculated, by means of some straightforward elasticity theory 1 as
(3.4)
where m is the Poisson ratio; since v~1/3, m~m*/2 as an approximate guide. For
accurate work the Poisson ratios of the constituents must be determined, and in
principle these will vary with the alloy content.
The intrinsic accuracy of mismatch measurements is ample. For example, a 1
μ m layer that is mismatched by 220 ppm from an InP substrate gives a peak
splitting of 28■ ; this can be measured to about 2%, giving a resolution of 4–5
ppm. The peaks begin to merge below about 100 ppm but splitting down to
about 50 ppm can be handled by simple methods; smaller splittings can be
determined through computer simulation and fitting of the whole rocking curve
as will be shown in Chapter 5.
3.4.2
Composition
For a ternary layer this may be determined from Vegard’s law; this simply states
that the lattice parameter of a solid solution alloy will be given by a linear
dependence of lattice parameter on composition, following a line drawn between
the values for the pure constituents. This process may be automated through
software that performs all calculations necessary to extract peak splitting, relaxed
mismatch and composition automatically for ternary layers. Vegard’s law 2 was
originally proposed for ionic salt pairs, e.g. KCl–KBr but has been widely
investigated for metals, in which it does not work too well, and widely assumed
for III–V semiconductors. It is based on elastic interactions between atoms, and
is thus reasonable when all electronic interactions in the alloy series are very
similar.
Quaternary multilayers present a substantial challenge because there are so
many variables. If the layers are not graded, and if there are only two types, e.g.
InP substrate and buffer layers and a single quaternary composition for the active
layer and cap, the mismatch may be obtained from the peak splitting. The
composition cannot be obtained from the mismatch since a degree of freedom
remains. However, a photoluminescent measurement of the bandgap plus the
mismatch measurement does suffice.
3.4.3
Substrate misorientation
Substrates are often specified at some controlled angle from (001), and this may
need to be verified. A specimen stage is required that can rotate the specimen
about its surface normal. If the specimen rotation method is used to bring the
reflecting planes of the specimen parallel to those of the reference crystal then in
the final configuration the reflecting plane normals in the specimen and reference
crystals and the specimen surface normal are all coplanar. Rotation of the specimen
about a further 180° (which is the other coplanar setting) will give a shift in the
position of the Bragg peak. This will be exactly twice the misorientation angle
between the reflecting plane and the specimen surface.
3.4.4
Tilt
If the layer is tilted relative to the substrate then this will result in a shift of the
layer peak relative to that of the substrate for reasons unconnected with
composition. We have two problems—how to measure true splittings, and how
to determine the tilt itself. The peak shift due to tilt will vary with the absolute
direction of the incident beam relative to the substrate (i.e. with respect to
rotation of the specimen about its surface normal), and may thereby be
distinguished from mismatch splitting. If the specimen is rotated λ about its
normal, the displacement of the layer peak from the position it would have were
there no tilt is λ cos λ , where λ is the angle of tilt. It follows that a true splitting
mismatch may be taken by rotating the specimen 180° in its plane and averaging
the two measurements of splitting, i.e.
λ λ =(λ λ 0+λ λ 180)/2 (3.5)
This is insufficient to measure the tilt itself since we do not know whether our
original measurement was in the direction in which the tilt would be a maximum
(i.e. with the reflecting plane normals in the reference crystal, substrate and layer
all coplanar). A third measurement (at least) is required, followed by fitting the
three measurements of layer peak deviation to a sine curve to find the maximum
deviation —this is then the tilt value. The best strategy is to make the third
measurement at 90° to the first. We then have measurements at values of ■
displaced by 0°, 90° and 180°. We do not yet know where the origin of λ is; let
this be at − λ from our zero value of λ . Then after extracting the mismatch part
of splitting as shown above, we have three values of splitting, ■ 0, ■ 90 and ■ 180
which are given in terms of λ (the tilt) and λ (the zero position of the tilt, when
all diffracting vectors are coplanar) as follows:
■ 0=λ cos λ (3.6)
■ 90=λ cos(90+λ ) =−λ sin λ (3.7)
■ 180=λ cos(180+λ )=−λ cos λ (3.8)
then from (6) and (7)
tan λ =■ 90/■ 0 (3.9)
which determines λ . The tilt λ can then be found by substitution in any of the
equations.
3.4.5
Dislocation content
Dislocations are commonly present in two regions. A layer with high mismatch
may relax so that interface dislocations are created to accommodate the strain. A
network at the interface is thus observed. Slip dislocations may be generated by
local plastic deformation due to thermal or mechanical strain and propagate
elsewhere in the layer. Dislocations in the layer itself may also be generated
during the growth process, due, for example, to the presence of inclusions.
Interface dislocations give a specified relaxation of strain between the
substrate and the epilayer, which gives quantifiable shifts in the positions of
peaks in asymmetric reflections as discussed later in this chapter. As structures
become more complex it is difficult to know which effects may be ascribed to
interface relaxation and which to the layer structure itself. It is therefore often
very useful to perform topography to see the dislocations directly, as discussed in

Chapters 8–10.
On the other hand, dislocations inside the epilayer may have any of the
possible Burgers vectors, and will on average contain roughly equal numbers of
each sign. These do not shift the rocking curve, but they both broaden it and add
diffuse scattering. A simple model for the broadening was given by Hirsch, 3
who showed that a reasonable estimate for the dislocation density λ is
(3.10)
in cm− 2, where λ is the broadening of the rocking curve in radians and b is the
Burgers vector in centimetres. Kaganer et al . 4 have recently published a much
better quantitative model for the broadening.
The diffuse scatter arises because dislocations are defects which rotate the
lattice locally in either direction. This gives rise to scatter, from near-core
regions, which is not travelling in quite the same direction as the diffraction from
the bulk of the crystal. This adds kinematically (i.e. in intensity not amplitude)
and gives a broad, shallow peak that must be centred on the Bragg peak of the
dislocated layer or substrate since all the local rotations are centred on the lattice
itself. We can model the diffuse scatter quite well by a Gaussian or a Lorentzian
function of the form:
(3.11)
A further effect that can be seen in thick layers with large mismatch, for example
GaAs on Si, is the decrease of dislocation density from interface to surface; this
is aimed for by the crystal growers, and gives a rocking curve that is hard to
simulate. However, an indication of the decrease in dislocation density can be
obtained using two measurements, one a usual symmmetric 004 and the other
highly asymmetric to confine the beam to a region much nearer the surface. 5 If
the dislocation content indeed decreases towards the surface, then rocking curves
taken with smaller extinction distances (and the same area measured) will show
less broadening.
3.4.6
Curvature and mosaic spread
If the specimen is curved, then the change in angle from one side of the beam to
the other will be cross-correlated with the rocking curve, and the latter will be
broadened. An estimate of this effect is straightforward. Let the radius of curvature
of the specimen be R and the diameter of the beam be s. The angular change of
the incident angle across the beam will then be s / R, as illustrated in Figure 3.4.
For standard measurements, we want this to be a small fraction of the intrinsic
rocking curve width; with III–V materials we might accept a broadening of 2■ or
10− 5 radians, which is 10−15% of the intrinsic width with CuK λ radiation. If
Figure 3.4 The change of incident angle across the beam on a curved specimen
the beam diameter is 1 mm then our criterion is satisfied if the specimen radius
of curvature is no less than 100 m (Table 3.3). This is not always true and
emphasises the need for small beams if accurate double-axis rocking curve
widths are to be obtained on stressed specimens, though the splitting is not
affected.
It is possible to compensate for curvature in theoretical simulation of the
rocking curve. If the beam profile is square, then the rocking curve is simply
correlated with
Table 3.3 The effect of curvature on rocking curve broadening

Beam size (mm) Radius of curvature (m) Broadening (arc seconds)
5 100 10
5 20 50
1 100 2
1 20 10
0.5 100 1
0.5 20 5
a box function whose width is the angular change across the beam, and this is
usually a good enough approximation. However, it is a consequence of the
convolution process that some information is lost—for example, fine interference
fringes will be washed out. The curvature for a given mismatch is less serious on
small specimens than on complete wafers, so if stress is suspected and details are
required, the measurement can be repeated on a cleaved-out portion of the
specimen.
Figure 3.5 The projection of measured points onto the dispersion plane
The curvature itself can be quite easily measured by translating the specimen a
distance x in its plane along a diameter (a circular wafer is assumed), repeating
the measurement and noting the shift λ λ in the absolute position of the Bragg
peak. Then, again,
R=s/λλ
This measures the curvature about an axis perpendicular to the dispersion plane,
i.e. the cylindrical curvature, and it may be necessary to rotate the wafer through
90° to get the orthogonal component. This may be related to absolute stress in
the wafer with knowledge of the wafer thickness, diameter and elastic modulus.
The most accurate method is to measure a number of points on a wafer and use a
linear regression formula for the average curvature.
If the specimen has been rotated (Figure 3.5) then we have
s=x cos λ +y sin λ
(3.14)
The linear regression (least squares) is performed on λ substrate and s to get 1/R.
The residuals give the uniformity of curvature.
3.4.7
Relaxation
Equation (3.4) contains the assumption that the interface is fully coherent. If it is
only partially coherent, i.e. it contains interface dislocations, it is said to be
relaxed, and equation (3.4) is not valid for the determination of the relaxed
mismatch. Note the two different usages of the word ‘relaxed’ in the last
sentence! It is necessary to measure the misfit parallel to the interface as well as
Figure 3.6 A side view of (a) coherent and (b) partially relaxed epilayers. The relaxation
process changes both the interplanar spacings of the epilayer and the angles between the
reflecting planes and the surface. (Courtesy K.M.Matney)
perpendicular. For this, we need an asymmetric reflection which is at as high an
angle to the surface as possible: 224 and 113 are both acceptable.
Figure 3.6 shows a coherent and a relaxed layer, and it is clear that both the
mismatch and the misorientation change between the substrate and the layer. The
tetragonal distortion changes. From equations (3.6) and (3.9) it is clear that the
effect of tilt on the splitting is reversed if the specimen is rotated by 180° about
its surface normal, but the splitting due to the mismatch will not be affected by
such a rotation. Thus we may make grazing incidence or grazing exit
measurements, Figure 3.7, to separate the tilt from the true splitting.
The resulting measured splittings ■ λ i and ■ λ e are now different between
these two geometries:
Grazing incidence:
(3.15)
Grazing exit:
(3.16)
Thus we may determine λ λ and independently. We need to know the lattice
parameter of the layer parallel and perpendicular to the substrate, i.e. the in situ
cell parameters a l , b l and c l of the layer. From these we may calculate the
relaxation and (together with the Poisson ratio) the fully relaxed lattice
parameter of the epilayer. Let the lattice parameter of the (cubic) substrate be a s
, let hkl be the (asymmetric) reflection, λ the wavelength, λ the Bragg angle and
the angle between the reflecting plane and the surface, with subscripts s and l
referring again to the substrate and epilayer. We consider first the case in which
the relaxation is symmetrical, i.e. the distortion is tetragonal and a l =b l . We
know λ , λ s and Using equations (3.10) and (3.11) we may therefore calculate
λ l =λ s +λλ (3.17)
(3.18)
From the equation for the interplanar spacing in tetragonal crystals,
Figure 3.7 (a) Grazing incidence geometry. (b) Grazing exit geometry. (c) Symmetric
scan. (Coustesy K.M.Matney)
(3.19)
and the Bragg law, we obtain
(3.20)
Assuming a 001 crystal surface, we use the formula for interplanar angles in
tetragonal crystals to obtain
(3.21)
From equations (3.16) and (3.17) we may solve for the cell constants of the
layer:
(3.22)
and
(3.23)
The accuracy of c l may be checked from measurement of the symmetric 001

reflection. The relaxation R (%) is defined as
(3.24)
where is the fully relaxed lattice parameter (cubic) of the epilayer. We may find
this if we know the Poisson ratio of the layer, v , since for any layer strained in
two coplanar directions λ xx and λ yy the vertical strain is given by 1
(3.25)
Since and
(3.26)
(3.27)
The fully relaxed lattice parameter of the layer, is the value that we need to use in
Vegard’s law to find the composition of the epilayer.
If we are confident that the (001) planes of the epilayer and substrate remain
parallel, as is reasonable if the substrate is cut parallel to (001)—but not if it is
offcut —we can in principle derive the relaxation from just the symmetric and
the grazing incidence reflections, without using the grazing exit. This may be
important in quality control to save measurement time. The derivation is lengthy
but straightforward. The symmetric reflection provides c l . We find expressions
for λ λ and in terms of the lattice parameters of the substrate and of and c l, using
the tetragonal interplanar spacing and angle formulae. We solve this for a l,
which gives
(3.28)
and we then proceed as before. This equation, which has not previously been
published, contains the approximations that ■ ■ and are small, and if this is not so
an iterative solution is preferable. 5 Beginning with the known c l and the value
for a l given by equation (3.27), we calculate ■ λ and iterate a l until the
agreement with the experimental ■ λ is good.
Once the parallel mismatch is determined, some information about average
dislocation density in the interface may be obtained. It is not possible
unambiguously to determine the types of dislocation present, since different
types of dislocation may combine to give the same strain. However, the parallel
mismatch is entirely due to dislocation content in (or very near) the interface and
thus
(3.29)
where b is the magnitude of the Burgers vector of the interface dislocations

projected in the direction given by the intersection of the incidence plane with
the interface, and s is their spacing in the same direction. If the nature of the
dislocations is known, for example from electron microscopy or X-ray
topography, and two asymmetric reflections are taken, the overall interface
density can be determined unambiguously.
3.4.8
Thickness
The layer thickness determines the relative intensity of the layer and substrate
peaks. Provided again that the structure is simple (later chapters deal with cases
that are not), then the intensity of the layer peak increases monotonically with
thickness. The values and the calibration constant will depend on the particular
system being measured and can be determined either empirically or by computer
calculation. An example is shown in Figure 3.8.
The proper value to use for intensity measurements is the integrated intensity—
the area under the peak—rather than the peak intensity. This is because it is less
variable with material structure. If many dislocations are present, for example in
a strained layer material such as InGaAs on GaAs of moderate thickness, then
the layer peak will be lowered and broadened, but to first order the integrated
intensity will be the same. It is not always possible to measure the integrated
intensity when peaks are close together. The measurement of integrated intensity
is taken out to the tails of the peak where it is indistinguishable from
background. Clearly, better values will be obtained if the background is low,
which emphasises the need for good experimental technique.
The rocking curves of Figure 3.8 show small periodic oscillations around the
layer peak. These are called thickness fringes and are a very good way of
measuring layer thickness. The most accurate way is through computer
simulation (Chapter 5), but it is often useful to get a start to the simulation by
measuring the interference peak separation, ■ λ p , which is given by
(3.30)
where λ is the wavelength, t the thickness and λ g the cosine of the angle between
the diffracted beam and the inward-going surface normal. For the reflection case,
this may be expressed as
(3.31)
where is again the angle between the reflecting plane and the surface, the
positive sign applies to grazing incidence and the negative sign to grazing exit.
For the common symmetrical case, we may simplify and rearrange to obtain
(3.32)
This very useful method also has the advantage that the equations do not contain
anything about the material or diffraction conditions other than the Bragg angle
and geometry. The independence from material parameters arises because the
refractive index for X-rays is very close to unity. The equations are, of course,
similar to those for optical interference from thin films, since the physics is the
same, but in the optical case we do need to know the refractive index.
If more than one layer is present, there will be interference fringes from each
of them. If their thicknesses are different these will superimpose and beat. Very
complex patterns may arise, as will be seen in the HEMT structure simulated in
Chapter 5.
Newcomers to the field often assume that a simple Fourier analysis will
suffice to pull out the periodicities automatically. However, the problem is not at
all simple, as has been shown by Hudson, Tanner and Blunt. 7 There are two
problems. One is that the interference fringes are quite weak, and relatively few
are observed. The
Table 3.4 Comparison of best Fourier analysis with specimen structure 7

Layer thickness from matching to Thickness extracted from Fourier transform
simulation
550 Å GaAs 990±20 Å
400 Å Al0.18Ga0.82As 1100±20 Å
180 Å In0.13Ga0.87As
intensity of the Fourier peak is therefore low, and may be obscured by noise. A
low-noise detector is essential. The second is that the transform is absolutely
dominated by the sharp substrate peak, which of course puts intensity throughout
Fourier space and masks any other modulation. The procedure that showed most
success was
1 Mask out the substrate peak from the rocking curve; in practice, truncate the
curve and study only one side.
2 From the average values of the curve determine a ‘background’ to the
thickness oscillations, and divide the data by the smoothed curve.
3 Subtract the residual background.
4 Perform an autocorrelation on the processed curve.
Figure 3.8 The effect of thickness on layer peak intensity. (a) Rocking curves, (b)
integrated intensity ratio between substrate and epilayer peaks. Calculated curves for Al0.
3Ga0.7As on GaAs. CuK λ radiation 004 reflection
5 Perform the Fourier transform.

The results of the analysis are shown in Table 3.4. It is seen that, although some
of the periodicities are extracted, some which appear are difference or beat
frequencies and do not correspond to a single layer. The conclusion is that
Fourier analysis, with the above procedure, is a powerful aid to a skilled
researcher, but it is not yet appropriate for automated analysis.
More complex integral transforms have been studied, e.g. the two-dimensional
Wigner transform, 8 but so far without success. Even the best-quality data are
too sparse and noisy to give reliable information when transformed in this way.
3.4.9
Area homogeneity
Whatever the crystal growers claim, epitaxial layers are not uniform across their
area! One per cent consistency is very good. It is therefore necessary to repeat
each measurement at a number of points across each wafer. A 9×9 grid across a
3■ wafer is thorough, whilst a 3×3 grid often suffices. Clearly, automated data
collection over such a grid is a great help. Figure 3.9 illustrates a ‘wafer map’
plotted as a mesh surface. This may be applied to any parameter that can be
extracted quantitatively from the rocking curve.
3.4.10
Limitations and problems
In the whole of this chapter we have taken an ‘intensity’ viewpoint, which
requires that frequencies are well separated and interaction is minimal. We have
shown that a very large amount of information may be derived in this way. But
of course, the true physics of waves must consider amplitude and not intensity,
and we should expect that interference effects dominate in certain situations.
These are in cases of
■ thin layers, say below 0.5 μ m

■ small mismatches, so that peaks overlap
■ multiple peaks
■ superlattices
Computer simulation is required to analyse such cases, as will be discussed in

later chapters.
3.5
Summary
In this chapter we have seen what information can be derived directly from the
rocking curve, without computer simulation. This has laid the foundation for
more detailed interpretation. This approach is also of immediate practical
Figure 3.9 Surface mesh plot showing the variation of In content in an InAlAs layer on
GaAs. (Courtesy T.Lafford)
application in the semiconductor industry, since the procedures discussed may

mostly be automated in software and used for very rapid analysis in quality
control applications. It is perfectly possible to measure and to analyse the
composition of a ternary epilayer, automatically, in less than thirty seconds.
References
. L.D.LANDAU & E.M.LIFSCHITZ, Elasticity (Pergamon, Oxford, 1972).

. L.VEGARD, Z.Physik, 5 , 17 (1921).
. P.B.HIRSCH, ‘Mosaic structure’, Chapter 6 in Progress in metal physics, eds.
B.CHALMERS & R.KING (Pergamon, New York, 1956).
. V.M.KAGANER, R.KöHLER, M.SCHMIDBAUER, R.OPITZ & B.JENISCHEN,
Phys. Rev B., 55 , 1973 (1997).
. J.W.LEE, D.K.BOWEN & J.P.SALERNO, Mat. Res. Soc. Symp. Proc., 91 , 193–8
(1987).
. K.M.MATNEY, Mat. Res. Soc. Symp. Proc., 379 , 257 (1995).
. J.M.HUDSON, B.K.TANNER & R.BLUNT, Adv. X-ray Analysis, 37 , 135 (1994).
. R.CLINCIU, MSc thesis, University of Warwick (1992).
4
X-ray Scattering Theory
In this chapter we first extend the kinematical theory, discussing the

strengths of X-ray reflections, forbidden reflections, the thin crystal
solution, anomalous dispersion and reciprocal space geometry. We
then review the results of the dynamical theory: deviation
parameters, dispersion space and geometry, boundary conditions, the
range of Bragg reflections, polarisation, intensity formulae and
penetration depth. We briefly discuss spherical wave theory, and
Penning-Polder theory for distorted crystals. Finally we give a
summary of useful formulae.
4.1
Introduction
The theory of X-ray scattering is highly practical! It is an accurate theory, based
on a few sound assumptions, and with implementations on personal computers it
may be used to interpret the structures of advanced industrial materials and
thereby to assist in process development and quality control. The characterisation
scientist or engineer who has a good grasp of the theory will therefore be able to
design better measurements and to interpret them more accurately.
It is not our intention to provide full derivations of X-ray scattering theory,
since this is primarily of interest to the specialist and may be found in many
excellent books and reviews. The characterisation scientist needs a qualitative
understanding in order to appreciate general features of experiment design and
the interpretation of high resolution rocking curves and images. He or she also
needs specific numbers such as the ideal rocking curve width or the penetration
depth for a particular specimen. Our aim is therefore simply to explain the
aspects of the theory that are relevant to later chapters and to summarise the
important formulae.
It is conventional and useful to approach X-ray scattering theory on two
levels, the so-called kinematical and dynamical theories. The simpler
kinematical theory assumes that a negligible amount of energy is transferred to
the diffracted beam, with the consequence that we can ignore rediffraction
effects. This is fairly accurate for the geometry of diffraction in all cases, and,
reasonably enough, is also fairly accurate for the intensities when the scattering
is very weak. Very thin crystals, surface scattering and diffuse scattering are
examples of weak scattering. When the scattering is strong, this assumption is
hopeless, for example for the diffracted intensities and rocking curve widths of
near-perfect crystals. Examples of these are the majority of the semiconductor
structures used in industry so it is easy to see why we must use the dynamical
theory for characterisation research. Fortunately, many of the concepts of
kinematical theory, such as structure factor and diffraction geometry, are also
used in the dynamical theory.
4.2
Kinematical theory
4.2.1
The structure factor
We began the discussion of kinematical theory in Chapter 1, showing how the
scattering from atoms is added up with regard to the phase to form the scattering
from a unit cell, the structure factor. We repeat this important equation here. The
structure factor for the hkl reflection is
(4.1)
where ( uvw ) are the fractional coordinates of the vector r , which runs from the
origin of the unit cell to the atom of type i whose atomic scattering factor is ƒ i
and the summation is over all atoms in the unit cell. Let us now look more deeply
at this equation and consider the assumptions made. These are:
1 The scattered intensity is very small. The loss of intensity due to re-
scattering is negligible and thus the refractive index is unity.
2 The point of observation is at a large distance compared with the dimensions
of any coherently illuminated scattering volume.
3 Scattered waves from different atoms are nearly parallel. We label these
with the single wavevector k h.
Conditions (2) and (3) are equivalent to the Fraunhofer or far-field

approximations in ordinary optics. The coherently illuminated region with usual
laboratory X-ray sources is a few micrometres across. We therefore expect this
theory to be useful in the cases of weak scattering but to be seriously awry for
strong scattering.
The atomic scattering factors f i are usually calculated in terms of the
scattering of an individual free electron. This is calculated as if the electron were
a classical oscillator—since the assumption is that the electron is a free charged
X-RAY SCATTERING THEORY 73
particle. It is set into forced oscillation by the radiation field of an incident X-ray
and then re-radiates in all directions at the same frequency as the incident wave
frequency. This is termed elastic or Thompson scattering. For an X-ray beam, the
intensity I scattered by one electron through an angle 2λ relative to the incident
intensity I 0 is
(4.2)
where R is the distance of observation from the particle and r e (the so-called
classical electron radius) is (in cgs units, in which this theory is most concisely
expressed)
(4.3)
where e is the electronic charge, m the rest mass of the electron and c the velocity
of light. C is a factor dependent on the polarisation. If the electric vector of the X-
ray wave is perpendicular to the dispersion plane then C=1; this is known as λ
polarisation. For λ polarisation, in which the electric vector is parallel to the
dispersion plane, C=cos 2λ . For an individual electron, the angular variation of
the scattering only arises through this polarisation term.
It does not matter that electrons in a solid are not classical oscillators, since
this is only a unit of reference, but it brings out the polarisation behaviour. In
fact, electrons in an atom behave surprisingly similarly to classical oscillators
with respect to elastic X-ray scattering. If all the electrons in an atom were
concentrated at one point, then we should just multiply the electron scattering
factor ƒ by Z , the atomic number, to get the atomic scattering factor ƒ i . This is
a good approximation when the scattering angle 2λ is small and thus all the
scattering is nearly in phase. However, atoms have finite sizes compared with X-
ray wavelengths and when 2λ ■ 0 we must add with regard to phase, just as we
added the scattering of atoms with regard to phase to arrive at the structure
factor, equation (4.1). We shall have a similar equation for the atomic scattering
factor, but since the distribution of electrons is continuous we express it as an
integral rather than a summation. In units of ƒ, or electron units (and hence itself
dimensionless), this is:
(4.4)
The scattering vector Q=k h−k 0 where k 0 is the incident beam vector and k h the
scattered beam vector, as illustrated in Figure 4.1.
It is seen from Figure 4.1 that the modulus of the scattering vector, Q, is given
by
(4.5)
Figure 4.1 The vector relationship between the incident beam vector k0, the scattered
beam vector k h and the scattering vector Q. The directions of the vectors correspond to
their real directions, and the magnitudes of k 0 and k h are both 1/λ
In equation (4.4), λ ( r ) dV is the probability that an electron lies in a volume
element dV of the atom at a radial distance r from the nucleus. λ ( r ) is the
electron probability density and is a meaningful and measurable quantity. The
scattered amplitude is therefore the Fourier transform of the electron density. We
see immediately that for a given atom the atomic scattering factor is a function
only of Q. Its angular dependence is important and is shown in Figure 4.2.
4.2.2
The strengths of X-ray reflections
The best equations to use for intensity are those of the dynamical theory, but
kinematical theory gives some useful insights, based simply upon the idea that the
scattering increases monotonically with the structure factor, equation (4.1).
The structure factors of reflections in different cubic structures are shown in
Table 4.1. Note that the structure factors may be complex, but this depends upon
the choice of origin of the unit cell. The intensity formulae always contain the
modulus
Figure 4.2 The variation of the atomic scattering factor f i with scattering angle 2■ . Values
for silicon (Z=14) and germanium (Z=32) are shown
Table 4.1 Values of |F hkl | for a number of useful reflections in cubic structures, with
examples of |F hkl |V in electron units per Å3 for one crystal of each type
Reflection fcc Al Diamond cubic Si Sphalerite GaAs
001 0 0 0 0 0 0
002 4ƒ 0.41 0 0 4(ƒ 1 −ƒ 2 ) 0.05
004 4ƒ 0.28 8ƒ 0.39 4(ƒ 1 +ƒ 2 ) 1.02
111 4ƒ 0.43 5.66ƒ 0.38 0.99
222 4ƒ 0.31 0 0 4(ƒ 1 −ƒ 2 ) 0.04
333 4ƒ 0.21 5.66ƒ 0.24 0.61
011 0 0 0 0 0 0
022 4ƒ 0.35 8ƒ 0.45 4(ƒ 1 +ƒ 2 ) 1.20
044 4ƒ n.a. 8ƒ 0.31 4(ƒ 1 +ƒ 2 ) 0.80
112 0 0 0 0 0 0
224 4ƒ 0.23 8ƒ 0.35 4(ƒ 1 +ƒ 2 ) 0.90
113 4ƒ 0.32 5.66ƒ 0.30 0.79
115 4ƒ 0.21 5.66ƒ 0.24 0.61
of the structure factor |F h |= = (or its equivalent in susceptibility, e.g. and also
1/V, where V is the volume of the unit cell. The modulus formulae are given in this
table. Also given are the actual moduli for aluminium, silicon and gallium
arsenide, to show the effects of material and of the scattering angle; these have
been divided by V to make direct comparison between the materials. A number

of the values are zero, for example the 001 and 002 reflections in the diamond
cubic structure possessed by silicon and germanium. These are often referred to
as forbidden reflections. Reflections such as 002 and 222 in sphalerite, the
structure of a very important class of semiconductors such as gallium arsenide
and indium phosphide, are not forbidden but are weak, since they depend upon
the difference between the atomic scattering factors of the constituent atoms.
They are sometimes called ‘quasi-forbidden’, and are very useful for
emphasising compositional differences in such materials, such as non-
stoichiometry or superlattice structures. An example is the common GaAs/
GaAlAs superlattice, in which the 002 reflection is most useful.
4.2.3
Intensity diffracted from a thin crystal
This is the most useful quantitative intensity formula that may be derived from
kinematical theory, since it is applicable to thin layers and mosaic blocks. We
add up the scattering from each unit cell in the same way that we added up the
scattering from each atom to obtain the structure factor, or the scattering power of
the unit cell. That is, we make allowance for the phase difference r i .Q between
waves scattered from unit cells located at different vectors ri from the origin.
Quantitatively, this results in an interference function J , describing the
interference of waves scattered from all the unit cells in the crystal, where
(4.6)
The summation is taken over all unit cells in the crystal. The overall scattering
amplitude J is given by A=F hkl J (where F hkl is the structure factor for the
hkl reflection) and the intensity I by the square of the amplitude
(4.7)
We may evaluate this for a small crystal. In order to consider the size effects, we
describe the crystal as a parallelepiped of side n 1 a 1, n 2 a 2 and n 3 a 3, where a
1, a 2 and a 3 are the unit vectors defining the unit cell and the n i are the number
of unit cells in each side of the parallelepiped. The interference function becomes
(4.8)
We know from the simple description in Chapter 1 that we shall not get
significant intensity unless the Bragg law is satisfied, so we shall formulate the
problem so that we look at deviations from the exact Bragg condition. From
equation (4.5), the latter is given when Q=1/d hkl , where d hkl is the spacing of a
plane with Miller indices hkl, and whose structure factor is not zero. In the
notation of Figure 4.1, the Bragg geometry and Bragg law are satisfied if the
vector Q=h, where |h|=1/d hkl , and the direction of h is perpendicular to the (
hkl ) planes (we shall see shortly that this is saying that h is a vector of the
reciprocal lattice of the crystal). Simple trigonometry on Figure 4.1 then gives
the Bragg law. We can express deviations from the Bragg law in terms of a
deviation vector q, i.e.
Q=h+q (4.9)
Taking one of the terms, for example that in a 1, the interference function then
simplifies drastically:
a 1 .Q=a 1 .(h+q)=a 1 .h+a 1 .q (4.10)
The first term a 1 . h is the Miller index component h, which is an integer. Thus
in the interference function it becomes unity since exp(2λ ni)=1. This corresponds
to strong Bragg diffraction when q=0. The second term when put in the
interference function becomes
(4.11)
where q 1 is the component of q along the first axis in the diagram of Figure 4.1.
This is in fact the first axis of the reciprocal lattice, which in cubic crystals is
parallel to the real-space axes of the lattice. The final expression for intensity is
(4.12)
At small values of the deviation vector q this reduces to the product of three
functions of sinc2 ( x ) type. It is shown for one component in Figure 4.3.
Figure 4.3 brings out the following features of the scattering from a thin (or
small) crystal:
1 The scattering is only intense near q=0, i.e. an exact Bragg reflection. The
weak scattering from atoms is here reinforced by successive planes of atoms
scattering in phase and strong intensity only occurs due to this
reinforcement. Therefore, X-ray diffraction techniques integrate the
scattering over many atomic layers.
2 The first zero occurs at n i q i =1, i.e. the width of the diffraction peak
varies inversely as the number of atoms. We expect appreciable broadening
of diffraction spots for crystallites less than a few nanometres on edge and
this is indeed observed.
3 The peak intensity, and also the integrated intensity, will be proportional to |
F h |2.
4 The scattered intensity is proportional to the volume of the crystal. This
implies that the scattering from a thin epitaxial layer, large in area compared
with the beam diameter, will be proportional to the layer thickness.
5 For large single crystals extremely narrow rocking curves are predicted by
the kinematical theory and these are not found. Dynamical theory is required
for these cases.
Figure 4.3 The scattered intensity as a function of one component of the deviation vector
q. (a) n 1=10, (b) n 1=1000
4.2.4
Anomalous dispersion
When a phenomenon appears inconsistent with a simple theory, it often acquires
the title ‘anomalous’, which persists long after the theory is properly understood.
The behaviour of any mechanical or electrical oscillatory system has features
near its resonance condition that are very different from those holding in the
(much more usual) off-resonance conditions. For the scattering of X-rays, the
resonant condition occurs near an X-ray absorption edge, when the energy of the
incoming photon is just above that required to excite electrons between energy
levels in the atom. We should expect to see more rapid variations than elsewhere
in both the scattering and absorption behaviour, and these are termed anomalous
dispersion. The treatment may be unified by notating the atomic scattering
factors ƒ i (and structure factors that are derived from them) as complex
numbers. The real part is the scattering that we have discussed so far, and the
imaginary part is the absorption. Since we use exponential notation for the
resulting wave, i.e.
(4.13)
where the complex function is linear with the atomic scattering factors, the
imaginary part will drop out automatically into the usual absorption equation
(4.14)
as discussed in section 1.1. We take the real part to predict the scattering in, for
example, a diffraction experiment. Near the absorption edge, and excluding all
other effects such as Compton scattering, the real and imaginary parts of the
scattering due to an oscillator must be related by a mathematical transformation
known as the Kramers-Krönig transformation.
The effects of an absorption edge are shown in an important case for
semiconductors in Figure 4.4. This shows the real and imaginary parts of the
scattering factors for gallium and arsenic near the K absorption edges. The
exchanges in values, and resultant intensities, that are shown can and have been
used for composition- sensitive experiments such as the determination of
stoichiometry in III–V semiconductors. 1 However, the tunability of a
synchrotron radiation source is required for such experiments in most cases.
A further variant in the scattering near the absorption edge is the extended X-
ray absorption fine structure (EXAFS). The energy absorbed by the material in
exciting an electron from one energy level to another depends, in detail, on the
atomic environment. This may be looked on alternatively as a back scattering
and mutual interference of the emitted electrons, dependent on the environment,
or as a modification of the state to which the ejected electron is excited, again
dependent upon the environment. This is a powerful method of crystal structure
analysis, especially for nanocrystalline or cluster regions in ‘amorphous’
materials. 2 It has only occasionally been combined with high resolution
diffractometry and then only at synchrotron radiation sources.
Figure 4.4 The real (scattering) and imaginary (absorption) parts of the atomic scattering
factor for gallium and arsenic near the K absorption edges. After Cockerton et al . 1
4.2.5
Reciprocal space geometry
We surreptitiously introduced reciprocal space in Figure 4.1, in an attempt to
show its usefulness before giving the formal definitions. It is so very helpful in
the interpretation of many diffraction experiments that we need to understand it
more fully. We shall use it extensively in the discussion of triple-axis
experiments, in which ‘reciprocal space mapping’ is an essential technique.
Figure 4.1 may be slightly extended to show a most useful construction, that
of the Ewald sphere. Since we have elastic scattering, it is always true that the
magnitudes of k 0 and k h are both 1/λ . A sphere of radius 1/λ can therefore
define all possible incident and scattered beam vectors. The incident beam vector
runs from the centre of this Ewald sphere to the origin, the scattered beam vector
runs from the centre to any point on the surface of the sphere, and the scattering
vector runs from the origin to the end of the scattered beam vector. Figure 4.5
shows a two-dimensional section of this three-dimensional construction. In this
figure the scattering vector Q has been made to coincide with a vector h that
satisfies the Bragg law (equation (4.9)) and we expect strong diffraction.
Of course, many reflections are possible from a regular lattice, and if the set of
vectors such as h is represented then we can graphically visualise all the
reflecting planes and consequent reflected directions in the crystal. We have
deduced virtually all the rules already, but we formalise the description below.
Figure 4.5 The Ewald sphere construction in reciprocal space
1 The scale of reciprocal space is reciprocal length. (Note that in most physics
texts this is scaled by a factor 2λ for mathematical convenience in treatments
of quantum theory. Here, and in general in crystallography texts, we use no
such scaling.)
2 All directions in real space are preserved in reciprocal space.
3 A reciprocal lattice vector is constructed for each plane of the real-space
lattice as follows:
(a) the direction of the vector is perpendicular to the plane in real space,
(b) the magnitude of the vector is the inverse of the interplanar spacing in
real space.
4 The end of each such vector, starting from the origin, is a reciprocal lattice
point (sometimes abbreviated ‘relp’).
5 The reciprocal lattice is the set of relps.
Thus the reciprocal lattice axes are perpendicular to the (100), (010) and (001)
planes in the real-space lattice. In cubic, tetragonal and orthorhombic crystals it
is also true that they are parallel to the [100], [010] and [001] directions, but this
is not true in other crystal classes. The general formulae for the reciprocal space
axes, a 1*, a 2* and a 3* in terms of the real-space axes a 1*, a 2* and a 3* are
(4.15)
Some examples will illustrate the concept. These are all drawn to scale for the
silicon lattice and structure. In Figure 4.6 a high resolution experiment is shown.
It is quite convenient to use reciprocal space to show the accessible reflections in
a given experiment. These are shown in Figure 4.7 for a silicon specimen, with a
(001) surface plane, and CuK λ radiation, for two of the orientations of the
incident beam. The large semicircle contains all the points that are cut by the
Ewald sphere as the incident beam is rotated 180° from just grazing the surface
in one direction to just grazing the surface in the opposite direction. The small
semicircle on the left contains points that cannot be accessed in reflection
because the incident beam would enter from below the crystal surface. The small
semicircle on the right likewise cannot be accessed in reflection as the diffracted
beam exits through the crystal. These two regions are accessible in transmission.
The reflection and transmission conditions are often called Bragg case and Laue
case respectively.
In the powder diffraction experiment, a fixed wavelength and incident beam
direction is used, and the random orientation of the powder grains means that the
reciprocal lattice is rotated about all angles, centred on the origin. A diffracted
beam ensues whenever a relp cuts the Ewald sphere. This is shown in two
dimensions in Figure 4.8. This construction also makes it easy to see when
reflections overlap, for example in Figure 4.9 we see the reciprocal space
representation of the Laue back-reflection method. The Ewald sphere here
sweeps through a range of wavelengths from the minimum given by the tube
voltage (largest diameter Ewald sphere) to the maximum, which is controlled by
absorption in the tube windows and in the air path. A crystal plane will diffract if
its reciprocal lattice vector intersects the Ewald sphere at a wavelength within
this range.
These examples show the great practical utility of the Ewald sphere
construction. We did once hear Paul Ewald say, some sixty years after he laid the
basis of X-ray scattering theory, that he wished people had named something
else after him, as it was such a trivial idea!
4.3
Dynamical theory
As we have mentioned, the kinematical theory is unsatisfactory for predicting
intensities of anything other than very thin or very small crystallites. The
assumption that no energy is transferred into the diffracted beam might suffice
for neutrino diffraction, but is very badly wrong for X-ray diffraction from most
crystals of practical interest in the semiconductor industry. It is not merely the
scaling factor that is wrong. There is a saturation in the intensity as the crystal
thickness increases and many additional interference effects caused by the
presence of strong wave-fields that are not predicted by kinematical theory. The
equations describing intensity as a function of thickness become seriously non-
linear. However, thanks to theoretical studies over eighty years, the dynamical
Figure 4.6 A high resolution experiment. In (a) the crystal is not yet aligned to the Bragg
position and no diffracted beam occurs. In (b) the incident beam has been rotated so that
the Ewald sphere falls on the 004 relp, and a diffracted beam ensues. The Ewald sphere is
to scale for CuK λ and the reciprocal lattice is to scale for silicon
theory, which corrects these gross errors, is very well understood. Moreover, the
recent arrival of packages providing numerical solutions on accessible personal
computers means that industrial scientists can obtain precise, quantitative
Figure 4.7 The accessible, allowed reflections in a high resolution experiment, shown in
reciprocal space, for a silicon specimen with an (001) surface, using CuK λ radiation. (a)
incident beam in (100) plane, reciprocal lattice section perpendicular to [100]. (b)
Incident beam in (110) plane, reciprocal lattice section perpendicular to [1 0]. Case (b)
would occur if the incident beam were parallel or perpendicular to the plane of the (110)
flat cut on many semiconductor wafers. Many more reflections are available in case (b). The
semicircular segments showing accessible and inaccessible reflections are sections of
hemispheres whose axes are vertical on this diagram
descriptions of the scattering that can be used in process development and quality
control.
The simulation approach is based upon the idea that the diffracted beam from
a set of reflecting planes is at the right angle to be rediffracted by the same
planes. This is illustrated in Figure 4.10. It is seen that the energy is spread
throughout a triangular region in this section known as the Borrmann fan. We
expect this to result in a complicated expression for intensity, as indeed it does,
though it can easily be implemented on a personal computer if the region is
uniform. If it is non-uniform, as are all interesting industrial materials, this
Figure 4.8 The powder diffraction experiment. (a) Reciprocal space notation. The Ewald
sphere is fixed, and the lattice is rotated about all angles about the origin. Only the
rotations about [100] are shown in this two-dimensional section. Intersections with the
Ewald sphere define the diffracting conditions. (b) The corresponding diffracted beams in
real space
coupling of the diffracted and forward-diffracted beams (we can no longer think
of the incident beam as transmitting unchanged through the crystal) must be
treated locally around the inhomogeneities. The Takagi-Taupin theory is based
upon the formulation and solution of a coupled pair of differential equations
Figure 4.9 The Laue back-reflection method. The direction of the incident beam and the
reciprocal lattice are fixed in space, and wavelengths are selected out of the beam. The
Ewald sphere may be any diameter between the short and long wavelength cutoffs. The
larger circle shows an intermediate wavelength and the diffracted beams that result at this
wavelength
which represent the changes in amplitude in each of the forward-diffracted and
diffracted directions.
Figure 4.10 The diffraction and rediffraction of an X-ray beam from a set of reflecting
planes. The triangle bounded by the incident beam and the diffracted beam from the entry
surface is called the Borrmann fan
The Takagi-Taupin approach and the simulation methods that it allows are so
powerful and useful that they are discussed in detail in the next chapter.
However, in this approach we lose everything but the numerical amplitudes and
intensities. We do not know how beams propagate through the crystal, how and
when they are likely to interfere, what parameters control rocking curve width
and shape; in other words the physics is buried. We therefore begin our
discussion of dynamical theory by discussing the classical physical treatment and
explaining the concepts that result. These will illuminate discussions of many
aspects of experimental measurement and interpretation throughout the rest of
the book. We shall not attempt to give the full derivations, since these are well
treated in many excellent texts 3 − 11 but rather present the basis and
assumptions of the theory and explain and interpret the conclusions and results.
We largely follow the treatments of Batterman and Cole 6 and Hart. 7
4.3.1
Fundamental equations
We are trying to discover what waves are generated inside and outside a crystal
when it is illuminated by X-rays. The propagation of any electromagnetic waves
in any medium is accurately described by Maxwell’s equations. These are, in
vector notation,
(4.16)
where E is the electric field, B the magnetic induction, D the electric flux density,
H the magnetic field and c the velocity of light in vacuo. This assumes that (at X-
ray frequencies) the electric conductivity is zero and the magnetic permeability is
unity, and is in Gaussian cgs units.
We want to introduce the properties of the crystal and of the X-rays and solve
f or the electric displacement or flux density, D. Hart 7 gives a careful discussion
of the polarisability of a crystal, showing that a sufficient model of the crystal for
X-ray scattering is a Fourier sum of either the electron density or the electric
susceptibility over all the reciprocal lattice vectors h. Thus the crystal is
represented as
(4.17)
and λ h can be put in terms of the now-familiar structure factor by
(4.18)
where V is the volume of the unit cell over which the structure factor is
calculated. The X-ray properties are introduced through the wavelength λ in this
equation and we use the observed fact that the X-ray refractive index is very
close to unity to make some simplifying assumptions, such as that λ is very small
at these frequencies. Substituting in the Maxwell equations and eliminating the
magnetic component, we find that
(4.19)
What can we guess about the solution? It should clearly be a wave equation. We
expect (from knowledge of Bragg diffraction) it to be a plane wave, or sums of
plane waves. We use capital K 0 and K h for wavevectors inside the crystal to
distinguish them from k 0 and k h outside the crystal. Inside the crystal the
allowed wavevectors should satisfy conservation of momentum, that is
K 0+h=K h
This Laue condition is a little less restrictive than the Bragg law, in that we no
longer have the condition that |K 0|=|K h|=1/λ , but we still expect strong
diffraction only when we are near the Bragg condition. Ewald proposed, and
Bloch 12 showed that waves that exist in a crystal must have the periodicity of the
lattice, that is, the solutions should look like
(4.21)
with the K h linked by equation (4.20). A qualitative argument in support of the
Bloch theorem is that any waves that do not have this periodicity will be strongly
damped by conflict with the ‘electron density waves’ that are the atoms
themselves. We can further restrict ourselves to the case where two beams are
very much stronger than any others. These are the diffracted and forward-
diffracted beams shown in Figure 4.10. A glance at Figure 4.6 shows that it will
require extremely careful adjustment to excite more than one diffracted wave
with X-rays, so the two-beam assumption is very good. This is in contrast to the
situation with electron diffraction in which the Ewald sphere is almost planar on
the scale of this diagram and many beams are excited in almost all cases. The
solution of the Maxwell equations inside the crystal (see for example Batterman
and Cole) 6 is then expressed in terms of the amplitudes D 0 and D h of these
two beams, and is:
(4.22)
The polarisation factor, C=D 0 .D h. This is unity for ■ polarisation, in which the
electric field vector is perpendicular to the dispersion plane and cos(2λ B) for λ
polarisation in which the electric field vector lies in the dispersion plane. We
may apply equation (4.20), which since h is real implies that Im(K 0)=Im(K h),
and, in order to express the solution simply, introduce the very important idea of
deviation parameters, λ 0 and λ h . These express the deviation of the incident
and diffracted wavevectors from the kinematic assumption where |k 0|=|k h|=1/λ ,
and are defined as
(4.23)
In effect, the deviation parameters define the local refractive index of the crystal.
The solution may now be expressed in terms of the relationship between the
permissible deviation parameters
(4.24)
This equation defines the geometry of diffraction in the two-beam dynamical
theory. The amplitude ratio may also be expressed in terms of the deviation
parameters as
(4.25)
and also that
(4.26)
These are the fundamental equations of two-beam dynamical theory, which allow
us to predict the wavefields and their intensities inside (and outside) the crystal.
Of course, there are many consequential derivations and solutions, especially
when we apply the boundary conditions, which consist of the matching of waves
so that they are continuous across the boundaries, and we shall quote such
equations as the need arises. We shall first interpret these equations visually to
demonstrate the principal qualitative conclusions.
4.3.2
Dispersion space and geometry
We have said that equation (4.24) gives us the allowable deviation parameters.
Waves that satisfy this equation may exist inside the crystal for a given incident
beam wavelength, others may not. We are going to draw a diagram that shows
all the allowable wavevectors near the Bragg condition. Since equation (4.24)
relates wavevectors to the local refractive index, it is a dispersion relationship.
The diagram of all allowable wavevectors is therefore called a dispersion
surface.
We should first correct the wavevector inside the crystal for the mean
refractive index, by multiplying the wavevectors by the mean refractive index (1
+λ /2). This expression is derived from classical dispersion theory. 7 Equation (4.
18) shows us that λ is negative, so the wavevector inside the crystal is shorter
than that in vacuum (by a few parts in 106), in contrast to the behaviour of
electrons or optical light. The locus of wavevectors that have this corrected value
of k lie on spheres centred on the origin of the reciprocal lattice and at the end of
the vector h, as shown in Figure 4.11 (only the circular sections of the spheres
are seen in two dimensions). The spheres are in effect the kinematic dispersion
surface, and indeed are perfectly correct when the wavevectors are far from the
Bragg condition, since if D h ■ 0 then the deviation parameter λ h ■ 0 from
equation (4.25). ‘Far’ means a few minutes of arc at most!
Indeed, Figure 4.11 itself is perfectly correct on the scale of the drawing. All
the effects that the dynamical theory considers take place within the thickness of
the lines at the intersection of the two circles at the point L0. The shift in
wavevector due to the mean refractive index is also undetectable on this
drawing. Let us then magnify Figure 4.11 about a million times; the drawing is to
scale and O and H are now about 50 km to the right. The circles centred on O
and H are now indistinguishable from straight lines. The effect of the mean
refractive index is shown as the shift from the intersection of these lines in vacuo
at the Laue point L to the point L0. However, equation (4.24) tells us that the
deviation parameters, which are measured perpendicularly from these lines, can
only be zero if the susceptibility is zero. In real cases, there must be a deviation,
and its form, again from equation (4.24), is a hyperbola. The circles centred on O
and H therefore do not represent solutions of Maxwell's equation when
diffraction occurs. The hyperbola representing the locus of deviation parameters
does. This is the dispersion surface.
For completeness we note that the dispersion surface is only strictly
hyperbolic when the circles are accurately represented by straight lines. In three
Figure 4.11 The kinematic dispersion surface. The circles centred on the origin O and the
relp H, with radius (1/λ vacuum )(1+λ /2) represent, in the plane shown, the allowable
wavevectors in the crystal far from the diffracting condition. A section of the Ewald
sphere is shown
dimensions, of course, the lines become cones and the dispersion surface
becomes a hyperboloid of revolution. We shall also have two independent
dispersion surfaces for the two polarisations, λ and λ . These must be taken into
account in calculations with a randomly polarised source such as a laboratory
generator, but we shall only draw one of the polarisations in the diagrams.
The dispersion surface has two branches, labelled 1 and 2. Waves from the
two branches are in antiphase. This will be seen later to affect their absorption.
Which of the allowable points (solutions of the Maxwell equation) that are
selected depend on the incident angle and on the boundary conditions as
discussed in the next section. One such point, labelled A, is shown in
Figure 4.12. The values of the deviation parameters at A are shown. The vector
from A to O is K 0 and that from A to H is K h. Once these are fixed, we know
the amplitude ratio from equation (4.25). This equation also tells us that the
diffracted beam intensity will be a maximum when λ 0=λ h , at the centre of the
dispersion surface. However, because of the displacement between L and L0, the
incident beam vector at this point is not quite parallel to the vector given by the
conventional Bragg law calculation. There is a refractive index shift, always in
the sense of increasing the apparent Bragg angle, usually around a few seconds
of arc. The diameter of the dispersion surface is approximately λ h k, about 10−5
k, and thus increases with the structure factor. We shall see later that this is
directly proportional to the width of the rocking curve of the crystal.
It is qualitatively very helpful if instead of having to solve the equations or
simulate the solutions every time, we can get some physical understanding which
reduces the task. Ray optics is one such method; we look at the geometrical
optics before the physical optics. Rays are reasonable concepts if diffraction
broadening is not too great, that is, for a ray path■ ■ 2/λ , where A is the width of
Figure 4.12 Magnification of the region near L0 in Figure 4.11. L is the Lorentz point (the
Laue point corrected for the mean refractive index). A is one of the tie-points selected on
the dispersion surface, and λ 0 and λ h are the deviation parameters at that point (shown on
branch 2)
the beam. For an X-ray beam 0.1 mm wide, this distance is about 1 mm, so the
ray approximation does give a useful start. The direction of the ray is found by
an interesting property of the dispersion surface. It is formally described by the
Poynting vector, parallel to E × H , and Kato 13 , 14 showed that this is
perpendicular to the dispersion surface at the point where the amplitude ratio is
defined. Thus, once the active points (solutions of Maxwell equations) are
selected, the direction of energy flow or of rays is easily found. We shall find
this very useful in X-ray topography in considering the contrast to be expected
from defects at various depths in the crystal.
4.3.3
Boundary conditions
Any point on either branch of the dispersion surface is an equally good solution
of the Maxwell equations. However, the only points that will be selected are
those that allow continuity of the wave—specifically of the electric displacement

D—in all places, including the transition from outside to inside the crystal. The
analogy with water waves may be useful. Here again there is an important
dispersion relation as the velocity of waves in the deep ocean is higher than that
in shallow coastal waters, and the resultant pile-ups are the surfer’s delight.
Nevertheless, the surface of the water is continuous through all these transitions.
Expressed formally, the wave must be matched in amplitude at the surface and
in phase velocity parallel to the crystal surface. This implies that the tangential
components of D and H must be continuous across the surface, and the
components in the crystal surface of the wavevectors inside and outside the
surface must be the same. If n is a unit vector normal to the crystal surface,
whatever the values of k 0 or the resulting Bloch wave inside the crystal, then
K 0−k 0=λ n (4.27)
where λ is a scalar variable. This construction is shown diagrammatically in
Figure 4.13. Just drop a line normal to the surface from the tail of the k 0
wavevector, and the tails of the K 0 wavevector(s) must lie somewhere on this
line. But they must also lie on the dispersion surface. The intersection therefore
gives the excited points, known as tie-points. Figure 4.14 shows the full
construction for the Laue (transmission) case.
In the Laue case there are always two tie-points selected, one on each branch,
labelled A and B. From each tie-point we shall generate wavevectors directed
towards each of O and H. There are thus four wavevectors generated in the
crystal for each polarisation, eight in all. The energy flow through the crystal is
in the direction of the Poynting vector, that is, normal to the dispersion surface at
the tie-points, and only at the exit surface do the waves split up into diffracted
and forward-diffracted beams, as shown in Figure 4.15. We should repeat the
boundary-condition construction for waves leaving the exit surf ace of the
crystal, but it is identical to the entry-surface case if the surfaces are parallel. The
displacement of the forward-diffracted beam from the ordinary attenuated direct
beam is quite large enough to measure on a photographic plate, with suitable
crystal thickness, and this formed one of the experimental tests of dynamical
theory in highly perfect crystals. 15
The Bragg (reflection) case is shown in Figure 4.16. Here the situation is a
little less clear, because, according to the geometry, the normal from the surface
intersects either two tie-points on the same branch of the dispersion surf ace or
none at all. The Poynting vectors associated with the two tie-points are different;
the energy flow from one point is directed into the crystal, but that from the other
is directed outwards. The latter therefore does not generate any wavefields inside
the crystal and can be ignored. Thus a single wavefield is generated for each
polarisation.
In the cases where no tie-points are selected, no wavefields at all (other than a
very rapidly decaying evanescent wave) are generated inside the crystal. The X-
rays are effectively excluded from the crystal and the reflectivity, with a zero-
absorption crystal, is 100%. This is the range of total reflection.
Figure 4.13 Matching of wavevectors across a boundary. The requirement that the
component tangential to the surface is constant is satisfied by the above geometric
construction
4.3.4
Absorption
The above discussion has in effect been for materials with zero absorption, but this
affects only the intensities. The construction of the dispersion surface and the
wavevector matching are all performed on the real part of the wavevectors.
When absorption is considered, the reflectivity in the Bragg case falls below
100% but it can still be over 99% for a low-absorption material such as silicon.
Far from the Laue condition the absorption shows the normal photoelectric
absorption, as would be measured (with allowance for density) in a liquid or gas
of the same atomic species. Close to the Laue condition, the absorption is
quantified by the imaginary parts of the susceptibilities, leading to imaginary
components of the wavevectors. These imaginary components are always normal
to the crystal surface and hence the planes of constant attenuation are parallel to
the surface. The attenuation coefficient μ (n) normal to the surface is given by
μ (n)=−4λ Im(K 0) (4.28)
It is more complex to express the attenuation along the direction of the Poynting
vector, 16 so that we may see the absorption of ‘rays’ in the material, but it is
interesting to quote one result, calculated by Batterman and Cole, 6 for 1 mm
thick
Table 4.2 Absorption (expressed as dimensionless μ t values) for the various beams
resulting from 220 diffraction in a 1 mm thick crystal of germanium
Branch Polarisation μt
2 λ 1.9
2 λ 12.5
1 λ 63.5
Figure 4.14 The selection of tie-points on the dispersion surface for the transmission
(Laue) case, using the construction of Figure 4.13 (branch 2 is on the left, branch 1 is on
the right)
Branch Polarisation μt
1 λ 74.0
germanium with CuK λ radiation. Far from the diffracting condition the normal
photoelectric absorption gives μ t=38, thus the transmitted intensity would be exp
{−38} or 3.14×10− 14 of the incident intensity and quite undetectable. However,
the two branches of the dispersion surface and the two polarisations show
different absorptions when diffraction is important, as shown in Table 4.2 for
220 diffracting planes. The 2λ ray is 15% of the incident intensity and is easily
measurable. This is known as the Borrmann effect or anomalous transmission.
Again, it is not anomalous once the theory is understood, and it is very useful in
the study of thick crystals by X-ray topography, since any defect in the
Borrmann fan will disrupt the high transmission and give contrast in the image.
Figure 4.15 The energy flow through the crystal and the geometry of diffracted and
forward-diffracted beams, for the case shown in Figure 4.14. P A and P B are the Poynting
vectors associated with the tie-points A and B
The values quoted in Table 4.2 are for symmetrical transmission; the 220
planes are normal to the surface, hence from the boundary-condition construction
the tie-points excited are at the diameter of the dispersion surface. The Poynting
vector is perpendicular to h and therefore the energy flow is along the Bragg
planes. The differing absorption coefficients have a simple physical explanation.
The K 0 and K h wavefields have components that travel in opposite directions,
parallel and antiparallel to h , with wavevectors connected by the Laue equation.
As with any other such pair of travelling waves they will set up standing waves
in the crystal, normal to the Bragg planes, and it is straightforward to show that
the periodicity in reciprocal space is just d, the interplanar spacing. It is also
possible to determine the phase of the standing waves. For branch 2 waves, the
nodes of the standing wavefield are on the atomic planes themselves whilst for
branch 1 they are between the atomic planes, as illustrated in Figure 4.17. It is
thus reasonable that photoelectric absorption, which is caused mainly by
interaction with the inner K and L electron shells, should be a minimum when
the antinodes of the standing wave (the regions of high electric field) fall
between the atomic positions rather than on the atomic planes themselves. The
Figure 4.16 The selection of tie-points on the dispersion surface for the reflection (Bragg)
case, using the construction of Figure 4.13
interesting and useful point is the dramatic effect that this shows in the
magnitude of the absorption.
4.3.5
Penetration depth
The penetration in the absence of absorption is governed by the extinction
distance, λ h . This is the depth at which the intensity drops by a factor 1/e in a
perfect crystal, and also the depth at which the Bloch wave in the crystal changes
phase by a factor 2λ .
(4.29)
Absorption effects combine with this to make the beam attenuate more rapidly.
In the absence of extinction, the absorption depth t λ is given by
(4.30)
where μ is the linear absorption coefficient and λ is the angle between the
incident beam and the crystal surface.
4.3.6
Spherical waves
As has already been emphasised, it is very difficult to approach a plane wave
with the laboratory sources, which are properly described as spherical-wave
sources. A single-crystal reflectivity profile from such a source will be
dominated by the source profile rather than by the above formulae; the integrated
intensity expressions will be multiplied by the source profile function.
The effect that a divergent source has inside the crystal is of great importance for
X-ray topography, and it is useful to consider it qualitatively. Since the rocking
curve profile corresponds to exciting tie-points on the dispersion surface
sequentially over its hyperbolic range (Figure 4.18), it follows that a source with
divergence greater than the rocking curve width will excite all the points on the
dispersion surface within this range. A graphic way of putting it, due to Kato, 14
is that the whole dispersion surface is illuminated. As shown in Figure 4.18(b)
this gives rise to energy flow within the Borrmann fan bounded by the incident
and diffracted beam directions. This should be compared with Figure 4.10. This
concept will be most useful in the interpretation of transmission X-ray topographs.
Defects at any point within the Borrmann fan may contribute to the change of
diffracted intensity at the exit surface of the crystal, and hence to image contrast.
4.4
Scattering in imperfect crystals
We will not treat in any detail in this chapter the more advanced topic of
scattering of X-rays by distorted crystals. Appropriate theory will be introduced
in later chapters as required. In this section we will merely give qualitative
descriptions of the effects of various imperfections and quote some results.
4.4.1
Thermal vibrations
The atoms in a crystal are not in fixed positions but in continual motion due to
thermal vibration. This has the effect of reducing the strength of the scattering.
Zachariasen 9 shows that the effect on Laue-Bragg scattering is to replace the
atomic scattering factors ƒ i by g i , the latter being given by the expression
(4.31)
Figure 4.17 The standing wavefields set up in symmetrical Laue-case reflections. The
nodes of wavefields from branch 2 of the dispersion surface lie on the atomic planes and
the wavefields experience low absorption. The antinodes of wavefields from branch 1 of
the dispersion surface lie on the atomic planes and the wavefields experience high
absorption
The exponent M k depends on the mean square displacement of the atom from its
equilibrium position and hence upon temperature. It is linear with ( kT / m )(sinλ /
λ )2 where k is the Boltzmann constant, T the absolute temperature, λ the
scattering angle, λ the wavelength and m the atomic mass (for a monatomic
material). In addition there are complicated expressions dependent upon the
crystal symmetry. As an example, for silicon at room temperature the ƒ i are
reduced by approximately 6%. With this correction all the equations of
dynamical theory still apply.
Figure 4.18 The effect of spherical incident waves on the excitation of Bloch waves. (a)
Reciprocal space: the divergent incident beam has wavevectors ranging from P 1 O to P 2
O. (b) Real space: energy is distributed throughout the Borrmann fan ABC. The beams
generated outside the crystal are indicated
4.4.2
Thermal diffuse scatter
When the atoms are displaced from their lattice positions, they obviously spend
part of the time in different positions. This means that they scatter X-rays in
different directions from those predicted by the Bragg law. This is the thermal
diffuse scatter. The maximum in the thermal diffuse scatter is displaced by
typically a few minutes of arc from the Bragg positions, and is some six orders
of magnitude weaker. In certain crystal structures, the time-averaged lattice
vibrations may not have the same symmetry as the fundamental lattice and this will
furthermore lead to diffuse Bragg reflections, which may appear between
reciprocal lattice points of the fundamental structure.
4.4.3
Mosaic structure
Mosaic structure is the commonest defect in most real materials. A mosaic block
is assumed to be a perfect crystal of a certain size, separated by an abrupt
orientation change (larger than the angular width of the rocking curve) from
adjacent blocks. These different blocks scatter independently. The effect on the
scattering depends upon the scale of the mosaic blocks with respect to the
extinction distance. If they are significantly larger than the extinction distance
then dynamical scattering formulae apply to each block, and the incident beam is
attenuated by this scattering before it reaches the next block. This is known as
primary extinction. If they are significantly smaller than the extinction distance
then kinematical scattering formulae apply and the beam is only attenuated by
photoelectric absorption; this is secondary extinction. Extinction calculations for
mosaic crystals are important in crystal structure analysis of imperfect crystals,
but are less significant for high resolution diffractometry, We will have occasion
to consider the scattering from mosaic regions, for example in reciprocal space
mapping, but we rarely need to perform quantitative calculations on a three-
dimensional assembly of mosaic crystals.
4.4.4
Dislocation scatter
One case that may be considered as mosaic-region scattering is the distorted
region around an isolated dislocation. Two limiting cases are distinguishable. If
the dislocations are well separated, that is, the kinematical scattering from a
given dislocation is neglible when that from a neighbouring dislocation is
significant, then the distorted volumes within the rocking curve width may be
calculated and the contribution added. 17 If the dislocation strain fields overlap
significantly, then a correlation function describing the averaged displacement
fields must be constructed, from which the scattering may be calculated. With
modern triple-axis diffraction equipment, the scattering from dislocation groups
18 and even from individual dislocations 19 is strong enough to measure, and
quantitative information about dislocation density may be obtained.

Figure 4.19 The range of strong diffraction in (a) the Laue case, (b) the Bragg case. The
heavy lining shows the region of the dispersion surface excited as the rocking curve is
traversed
4.5
Formulae for rocking curve widths, profiles and intensities
In this section we collect all the useful formulae for describing the shapes,
widths and integrated areas or rocking curves.
There will only be a significant diffracted wave if both the deviation
parameters are significantly greater than zero (equation (4.25)). A graphical way
of looking at this is that this is the region in which the dispersion surface is
hyperbolic, i.e. differing significantly from the circles centred on O and H.
Figure 4.19 shows how the tie-points are selected in each case as the range of
reflection is traversed. Note that tie-points on both branches are excited in the
Laue case, but on one branch or the other in the Bragg case with, in the latter, a
range of total reflection in which no tie-points are excited and the wave is
excluded from the crystal.
In Figure 4.19 the hyperbolic region is shown separately to clarify its
dependence on the diameter ■ of the dispersion surface. Since K 0 is large, we may
write
(4.32)
Since h=2K 0 sin λ B (this is the Bragg law),

(4.33)
for the symmetric case (reflecting planes parallel or perpendicular to the
surface). In the asymmetric case the dispersion surface is ‘viewed’ obliquely,
which will give a wider or narrower rocking curve for grazing incidence or
grazing exit, respectively (Figure 4.20).
Figure 4.20 The dependence of the range of strong diffraction on the diameter of the
dispersion surface. (a) Symmetric reflection. (b) Asymmetric reflection
Equation (4.33) is at least approximately true for both transmission and
reflection, but the details of the rocking curve are different in each case. The
individual expressions for the profile are complicated to derive, and we shall
simply quote the results. For generality the complete expressions for asymmetric
diffraction are given, using the direction cosines λ 0 and λ h . These are defined as
the cosines of the angles made by the incident and diffracted beams,
respectively, with the inward-going surface normals (Figure 4.21). These are
general definitions, which may be simplified when the actual geometry is
known. The asymmetry factor b is given by
b=λ 0/λ h
Figure 4.20 (cont.)

where b=1 for symmetric Laue case and b=−1 for symmetric Bragg case
diffraction.
4.5.1
Bragg case formulae
The rocking curve half-width w (FWHM) is given by
(4.34)
Note that if w s is the width of the rocking curve for the symmetric reflection,
Figure 4.21 Definition of direction cosines λ 0 and λ h. (a) Laue case, (b) Bragg case
(4.35)
The simplest expression for the rocking curve profile is in terms of a deviation
parameter λ , which varies from −1 to +1 in the range of total reflection. The
reflectivity as a function of angle is then
(4.36)
where λ is defined as
(4.37)
for the general asymmetric case in zero absorption (i.e. λ 0 and λ h are real). The
positive square root corresponds to tie points on branch 1 and the negative root to
branch 2. Adding absorption we obtain real and imaginary parts of the deviation
parameter:
(4.38)
(4.39)
where λ =λ h ■ /λ 0■ .
Examples of rocking curves with low and high absorption will be shown in the
next chapter.
The integrated intensity I hi in the asymmetric Bragg case without
absorption is
(4.40)
and with absorption is
(4.41)
where
(4.42)
(4.43)
and
(4.44)
Figure 4.22 A plot of intensity as a function of deviation parameter and of crystal

thickness
4.5.2
Laue case formulae
The Laue case rocking curve width in the limit of a thick crystal and zero
absorption is
(4.45)
The rocking curve profile in the asymmetric Laue case without absorption is
(4.46)
where R 0 and R h are the reflectivities of the forward and diffracted beams
respectively.
Note that the intensity is now periodic with both the crystal thickness t and the
reciprocal of the dispersion surf ace diameter, 1/■ . This is illustrated vividly in
Figure 4.22, which shows the rocking curve as a function of crystal thickness.
The integrated intensity in the Laue case without absorption is given by
(4.47)
where J 0 is the zero-order Besel function.

Here is seen the most obvious contrast with the kinematical theory. Both
curves are plotted as a function of thickness in Figure 4.23. It is seen that the
formulae give similar results for small thicknesses, that the kinematic theory
diverges drastically at larger thicknesses, and that the dynamic intensity shows
oscillations about a saturated level after a thickness of about λ h .
With absorption the reflectivity is given by
(4.48)
where
(4.49)
and
(4.50)
4.6
Summary
We have seen that we may obtain a quantitative description of X-ray propagation
and scattering in perfect crystals. This theory is very well established and its
accuracy has been tested very many times over more than 80 years. Crystals such
as silicon and germanium are perfect enough to show quantitative agreement
with the theory at least over seven or eight decades range of scattered intensity.
The theory is routinely used for the design of X-ray optics and the modelling of
scattering in epilayer crystals.
References
. S.COCKERTON, G.S.GREEN & B.K.TANNER, Mater. Res. Soc. Symp. Proc., 139 ,
65 (1989).
Figure 4.23 The transmitted integrated intensity as a function of thickness, as described

by kinematical and dynamical theories
. G.S.BROWN, in: Synchrotron radiation research, eds. H.WINICK & S.DONIACH

(Plenum Press, New York 1980), p. 387 ; J.EVANS, in: Catalysis, Royal Society of
Chemistry, Specialist Periodical Report 8 , 1 (1989).
Connecticut, 1982).
. A.AUTHIER, in: Advances in structure research and diffraction methods, Vol. 3 , eds.
BRILL & MASON (1970) p. 1.
. A.AUTHIER, in: X-ray and neutron dynamical diffraction: theory and applications, eds.
A.AUTHIER, S.LAGOMARSINO & B.K.TANNER (Plenum Press, New York,
1997), p. 1.
. B.W.BATTERMAN & H.COLE, Rev. Mod. Phys., 36 , 681 (1966).
. M.HART, in: Characterization of crystal growth defects by X-ray methods, eds. B.K.
TANNER & D.K.BOWEN (Plenum Press, New York, 1980) p. 216.
. Z.G.PINSKER Dynamical scattering of X-rays in crystals (Springer, Berlin, 1977).
. W.H.ZACHARIASEN, Theory of X-ray diffraction in crystals (Wiley-Dover reprint,
New York, 1945).
. M.HART, Rept. Prog. Phys., 34 , 435 (1971).
. M.VON LAUE, Rontgenstrahlinterferenzen (Akad. Verlag, Frankfurt, 1960).
. J.M.ZIMAN, Principles of the theory of solids (Cambridge University Press, London,
1964).
. N.KATO, Acta Crystallogr., 11 , 885 (1958).
. N.KATO, Acta Crystallogr., 13 , 349 (1960).
. A.AUTHIER, Comptes Rendues Acad. Sci. Paris, 251 , 2003 (1960); A.AUTHIER, F.
BALIBAR & Y.EPELBOIN, Phys. Stat. Sol., 41 , 225 (1970).
. U.BONSE, Z.Phys., 177 , 385 (1964).
. V.M.KAGANER, R.KOHLER, M.SCHMIDBAUER, R.OPITZ & B.JENICHEN, Phys.
Rev. B, 55 , 1793 (1997).
. E.KOPPENSTEINER, A.SCHUH, G.BAUER, V.HOLY, G.P.WATSON & E.A.

FITZGERALD, J.Phys. D: Appl. Phys., 28 , A114 (1995).
. M.S.GOORSKY, M.MEHKINPOUR, D.C.STREIT & T.R.BLOCK, J.Phys. D: Appl.
Phys., 28 , A92 (1995).
5
Simulation of X-ray Diffraction Rocking
Curves
In this chapter we discuss the computational implementation of the

dynamical theory of diffraction: Takagi-Taupin theory, the solution
for grids and thin layers, HRXRD simulation, deviation parameters,
strategies, effect of strain, dislocation and defect arrays. We then
give a number of examples of simulations.
5.1
Introduction
In Chapter 3 we went as far as we could in the interpretation of rocking curves of
epitaxial layers directly from the features in the curves themselves. At the end of
the chapter we noted the limitations of this straightforward, and largely
geometrical, analysis. When interlayer interference effects dominate, as in very
thin layers, closely matched layers or superlattices, the simple theory is quite
inadequate. We must use a method theory based on the dynamical X-ray
scattering theory, which was outlined in the previous chapter. In principle that
formulation contains all that we need, since we now have the concepts and
formulae for Bloch wave amplitude and propagation, the matching at interfaces
and the interference effects.
Unfortunately, the classical dynamical theory, whilst greatly illuminating the
physics of wave propagation, is not very convenient to use. There are many
tedious calculations and, in particular, wave matching and beam multiplication
problems to be handled at every interface. If the theory had remained in that state
it is certain that high resolution X-ray diffractometry would not be the major tool
that it is today. Ideally we want a theory that will take in our experimental
rocking curves and give out the structure. That is not possible yet; it is one of the
class of inverse problems with limited experimental information (in particular the
phase of the X-rays and the limited sampling of reciprocal space which a rocking
curve provides). What we do have is a practical theory that can be implemented
on personal computers for simulating the rocking curve of a material whose
structure is known. Comparison of the features and then the intensities of the
simulated and experimental curves permits iterative refinement of the simulated
structure. Simulation of rocking curves is an extremely powerful method of

interpretation of complex structures. In addition, it is very valuable for the design
of experiments, optimisation of data collection strategy and education of new
researchers and operators in high resolution diffractometry.
5.1.1
The Takagi-Taupin generalised diffraction theory
This generalised diffraction theory, developed independently by Takagi 1 , 2
and Taupin 3 can be used to describe the passage of X-rays through a crystal
with any type of lattice distortion. This theory is not formulated in terms of
wavefields, but is a multiple scattering theory, familiar to electron microscopists
as the Howie-Whelan equations. It assumes that X-rays are propagating as plane
waves and that scattering is occurring both into and out of the diffracted beam.
While this appears to be physically wrong, we must remember that Bloch waves
are just a sum of plane waves in the forward and diffracted directions. Thus
mathematically we can always keep the account straight by adding the correct
phase factors into the scattering from forward to diffracted wave and vice versa.
The important thing about this theory for simulating the rocking curves of
multiple and multilayers is that we assume a single wavevector and no matching
at the boundaries needs to be done. So at the expense of understanding what
physically goes on inside the crystal we have a mathematical calculating method
which works splendidly. The theory can be applied equally well to deformed and
distorted crystals as to a perfect crystal. It has therefore become the most
powerful method both for interpreting rocking curves of complex epilayer
systems and for simulating the contrast of defects such as dislocations in X-ray
topographs. Using simulation techniques you can relate the scattered X-ray
intensity to the microscopic lattice strains in the crystal.
The wavefield inside the crystal is described in a differential form; D 0 and D
h now are the total amplitude of the wave in the forward and diffracted beam
directions, which may be slowly varying functions of position. How slowly?
Surprisingly, the theory works very well for changes as abrupt as stacking faults
and relaxed epilayer interfaces. K 0 and K h are again the incident and scattered
wavevectors. We also take |K 0|=nk, where n is the refractive index far from
Bragg reflection. As is usual and accurate in most X-ray diffraction we use the
two-beam approximation, that is, only the forward and diffracted beam
wavefields have appreciable intensity. With these assumptions, Takagi and
Taupin took a modified Bloch wave representation of the wavefield and obtained
two coupled second-order partial differential equations expressed along the
forward and diffracted beam directions s 0 and s h (these are unit vectors in the
directions of K 0 and K h):
(5.1)
SIMULATION OF X-RAY DIFFRACTION ROCKING CURVES 113
(5.2)
where C is the polarisation factor and λ h represents the deviation of the incident
wave from the exact Bragg condition. This is a key parameter as it is this that we
vary when we scan a specimen to collect the rocking curve. We treat mismatched
layers, graded layers and all kinds of defects by their effect on a local distorted
reciprocal lattice vector and hence on the deviation parameter.
For a perfect, uniform crystal, whether in bulk or as a thin layer, the Takagi-
Taupin equations can be solved exactly as given in the next section. For the
general case with multiple layers, however, it is necessary to integrate them
numerically. The concepts of the dispersion surface are lost, and we cannot tell
directly in which directions wavefields are propagating. They do give directly the
intensities of the direct and diffracted beams emerging from the crystal, and all
interference features are preserved.
5.2
Numerical solution procedures
All that we need to do now is to solve the Takagi-Taupin equations and plug in
the deviation parameter to predict any intensity. For the general distorted crystal,
numerical solution is used. The most common is that over a grid of points, which
may be distorted in any way, for example to model the strains caused by a
dislocation or precipitate at a given depth and hence to simulate the X-ray
topographic image. This is discussed in Chapter 8. There have been many attempts
to solve Takagi’s equations analytically. The only solutions of real interest to us
here are for the perfect crystal and the thin layer.
5.2.1
Thin-layer and substrate solutions
Halliwell, Juler and Norman obtained an important solution of the Takagi-Taupin
equations for a uniform layer of known composition, structure and thickness. 4
This allows any one-dimensional strain distribution to be obtained by splitting up
the crystal into lamellae of constant strain. The solution is expressed in terms of
the variables
A=Cλ −h (5.3)
(5.4)
(5.5)
E=−Cbλ h (5.6)
where b=λ 0/λ h . We also write
(5.7)
These are all complex variables since the susceptibilities are complex. We obtain
the amplitude ratio at the top (exit) of the layer, X, in terms of that at the bottom
(entrance), Xλ:
(5.8)
The variable z is the depth above the depth w at which the amplitude ratio is the
known value Xλ, in effect the thickness of the layer or lamella. We know the
amplitude ratio deep inside the crystal; here, both the diffracted intensity and the
incident amplitude are zero, but since their ratio must always be <1 then the
amplitude ratio Xλ=D h /D 0 ( z ) must also be zero. Using the above we may
derive the reflectivity of a thick crystal,
(5.9)
The parameters A, B, D and E are of fundamental importance. They depend upon

the crystal susceptibilities λ 0, λ h and λ −h , the cosines of the inclination angles
of the incident and diffracted beams with the inward-going surface normals (λ 0
and λ h , with the asymmetry factor b=λ 0/λ h ), the polarisation factor C, the
wavelength A and the deviation parameter λ h . Except for λ h , we may look up
or calculate all of these from the composition and structure of the material and
the geometry of the experiment. Thus the problem is reduced to the calculation
of the deviation parameter.
5.2.2
Calculation of strains and mismatches
At reasonably large angles the deviation parameter is given by: 5
λ h =−2■ λ h sin 2λ B (5.10)
where ■ λ h is the local deviation from the exact Bragg angle, taking account of
lattice strains (with the sign convention that incident angles below the Bragg
angle are negative) and λ B is the local exact vacuum Bragg angle. At small
angles, which are very important for analysing thin layers, a better expression is:
6
(5.11)
The calculation of the Bragg angle and the deviation parameter requires a
rigorous procedure for dealing with strains and mismatches, whether caused by
composition changes or (for example) ion implantation. A suitable method is as
follows. All deviation angles are referred to the ideal Bragg angle of the
substrate. The procedure for calculating the deviation parameter for an arbitrary
layer is then:
1 Find the unit cell of layer material in the fully relaxed (i.e. bulk) condition
from materials databases. For alloys of more than one material, Vegard’s law
is applied to the lattice parameters, Poisson ratios and structure factors. Find
the susceptibility of the layers (the extremely small change in susceptibility
when a layer is strained may be ignored).
2 Calculate the strains λ xx and λ yy that would be applied if the lattice parameters
in the interface plane of the layer were forced to conform to the substrate
(full coherent epitaxy). Multiply these by (1−R) where R is the (fractional)
relaxation of the layer. (See the discussion of measurement of relaxation in
Chapter 3.)
3 Calculate the layer strain normal to the interface, λ zz from the relationship 7
(5.12)
4 Apply the strains to the layer material, giving a new unit cell. Hence obtain
the Bragg angle of the layer, the difference ■ λ between the Bragg angle of
the layer and that of the substrate, and the angle of tilt between the
diffracting planes in the layer relative to the substrate. The deviation
parameter λ h is then calculated separately for the different diffraction
geometries, as follows:
Symmetric case:
λ h =■ λ s −■ λ (5.13)
Asymmetric, glancing incidence:
λ h =■ λ s−(■ λ −■ λ ) (5.14)
Asymmetric, glancing exit:
λ h =■ λ s−(■ λ +■ λ ) (5.15)
where ■ λ , the deviation from Bragg angle in the substrate, is the controlled
parameter, which forms the abscissa of the eventual rocking curve graph.
Note that the procedure in (3) is more general than the more usual approximate
formula:
(5.16)
m and m* are the relaxed and X-ray mismatches. Note that the X-ray mismatch is
not the strain in the layer but the mismatch between the substrate and the layer in
the strained condition. This formula neglects second-order terms, which can
become large for layers with large mismatch, and is less suitable for
incorporating relaxation. The new procedure is much more general, allowing
accurate calculation of asymmetric reflections and avoiding approximations in
implementation at large values of the deviation parameter.
It is seen from the above equations that a tetragonal distortion is assumed in

the layer. This is only strictly true if the substrate surface orientation is (001),
though for symmetric reflections the treatment is valid for any orientation under
the assumption of isotropic elasticity. However, the distortion would, for
example, be trigonal on a (111), surface which would seriously affect the Bragg
angle calculations.
5.2.3
Calculation of reflectivity
Thus we calculate the reflectivity of a whole layered material from the bottom up,
using the amplitude ratio of the thick crystal as the input to the first lamella, the
output of the first as the input to the second, and so on. At the top of the material
the amplitude ratio is converted into intensity ratio. This calculation is repeated
for each point on the rocking curve, corresponding to different deviations from
the Bragg condition. This results in the plane wave reflectivity, appropriate for
synchrotron radiation experiments and others with a highly collimated beam from
the beam conditioner.
5.2.4
Approximations and limitations
The dynamical diffraction theory is very accurate and highly practical, but no
theory provides a perfect description of the complexity of real specimens. What
are the limitations and approximations, and how serious are they?
5.2.4.1
Two-beam approximation
Most practical simulation programs use only two beams, the forward and the
forward-diffracted. That is, only one strong beam is present other than the
incident beam. For most cases this is an excellent assumption, since it is quite
difficult to set up a true multi-beam X-ray diffraction experiment. This is in
contrast to the situation in transmission electron microscopy, where it is almost
impossible to avoid multiple diffraction, and is caused by the longer wavelength
(sharply curved Ewald sphere) used in most X-ray experiments. For very high
energy diffraction, for example at advanced storage rings, multiple diffraction
may become more significant but it may be ignored in ordinary characterisation
experiments.
However, at grazing incidence angles near the critical angle the reflected beam
becomes important, and eventually becomes dominant. This is an important
effect and the two-beam case is seriously in error for angles of incidence close to
the critical angle. Fortunately, the reflected beam decreases in intensity as the
fourth power of the scattering angle, and is weak compared with the diffracted
beam (close to a Bragg peak) above about two or three times the critical angle.
The latter is 0.22° for silicon and 0.31° for GaAs, with CuK λ radiation and is
linear with wavelength.
5.2.4.2
Vegard’s law
It is common to approximate alloy behaviour in both lattice parameter and
Poisson ratio by a linear interpolation between the end points of the alloy
system, e.g. the values for Ga0.5Al0.5As will be taken as half-way between
those for GaAs and AlAs. In the absence of extensive experimental data, which
are not available, this is the only practical general method. There is no general
theory that describes departures from linear behaviour (known as Vegard’s law,
originally proposed 8 for the lattice parameter variations in well-behaved ionic
crystals!). In systems that have been studied, the lattice parameter interpolation
turns out to be reasonable. For example the silicon-germanium system departs only
about 2% from Vegard’s law for any mixture. 9 There is very little information
available on the behaviour of the Poisson ratio with alloy content in
semiconductors, but elastic constant behaviour is known to depend strongly on
electronic interactions. It would not be surprising if there were large deviations
from Vegard’s law in certain pathological cases. It is of course impossible to
estimate the magnitude of this error in ordinary cases without hard experimental
evidence. Vegard’s law errors are, of course, not in the X-ray scattering theory
but in the data put in to model the structure. If good data are available this is not
a source or error.
5.2.4.3
Dislocation broadening
At present this may only be modelled approximately, as discussed in Chapter 4,
rather than fully incorporated into the dynamical theory. This is an important cause
of peak broadening in ‘difficult’ materials such as strained layers.
5.2.4.4
Neglect of diffuse scatter
As also discussed in Chapter 4, dislocations also give rise to diffuse scatter from
the near-core region. Point defects of all types also give diffuse scatter. Although
we do not yet have complete theories to describe this at present, we can model
the diffuse scatter quite well by a Lorentzian function of the form:
(5.17)
In some cases, a higher power Lorentzian can be found to give a better fit.
5.2.4.5
Computational approximations
These are remarkably few in current commercial simulation packages (though
they do differ in the extent to which they incorporate all the effects discussed in
this chapter). The main sources of error are in the incorporation of the instrument
function in which there is a significant trade-off between speed and accuracy for
large data sets.
5.2.4.6
Overall accuracy and sensitivity
Although there is an uncertainty that we cannot quantify if Vegard’s law is used,
it is clear that other errors are below about 1% if the specimen is of good quality
and is accurately described. The very large number of successful analyses by
means of simulation, leading to successful operation of semiconductor devices
that are critically dependent on alloy content, leads us to believe that Vegard’s
law errors are also at the per cent level. We therefore expect that for a reasonably
good-quality specimen with dislocation density below about 103 cm− 2 we should
be able to obtain 2−5% agreement between simulation and experiment, for
overall positions and intensities.
This is not to be confused with sensitivity. Certain features will show far
higher sensitivity than this implies, as we shall see. The sensitivity is always
enhanced for weak structural features if it can modulate the effects caused by
strong features. Examples are, peaks that appear close to the substrate peak and
modify its shape; 10 or very thin layers with large atomic number differences to
adjacent layers.
5.3
Matching an experimental measurement
For laboratory-based systems the instrument function given by the effect of the
beam conditioner must now be introduced. In Chapter 2 we discussed beam
conditioners in detail and showed that they may be characterised in terms of an
intensity which is a function of both divergence and wavelength.
5.3.1
Effect of beam conditioner
We first consider the beam divergence. As discussed in Chapter 2 we must
perform a mathematical correlation (often miscalled convolution) between the
plane wave rocking curves of the beam conditioner and of the sample crystals. If
R 1 (λ , λ ) and R 2 (λ , λ ) are the reflectivities (in intensity) of the first and
second crystals as func tions of the angle of incidence λ , at a given wavelength, λ .
then the total double-crystal reflectivity R ( λ , λ ) at any angle λ of the first crystal
relative to the second crystal is given by 11
(5.18)
The denominator (normalising constant) is the integrated reflectivity of the first

crystal, and the λ dependence takes account of the spectral distribution of the
source (see the next section). The integrals are strictly taken to infinity on either
side, but have to be truncated to a reasonable value consistent with the accuracy
required.
The second property of the beam conditioning system is the wavelength
dispersion. When the beam conditioner is not the same crystal as the sample used
in the non-dispersive geometry, we observe a broadening of the rocking curve as
discussed in Chapter 2. Each wavelength will have its own independent rocking
curve. In the non-dispersive (+ n, − n ) geometry, or if the bandpass of the beam
conditioner is very narrow, it is not necessary to include this effect. In the
general case we must integrate the reflectivity at any point on the rocking curve
over the wavelength:
(5.19)
Since the R ( λ ) are all independent there is no normalising factor. The common
case is the transmission of the the two K λ lines in the dispersive (+ n, − m )
geometry. If the Bragg angles of specimen and reference are very different, then
the K λ 2 and K λ 1 lines are separated and all peaks become doublets with height
ratios 1:2. To simulate this, it is adequate to simulate two rocking curves for the
wavelengths corresponding to the K λ 1 and K λ 2 lines and add the results in the
ratio of 1:2.
However, if the dispersion is small, as in the case, for example, of GaAs and
InP crystals mixed in the 004 reflections as reference and specimen, the rocking
curve simply broadens. Then to a reasonable approximation, one can simulate
this by adding a curvature equal to the angular separation of the K λ 1 and K λ 2
Bragg angles.
5.3.2
Effect of detector noise
A small addition to a logarithmic curve will be very evident at the lower end of
the curve and virtually invisible at the top. For good matches with experimental
curves plotted on a logarithmic scale (to see small peaks) it is necessary to add a
background noise. With a well set up system, this is essentially the detector
background, which may be measured at an angle remote from the substrate peak.
Figure 5.1 A comparison between experimental and simulated rocking curves with and
without the inclusion of background. Si specimen 004 reflection CuK λ radiation
An automatic normalisation of the simulated curve to the substrate peak value

and the lower tail value of the experimental curve may also be useful. Figure 5.1
shows this effect.
5.3.3
Effect of sample curvature
If the specimen or reference crystal is bent, and this can occur with wafers very
easily if they are stuck onto the sample holder with too much adhesive, then the
rocking curve is broadened. This occurs because different parts of the sample
satisfy the diffraction condition at different angles. This may be simulated by
correlating the rocking curve with a rectangle (in angle space) of width equal to
the change in the incidence angle from one side of the beam to the other. The
error introduced by assuming that the incident beam intensity is a rectangular
function of angle is negligible. Figure 5.2 shows this effect.
5.4
Limitations of the simulation approach
The whole object of performing an X-ray scattering experiment is to be able to
deduce the microscopic strains from the scattered intensity. However, except in a
Figure 5.2 A (normalised) comparison between experimental and simulated rocking

curves. InP specimen 004 reflection, CuK λ radiation. (a) Comparison without
incorporation of curvature, (b) comparison with incorporation of curvature of 10 arc
seconds. x-axis scale in arc seconds
few special cases, the phase information is lost due to the fact that we can only
measure intensity, not amplitude, of the scattered wave. Thus, direct
determination of the structure by a Fourier-type transformation is not possible.
Instead, we must adopt the method of simulating the X-ray scattering distribution
from a model structure and comparing this with experiment. By sequential
adjustment of the structural parameters, we arrive at the best match to the
experiment and in an act of faith, conclude that the structure of the material
under investigation is that of the model. Notice that this method is adopted both
in the interpretation of high resolution X-ray diffractometry data and X-ray

diffraction topographs.
There are a number of unsatisfactory features about this procedure which it is
important to examine. The first is the uniqueness of the solution. From a
fundamen tal viewpoint, we may believe that the Uniqueness Theorem in
electromagnetism suggests that there is indeed only one possible perfect match
between experiment and simulation. However, even if this is the case, we can
never have sufficiently perfect data for this stringent condition to be valid. All data
are intrinsically statistically noisy, have a non-zero background and a finite range
of wavevector covered. In practice, there can be no truly unique solution and this
immediately leads to the second problem, that of local minima.
To obtain a match between experiment and simulation, we search for a global
minimum in the difference between the two data sets. When many parameters
are variable, such as in a superlattice structure, there appear a large number of
local minima in this difference. Many of these can be very shallow and very
difficult to leave; conversely the global minimum is often very deep but of very
limited range in parameter space. In practice then, you need to be very close to
the correct solution before the simulation will converge on experiment. At
present, use of automatic fitting algorithms such as downhill Simplex is very
hazardous, and while there are attempts to find algorithms which will identify
local minima, a satisfactory solution to the problem appears to be some way off.
A third problem with simulation of high resolution diffraction data is that there
is no unique instrument function. In the analysis of powder diffraction data, the
instrument function can be defined, giving a characteristic shape to all diffraction
peaks. Deconvolution of these peaks is therefore possible and fitting techniques
such as that of Rietveld can be used to fit overlapping diffraction peaks. No such
procedure is possible in high resolution diffraction as the shape of the rocking
curve profile depends dramatically on specimen thickness and perfection. Unless
you know the answer first, you cannot know the peak shape.
5.5
Examples of simulations
5.5.1
Simulation strategies
Extensive simulations take time; a typical, moderately complex material may
require a few million floating-point complex multiplications to 20 decimal place
accuracy! It is worth learning some effective time-saving strategies, which we
suggest below.
■ Decide what is the aim of the simulation before beginning. Is it to understand

the specimen thoroughly, or to check that it is within process-control limits?
■ Match the main features of the rocking curve first, using plane wave
calculations and a single (sigma) polarisation. It should be possible to fit all
the main peaks accurately in spacing and approximately in intensity. Then
begin refinement to match the intensities and widths.
■ Add the instrument function. If this is large it affects widths and relative peak
intensities, but not integrated intensities.
■ Add curvature (if physically reasonable). This has the same effect as the
instrument function.
■ Add the other polarisation state. This affects intensities by typically 20–30%
but has little effect on widths. It doubles the time of computation.
■ Note that instrument function and curvature corrections have a greater effect
on narrow peaks (e.g. the substrate) than on broad peaks.
■ Do not use too fine a step at the beginning. The computation of the plane
wave rocking curve is linear with the number of steps, and with the number of
layers. The correlation with the beam conditioner is independent of the
number of layers but goes as the square of the number of steps or, if an
integral transform method is used, with the logarithm of the number of steps.
■ Note that a step that is comparable in size to the fringe spacing gives spurious
periodicities—the aliasing effect, well known in control theory (Nyquist
theorem)—and the final simulation must be performed with a fine step.
5.5.2
An example simulation: deduction of material parameters
For a general review of the use of simulation in high resolution diffraction see
Tanner. 12 We now give an example simulation, showing the intermediate stages
and our thought processes. We use a sequence of procedures which will improve
the rate of convergence. The underlying philosophy is that, first, the features must
be identified and matched and, only secondly, the parameters must be refined.
■ Use all the information at hand from the crystal grower
This may seem elementary, but it is surprising how often one is asked to fit an X-
ray rocking curve for a specimen of which almost nothing is known about the
structure. Thus the first rule is to ask the crystal growers what structure they
think they have grown. The second rule is not to believe them! Reliability in the
answer can be assessed on several levels. It is unlikely that the grower has
muddled InGaAs with AlGaAs and one can usually be confident in the sequence
of compounds deposited. Exact compositions can be variable, but if other data
such as photoluminescence or RHEED flux calibration are available, the
reliability of your own crystal grower can be assessed over a period of time.
Similarly, approximate thickness of various layers can be determined from other
data, and always provides a starting point for trial simulations. Never, however,
accept any assertion from a crystal grower as absolute truth. There is one
splendid example of a commercial MBE kit on which the computer cannot count
correctly the number of layers in a multilayer! It was some time before the
computer screen was watched throughout the run and the evidence of the
diffraction data accepted. In addition, interrogate your grower as to whether there
are arbitrary capping or buffer layers present. These can lead to complex
interference phenomena and the same degree of control over the deposition is
desirable as over active layers. By taking care to establish a good working
relationship with the crystal grower, a tight initial specification can be obtained.
This will steer clear of many local minima and dramatically shorten time spent in
simulation.
■ Initial measurements from the experimental data
In Chapter 3 we discussed the extraction of data directly from the rocking

curves. We will now apply these ideas and equations to help obtain our trial
structure.
■ Identification and measurement of peaks
Before attempting a first trial solution, it is essential to identify the Bragg peaks
present in the rocking curve. A number of criteria may be applied.
(a) The narrowest peak is almost always that associated with the substrate. This
rule works well for compound semiconductors such as AlGaAs on GaAs,
CdHgTe on CdTe and InGaAsP on InP where the perfect crystal reflecting
ranges are approximately the same for semi-infinite crystals of both
substrate and layer materials. Where the rule will break down is for the
situation where the substrate has a smaller structure factor than the layer
material. For example, for a 1 μ m thick layer of Si on GaAs, it is
theoretically possible to have layer rocking curve halfwidth of 5 arcsec
whereas the substrate halfwidth is 12 arcsec for plane wave CuK λ radiation
in the 004 reflection. In practice, the Si will have relaxed by the creation of a
large number of dislocations and the rocking curve width will have been
broadened by the strains. (Further, it is unlikely that people will wish to
grow Si on GaAs rather than vice versa, but the warning is there. Maybe
diamond on silicon is a possibility which might arise in the future.)
(b) Look at the lattice parameters to see whether you expect larger or smaller
Bragg angles for layer or substrate. Except when covering a heavily relaxed
strained layer, 13 AlGaAs peaks are always at smaller Bragg angles than
GaAs, as the AlAs lattice parameter is larger than that of GaAs. Where the
lattice parameters of the binary components straddle that of the substrate, for
example InAs and GaAs with respect to InP, there is no way of knowing a
priori whether the substrate or layer peak will be at the higher angle.
Measure the integrated intensity under the peaks. (Most plotting packages
will do this if your data collection software does not.) Using Figure 3.9,
make an estimate of the layer thickness from the ratio of the integrated
intensities. If you regularly use another epitaxial system, it would be advisable
to produce the equivalent figure by a series of simulations.
(c) Measure the peak separation of the various peaks and tabulate them. Use the
equations of Chapter 3 to deduce obvious layer compositions, with the
caveat that some peaks may be produced by strong interference effects
rather than by actual layers.
(d) If the structure is a superlattice, identify the equi-spaced superlattice
satellite peaks and measure their separation. If possible, determine the zero-
order peak in the series of satellites. This correponds to the Bragg angle for
the mean composition of the multilayer. If the zero-order peak can be
identified, one further piece of data is obtained for the trial simulation. In the
case of a binary superlattice, for example AlAs/GaAs, the position of the
zero-order peak will give the well to barrier ratio. Together with the period,
determined from the superlattice peak spacing, the thickness of both
components is then known. For a ternary/binary system such as AlGaAs/
GaAs, the thickness and composition cannot be determined uniquely.
Superlattice simulation is discussed in more detail in Chapter 6.
■ Identification of fringe periods
A second feature which should be measured prior to a trial solution is the period
of any fringe system seen in the rocking curve. This can be measured manually
from the separation of adjacent peaks or by Fourier analysis of suitably treated
data. If this latter course is taken, the Bragg peaks must be removed from the
data by truncation of the angular range and the overall shape of the diffraction
profile fitted and then subtracted to leave the periodic variation about a
background of zero. Finally, autocorrelation of the data dramatically reduces the
noise and sharpens up the Fourier spectrum. Procedures for undertaking such an
analysis may be found in Hudson et al . 14 Except for the case where beating
can be seen in the fringes, suggesting the presence of two closely spaced periods,
this elaborate procedure is not normally necessary to a trial simulation. The
equations for fringe spacing are given in section 3.4.8
5.5.3
Example 1: single-layer structure
We add the above deductions to the information from the crystal grower to
obtain our starting trial structure. In this section, we use the above strategy to
make a first attempt at simulation of the rocking curve shown in Figure 5.3,
which we believe to be of a single layer of AlGaAs on GaAs.
We note:
(a) that the sharp peak on the right will be the GaAs peak,
(b) that the layer peak is separated by 120 seconds from the layer, thus
suggesting a composition of Al0.31Ga0.69As if the layer is totally strained,
(c) that the ratio of the integrated intensities is layer: substrate=0.76:1,
suggesting a layer thickness of l.5 μ m,
(d) the interference period is 12.5 arc seconds also suggesting a layer thickness
of 1.5 μ m. (If these disagreed we would expect the fringe spacing to be a more
reliable indicator of layer thickness.)
Figure 5.4 shows the fit between experiment and simulation on this first trial.
■ Parameter adjustment
The fit in Figure 5.4 is not quite satisfactory, in that the position of the peak is
displaced by 4 arc seconds in the simulation and the experiment. This is not
surprising, as we assumed a Poisson ratio of 0.33 in estimating the composition.
The fringe period is also too large, by about 10%. We should not be too worried
by minor discrepancies such as these. Although the Poisson ratio of GaAs is well
known at 0.281, that of AlAs is still in dispute and it is advisable to check the
database of any commercial program to see what value is there (see Tanner et al .,
15 Goorsky et al . 16 and Bocchi et al .). 17
From the peak splitting, we scale the composition in the ratio of 120/124
seconds and change the layer thickness to 1.6 μ m. As seen in Figure 5.5, with
this adjustment we get a good fit to both peak and fringe spacings.
The main difference between the experiment and simulation is now the broad
background, peaking at the substrate Bragg angle. This is likely caused by
diffuse scatter from dislocations and other defects in the substrate (and maybe in
the layer too). We can model the diffuse scatter quite well by a Lorentzian
function of the form given in equation (5.17). In this example, use of k=2.2×103
and A=1.2×109, with ■ λ in arc seconds, gives a good match to the experimental
data.
5.5.4
A second example: a HEMT structure
As a second example, we take the rocking curve from a high electron mobility
transistor structure and show exactly all the stages in the simulation sequence.
All rocking curves that were simulated when we first attempted the modelling
are shown; this is a ‘real-time’ example!
The structure consists of a thin In x Ga1− x As layer grown epitaxially onto
GaAs and covered with a GaAs cap layer. We know that the cap layer is about
1000 Å thick and that the In x Ga1− x As layer is about 20Å thick. The
experimental rocking curve was taken with CuK λ 1 radiation in the symmetric
004 reflection, so we can immediately estimate the In composition assuming that
Figure 5.4 The results of the first simulation, with parameters extracted from the rocking
curve
the In x Ga1− x As is totally strained and unrelaxed. This is done simply from the
relative position of the broad left-hand peak in Figure 5.6(a) with respect to the
sharp substrate peak on the right. From
(5.20)
and as ∆θ=9600 arcsec and θ=33°, we deduce an effective mismatch m *=0.
1432.Assuming a Poisson ratio of 0.333, and thus m *=2m, the real mismatch m
is 0.0716.With a(InAs)=6.0584Å and a(GaAs)=5.65375Å, we arrive immediately
at a trialvalue of x=0.13.
The high frequency fringe period corresponds to the combined thickness of
both layers, or possibly the thickness of the capping layer. Assuming the latter,
and measuring the fringe period ■ λ p to be 120 arcsec, gives from
(5.21)
t=1580Å for the GaAs cap layer. The ‘peak’ at −946 arcsec from the substrate
peak is in fact an interference fringe corresponding to the thickness of the InGaAs
layer. Using the same formula, we deduce a starting thickness of 200Å for the
InGaAs layer.
The simulated curve in Figure 5.6(a) shows the result of using these
parameters in the simulation. The simulated data have been normalised to
150000 cps in the substrate peak, to allow for detector non-linearity. It is
immediately clear that the layer peak is not in the correct position, and this arises
Figure 5.5 Comparison of rocking curve (small squares) with second trial simulation
(dashed line). The effect of adding diffuse scatter is shown by the solid line
from the fact that in deducing the starting parameters we assumed a Poisson ratio
of 0.333. In the simulation, a correct value of the Poisson ratio has been used. The
x value is then adjusted by the ratio of the simulated and experimental peak
splittings, i.e. to (2526/2267)×0.13= 0.145. Figure 5.6(b) shows the result of
using this value, and we see immediately that the peak positions now agree very
well. They differ by 45 arcsec and Figure 5.6(c) shows a similar small adjustment
in the same manner to x=0.1425. Excellent agreement is now achieved.
However, the splitting of the interference peak at −946 arcsec is too small in
the simulation by 140 arcsec, suggesting that the InGaAs layer is too thick in the
simulation. Figure 5.6(d) shows the result of decreasing the InGaAs thickness to
170Å and increasing the GaAs cap thickness to 1410Å to compensate. This was
clearly an error of logic as the effect was to move the interference peak closer to
the substrate peak! Figure 5.6(e) shows the result of increasing the InGaAs
thickness to 220Å and leaving the total thickness at 1580Å, as the short period
interference fringe period is very well matched. This is much better!
The result is now sufficiently good to perform a convolution with the
reference crystal. Note that the previous simulations were done under plane wave
conditions, reducing the computing time significantly. The result shows no
significant difference in the quality of the fit. What is now necessary is to adjust
the InGaAs layer thickness to get the short period fringe position correct.
Dropping the thickness to 215Å shows that this is clearly the wrong way. A final
alteration to 225Å in Figure 5.6(f) gives almost perfect agreement.
Figure 5.6 The HEMT structure and (a)–(f) the sequence of simulations as described in
the text
Addition of a Lorentzian diffuse scatter term of the form 1.5×107/(■ λ 2
+3×105) gives excellent agreement, as seen in Figure 5.6(f). Thus it has taken a
total of only eight simulations to get a very good match to the data.
Figure 5.6 (cont.)

5.6
Summary
We have seen that the full dynamical theory may be solved conveniently for
simple or layered materials that are uniform in the plane of the layer. The
approximations are small, and accuracies of 2−5% are normal. Sub-nanometre
sensitivity to layer thickness may be achieved in some cases, and simulation is
the most powerful tool for detailed characterisation. An intelligent approach to
Figure 5.6 (cont.)
simulation, using all available information, is shown to give rapid convergence to

a satisfactory model.
References
. S.TAKAGI, Acta Cryst, 15 , 1311 (1962).

. S.TAKAGI, J. Phys. Soc. Japan, 26 , 1239 (1969).
. D.TAUPIN, Bull. Soc. Fr. Mineral. Cristallogr., 87 , 469 (1964).
. M.A.G.HALLIWELL, J.JULER & A.G.NORMAN, Microscopy of semiconducting
materials, Inst. Phys. Conf. Ser., 67 , 365 (1983).
. Z.G.PINSKER, Dynamical scattering of X-rays (Springer-Verlag, Berlin, 1978).
. R.ZAUS, J.Appl. Cryst., 26 , 801 (1993).
. L.D.LANDAU & E.M.LIFSHITZ, Elasticity (Pergamon, Oxford, 1972), pp. 13, 14, 55 .
. L.VEGARD, Z.Phys., 5 , 17 (1921).
. A.Y.NIKULIN, A.W.STEVENSON & H.HASHIZUME, Phys. Rev., B53 , 8277
(1996).
. L.TAPFER & K.PLOOG, Phys. Rev. B, 40 , 9802 (1989).
Connecticut, 1982) p. 308 .
. B.K.TANNER, J.Electrochemical Society, 136 , 3438 (1989).
. B.K.TANNER, J.Phys. D.: Appl. Phys., 26 , A151 (1993).
. J.M.HUDSON, B.K.TANNER & R.BLUNT, Adv. X-ray Analysis, 37 , 135 (1994).
. B.K.TANNER, A.G.TURNBULL, C.R.STANLEY, A.H.KEAN & M.McELHINNEY,
Appl. Phys. Lett., 59 , 2272 (1991).
. M.GOORSKY, T.F.KEUCH, M.A.TISCHLER & R.M.POTEMSKI, Appl. Phys. Lett.,
59 , 2269 (1991).
. C.BOCCHI, C.FERRARI, P.FRANZOSI, A.BOSACCHI & S.FRANCHI, J.Crystal
Growth, 132 , 427 (1993).
6
Analysis of Thin Films and Multiple Layers
In this chapter we discuss the double-axis rocking curves of thin single

layers, the effects of thin interfacial layers and the Bragg case
interferometer. We discuss characteristic rocking curves of
superlattices and extraction of parameters, and show how the rocking
curve builds up as more layers are added. We give strategies for
analysis and recommend a starting point for simulations.
6.1
Review of rocking curves from thick layers
When the epitaxial layer thickness is quite high, typically of the order of one
micrometre, we can apply the simple criteria discussed in Chapter 3 to determine
the layer parameters from the rocking curve. The effective mismatch can be
determined by direct measurement of the angular splitting of the substrate and
layer peaks and the differential of the Bragg law. This simple analysis cannot be
applied when the layer becomes thin, typically less than about 0.25 μ m, where,
even for a single layer, interference effects become extremely important. We
consider these issues in section 6.2 below.
If a thick layer can be considered perfect, the integrated intensity can be used
as a measure of the layer thickness, as described in Chapter 3. However, if the
layer is thick and imperfect, the dynamical theory may not be appropriate. Then
the integrated intensity is larger than that predicted and an overestimate will be
made of the layer thickness. The criterion for a layer to behave as a perfect
crystal is essentially that it must not contain inhomogeneous strains which give
rise to effective misorientations greater than the perfect crystal reflecting range.
Such tilts or dilations will lead to broadening of the rocking curve. An example of
such a broadened rocking curve from a CdTe epitaxial layer grown on GaAs is
shown in Figure 6.1. The simulated rocking curve from an ideally perfect layer
has a rocking curve width of the order of 10 arc seconds, whereas the
experimental curve is some hundreds of arc seconds in width. One approach is to
assume that it is an ideally imperfect crystal consisting of many thin layers
misoriented with respect to each other. The rocking curve can then be built up by
Figure 6.1 Double-axis rocking curve of CdTe on GaAs showing the broadening due to
the very high dislocation density in the layer. (Courtesy R.I.Port, Durham University.)
The rocking curve width in such thick, high mismatched layers falls with increasing layer
thickness 1 , 2
adding the intensities of such layers convolved over the misorientation

distribution function, in this case a Gaussian distribution. We may then expect
that the integrated intensity will vary linearly with thickness and by use of the
calibration table for the relative intensity of layer to substrate peak at very low
layer thickness (Figure 6.2), we can deduce the thickness. It is difficult to specify
the exact amount of broadening of the rocking curve necessary for this
kinematical model to be appropriate as this is a function of the strain distribution
through the layer.
In this circumstance, it is more reliable to measure the absolute integrated
intensity of the substrate peak and compare this with the integrated intensity from
an equivalent crystal of the substrate material on which no layer has been grown.
In the angular position for diffraction from the substrate, the layer will not
diffract and the substrate peak intensity will be simply reduced by normal
photoelectric absorption. For a symmetric reflection, it is easy to see that the
integrated intensity I s of the substrate peak with the layer of thickness t present
is related to the integrated intensity of the bare substrate I 0 by
I s =I 0 exp[−μ t/(2sin λ B)] (6.1)
where μ is the linear absorption coefficient and λ B is the Bragg angle. Provided
that the layer composition is known, μ can be calculated. Note therefore that for
ternary alloys, the Bragg peak splitting must be used to determine the
ANALYSIS OF THIN FILMS AND MULTIPLE LAYERS 135
Figure 6.2 Ratio of integrated intensity of layer peak to substrate peak as a function of
thickness for a thin AIGaAs layer on GaAs
composition and hence μ . When the reflection is not symmetric, and the Bragg
planes make an angle λ with the surface, equation (6.1) is generalised to
I s =I 0 exp[−μ t/{sin(λ B+λ )+sin(λ B−λ )}] (6.2)
due to the different path lengths of the incident and diffracted beams. This
approach is valid independently of the layer perfection.
6.2
Thin single layers
We have already seen something of the complexity and subtlety of the rocking
curves which can arise when there are several thin layers present. This can be
understood in terms of contributions from individual thin layers and thus it is
important to have a clear picture of the effects of thin single layers.
The first point is one of simplicity, as the integrated intensity varies linearly
with thickness when the layer is sufficiently thin to be in the kinematical
diffraction limit. This is again valid irrespective of whether the layer is imperfect
or perfect. It may be more convenient to measure the ratio of layer to substrate
peak, which also varies linearly with thickness provided that the layer remains
sufficiently thin for absorption to be neglected (Figure 6.2).
As discussed in Chapter 3, the interference (thickness) fringes can be used as a
very direct method of layer thickness measurement. We note that the period is
Figure 6.3 Rocking curve for a layer of ‘germanium’ with lattice parameter 5.43 Å,
exactly lattice-matched to a silicon substrate. Interference fringes are seen
related to thickness (equation (3.32)) only by geometrical parameters and does

not involve the structure factor. Thus, we might expect that for a layer of
identical composition to the substrate, but with a different lattice parameter, the
interference fringes will be present. This effect is seen very clearly in rocking
curves of doped and ion implanted layers. What is not quite so obvious is that the
interference fringes will be seen when the layers are of different composition but
exactly lattice-matched. This effect is shown in the simulation of Figure 6.3,
which is for the rocking curve from a layer of ‘germanium’ with a lattice
parameter 5.43 Å on a silicon substrate. Although the example is somewhat
artificial, the situation does occur in ternary alloys which are exactly lattice-
matched to the substrate.
When the epitaxial layer is thin, typically below 0.25 μ m, the differentiated
Bragg law can no longer be used to determine the effective mismatch. As originally
shown by Fewster and Curling 3 and subsequently studied in detail by Wie, 4
the peak splitting for a given composition varies as a function of layer thickness.
As shown in Figure 6.4, the effect is negligible for layers of thickness greater
than 0.25 μ m, but varies very rapidly below 0.1 μ m. The shift is dependent on the
mismatch and is not related in any way to relaxation. It arises as a result of the
interference terms in the expression for the intensity scattered from the thin
layer, but a direct physical explanation has so far eluded the present authors. The
effect is contained within the dynamical theory formalism of the rocking curve
simulation programs based on the solution of the Takagi-Taupin equations, the
Figure 6.4 Variation in the substrate-epilayer peak splitting as a function of thickness for
Al x Ga1− x As layers of constant composition (x=0.5 and 1) on GaAs. Note that the absolute
separation decreases with decreasing thickness. CuK λ 004 reflection
data in Figure 6.4 being the peak splitting as calculated using the Bede Scientific
RADS program. 5 Thus, in order to determine the mismatch in a thin layer, it is
essential to use the simulation technique, or at the very least a pre-computed
calibration curve. It is not a problem in cases where the mismatch is large, and is
then often neglected.
6.3
Graded thick layers
If the composition varies with depth through an epilayer, the rocking curve peak
from the layer is characteristically broadened. The profile can be simulated by
splitting the layer up into a sequence of thin lamellae of constant lattice
parameter, which are assumed to scatter coherently (Figure 6.5). To a good first
approximation, the scattering from the component lamellae is kinematical, and
the rocking curve can be simulated by adding the amplitudes and phases from
individual thin layers. The commercial simulation packages all permit thin layers
of varying composition to be simulated in this way and, although it is within the
dynamical theory formalism, the results are often identical to those of the
kinematical theory.
Figure 6.5 The approximation of a continuously varying lattice parameter into a series of
lamellae of constant lattice parameter
When there is a linear grade in lattice mismatch between the top and bottom of
the epilayer, the rocking curve has a characteristic wedge shape such as shown in
Figure 6.6(a). For a quadratic dependence on depth, the curve becomes strongly
asymmetric and falls more steadily to zero on one side of the peak (Figure 6.6
(b)). The interference fringes can be quite complex and are not easy to analyse
qualitatively. The period is neither associated with the thickness chosen for the
component lamellae nor characteristic of the total thickness. As seen in Figure 6.6
(c), the interference fringe spacing changes on going from a linearly
graded structure to a quadratically graded structure. Although the end points are
the same, here 0% aluminium to 100% aluminium in Al x Ga1− x As, not only
does the shape of the curve change, but so too does the interference fringe
character. These fringes are genuine and not simulation artefacts. Experimental
examples can be found in the literature.
Such rocking curves were found very commonly in the early days of
development of quaternary systems for long-haul telecommunications links, and
epilayers grown by chemical vapour epitaxy showed the effects particularly well.
6 Current III–V layers do not often show the effects, but this is not true for II–VI
compounds containing mercury, such as the important infrared detector material

Hg x Cd1− x Te. Figure 6.7 shows an example of a rocking curve from a thick
layer of another mercury compound, Mn x Hg1− x Te, grown by metalorganic
vapour phase epitaxy on a CdTe buffer layer on GaAs. The simulated rocking
curve assumes that the layer is divided into 20 lamellae, each differing in the Hg
concentration by x=0.05. Such Hg segregation is also found in HgxCd1− x Te
devices used for infrared sensors.
There are examples in the literature where there have been random
fluctuations in the composition through the depth. Such curves are almost
impossible to simulate in detail, but typical characteristics have been simulated
by Lyons. 8
6.4
Bragg case interferometers
In Chapter 5, we have already seen an example of the rocking curve from a high
electron mobility transistor structure, where a thin, highly mismatched layer is
situated between two identical layers. Figure 6.8 shows another such rocking
curve, this time for an Al0 . 6Ga0 . 4As layer between two Al0 . 3Ga0 . 7As layers
on a GaAs substrate. Although the peak from the thin Al0 . 6Ga0 . 4As layer is itself
of low intensity, the presence of this layer has a dramatic impact on the two Al0 .
3Ga0 . 7As layers. The latter peak is modulated very strongly and almost appears
to be split. As the sandwiched layer thickness is changed, the position of the
peaks changes, altering the relative heights of the main and subsidiary peaks from
the Al0 . 3Ga0 . 7As layers. This example is a particular case of a general class of
epilayer systems which behave as Bragg case interferometers.
The Laue (transmission) case X-ray interferometer was invented by and
characterised in detail by Bonse and Hart in the mid-1960s. 9 Recent
applications have been reviewed in detail by one of the present authors in the
1996 Erice school proceed-ings. 10 The realisation that the ‘ABA’ epilayer
structure behaved like a Bragg case X-ray interferometer led to a very direct
physical picture of the origin of the interference fringes in the rocking curves. 11
In Holloway’s analysis, he pointed out that if the B layer had the same lattice
parameter as the A layers, then provided the B layer was thin, the rocking curve
profile would be independent of the B layer thickness as it simply inserted an
integer number of Bragg planes into the structure. The phase difference between
waves scattered from the top and bottom A type layers was then only changed by
a multiple of 2λ , leaving the rocking curve unaltered. If there was a mismatch
m* between A and B type layers, then the phase was altered by an additional
amount given by 2λ m*t/d where t is the B layer thickness and d is the Bragg
plane spacing. For a layer of thickness d / m* , the interference fringes in the
rocking curve fringes move along the angular axis by one period. The rocking
curves for layers whose thickness increases by integer multiple of d / m* are, to
first order, identical.
Thus the sensitivity of the rocking curve to the presence of thin buried layers
can be very high if the mismatch is high. Tapfer and Ploog, 12 Green et al . 13
and more recently Li et al . 14 have used this phenomenon to deduce the
presence of a thin interfacial layer with a high mismatch. For high mismatch
(e.g. In0 . 2Ga0 . 8As and GaAs) the sensitivity is at a single atomic layer. Note,
however, that if the mismatch is low as in Figure 6.8, the thickness sensitivity is
low and changes in thickness of several hundred angstroms may be necessary to
observe effects.
Figure 6.6 Simulated rocking curves of InxGa1− x As on InP for: (a) a linearly varying
composition with distance, x=0.0067t+0.525 where t is the distance in micrometres from
the substrate interface; (b) a quadratic dependence on distance from the substrate, x=−0.
022t 2+0.0133t+0.525, total layer thickness 3 μ m, 004 reflection, CuK λ radiation, InP
reference crystal; (c) 2 μ m thick Al x Ga1− x As layer on GaAs varying linearly and
Figure 6.7 Experimental and simulated rocking curves of an Mn x Hg1− x Te grown on a

CdTe buffer layer on GaAs. Slit collimated λ –2λ scan. (Courtesy C.D.Moore) 7
6.5
Superlattices
Modern high-speed and optoelectronic devices regularly contain periodic
sequences of layers of alternating composition. By use of such a period
superlattice of small repeat layer thickness, the band structure of semiconductors
can be dramatically altered and such multiquantum wells can exhibit a wide
range of novel phenomena. As the effective bandgaps can be tailored by varying
the thickness of the quantum well, this provides a very flexible design tool for
the electronic engineer. Additionally, it has been found that superlattice
structures can be used to prevent mismatch dislocations propagating, thereby
providing a means of eliminating relaxation, which would lead to the presence of
defects which reduce unacceptably the lifetime of light-emitting devices.
The rocking curves from such one-dimensional ‘artificial crystals’ are quite
striking and a typical example is shown in Figure 6.9, where we see a sequence
of equispaced satellite peaks associated with the superlattice of AlGaAs and
GaAs. We can understand the origin of the superlattice satellites by noting that
the artificial periodic structure will have a corresponding set of points in
reciprocal space. The period in real space is large and therefore the separation of
reciprocal lattice points in reciprocal space is very small. Additionally, it is in
one dimension only and thus we have a line of reciprocal lattice points such as
shown in Figure 6.10. The important point to note is that all reciprocal lattice
points in the basic crystal, including the 000 point (which corresponds to grazing
incidence reflectivity measurements) are modulated by the superlattice. The
Figure 6.8 Rocking curve from a layer of Al0.6Ga0.4As sandwiched between two 1 μ m
Al0.3Ga0.7As layers on a GaAs substrate. Note the interference fringes and the change in
phase of the fringes with respect to the layer peak as the AlGaAs layer thickness is
changed. Al0.6Ga0.4As layer thickness (a) 0.2 μ m, (b) 0.3 μ m
origin of the satellites then becomes clear and in triple-axis reciprocal space
maps (Chapter 7) the reciprocal space structure is clearly seen.
Figure 6.9 Double-axis rocking curve from an AlGaAs/GaAs superlattice CuK λ

radiation, 004 reflection
Superlattice structures containing only a few repeats can be difficult to analyse
and result in some quite unexpected effects. Consider for example an epilayer of
total thickness 1 μ m which is composed of two layers L and M, each 0.5 μ m
thickness and mismatched from the substrate by −600 and −1200 ppm
respectively. The rocking curve (Figure 6.11) is very clearly composed of two
layer peaks, one corresponding to each layer. However, if the total thickness is
held the same, but the epilayer divided up into four layers each of 0.25 μ m
thickness, then we have six peaks. Although the two strongest peaks are close,
there is no peak which corresponds to the composition! Differentiation of the
Bragg equation would suggest that there are six layers present, each of different
composition. Further subdivision results in five peaks for six layers and, on
division into ten layers of thickness 0.1 μ m, we find a main peak appearing at the
average composition of the structure. This is the so-called zero order peak.
The positions of the subsidiary peaks can be recognised as being determined
by the thickness of the layers, or in another form, the superlattice period. These are
the satellite peaks which we described earlier. It is best seen in a similar example
to Figure 6.12, but with a larger mismatch between layers and plotted on a
logarithmic scale. Here we show the simulated curves for an In x Ga1− x As
structure on InP, of total epilayer thickness of 1 μ m. The InxGa1− x As is
composed firstly of two layers of composition x=0.5 and x=0.43, each of
thickness 0.5 μ m. On subdivision into four alternating layers of composition x=0.
5, 0.43, 0.5, 0.43, each of thickness 0.25 μ m, we see that the layer peaks are
strongly modulated by the thickness fringes. Note that this modulation splits the
Figure 6.10 (a) Real-space superlattice structure and (b) the corresponding structure in
reciprocal space
Figure 6.11 Simulated rocking curves from an epilayer of total layer thickness 1 μ m,
subdivided into 2, 4, 6 and 10 layers of alternate composition. The curves are displaced
for clarity
peaks, leaving a minimum at the position corresponding to the differentiation of

Bragg’s law. Also note from the middle curve in Figure 6.12 that alternate
fringes have been enhanced in intensity. Subdivision into eight layers (four repeats
of x=0.5 and x=0.43 layers) results in the appearance of clear satellite peaks, the
separation of which is determined by the sum of the pair of x=0.5 and x=0.43
layers. While this is confusing on a linear scale, examination of the data on a
logarithmic scale makes the origin of the individual features clear.
Figure 6.12 Simulated rocking curve from sequence of layers of total thickness 1 μ m,
divided into 2, 4 and 8 repeats of In x Ga1− x As layers of composition x=0.5 and x=0.43.
The curves are displaced for clarity
It is worth noting that for superlattices where the two components differ only
by the substitution on one of the zincblende structure sites (e.g. Al x Ga1− x As/
GaAs) it may be advantageous to use quasi-forbidden reflections such as 002. In
such reflections, the intensity is determined by the difference in the atomic
scattering factor of the atoms on the two zincblende sites, for example |ƒGa−ƒAs|
for the case of GaAs. For Al x Ga1− x As, the intensity becomes |(1−x)ƒGa+xƒAl
−ƒAs|. There is thus a large change in scattering factor between the two
components and excellent contrast of the superlattice satellites is obtained.
However, the penalty is that the overall intensity is down and, depending on the
signal-to-noise ratio inherent in the experimental arrangement, it may be better to
use the lower contrast but higher signal of the allowed 004 reflection rather than
the higher contrast but lower signal of the quasi-forbidden 002 reflection.
6.6
General characteristics of large repeat superlattices
Device engineers and crystal growers are concerned with determining the
following parameters of a superlattice structure:
■ the spatial period of the structure

■ the thickness of the repeating unit

■ the compositions of the layers
■ the dispersion in the repeating period
■ the interface roughness
■ the interface grading
These parameters can be found from the rocking curve, with the exception that
roughness cannot at present be distinguished from grading if only double-axis
diffraction is used.
Although the rocking curve from a superlattice or multiquantum well structure
may be quite complex, there are a number of common features which can be
used for the analysis. Assuming that we have a substrate of material A (e.g.
GaAs) and a superlattice or MQW with a stack of AB layers, where B is the alloy
(e.g. Ga1− x Al x As), as illustrated in Figure 6.9, the rocking curve will show the
following features:
1 A substrate peak from the A substrate.

2 A peak caused by the addition of Bragg reflections from the A and B
components of the MQW. This is the zero-order or average mismatch peak,
from which the average composition of the A+B layers may be obtained by
differentiation of Bragg’s law.
3 A set of subsidiary ‘satellite’ peaks symmetrically surrounding the zero-
order peak, with spacing determined by the periodicity (total thickness of the
repeating layers) of the MQW. These may also be regarded as the sum of the
interference or gap fringes arising from interference between each of the
layers comprising the MQW.
These peaks may be analysed to give much of the information required by the
crystal grower. The methods of analysis have been developed by Segmuller et al .,
15 Kervarac et al . 16 and by Fewster 17 , 18 whose treatment we use here.
First we shall give a qualitative treatment of the effects of multilayers in

general upon diffraction. As we have noted above, the term ‘superlattice’ may be
taken literally—we have a large-spacing lattice with the periodicity of the
spacing of the units making up the superlattice. If this is 200 Å, a typical value,
we could use this as a‘Bragg’ plane, apply the Bragg law and with CuK λ
radiation there should be a first-order Bragg peak at λ B=0.221°. Such peaks can
indeed be seen with grazing-incidence reflectometry.
The number of units in an MQW will be much more limited than the number
of atomic planes sampled by the X-ray beam in a standard reflection. The
intensity will be low, but also the MQW will behave as a ‘thin crystal’—the
reciprocal lattice points will be extended into rods perpendicular to the crystal
surface. This will broaden the reflection, and thus the width of each satellite peak
is determined by the number of units in the MQW. It has even been possible, by
careful analysis, to count the number of units more accurately than the prediction
by the crystal growers.
The number of satellite peaks will depend on the shape of the interface
between the units. It is convenient to think of the diffraction pattern in the
kinematic approximation as the Fourier transform of the structure. If the layers in
the units were graded so that the overall structure factor variation were
sinusoidal, this would have only one Fourier component and thus only one pair
of satellites. If the interface is abrupt, this is equivalent to the Fourier transform
of a square wave, which consists of an infinite number of odd harmonics; the
corresponding diffraction pattern is also an infinite number of odd satellites. The
intensities of the satellites therefore contain information about the interface
sharpness and grading.
If there are variations in the period (period dispersion) then there are in effect
different periods within the superlattice. The zero-order peak for each of these
must be the same—because it is simply the average mismatch of the alloy—but
the distribution of intensity in the various interference fringes will be slightly
different. This will tend to affect the higher-order satellite peaks more than the
lower orders, and if measured with a low resolution instrument there will appear
to be an increase in the width of the satellite peaks with order of the satellite, having
taken out any instrumental functions.
6.6.1
Average mismatch
Subject to the caveat that there can be a significant shift in peak position when the
total layer thickness is sub-micrometre we can determine the average
composition of the MQW using the zeroth order, or average mismatch, peak.
Asymmetric reflections are often used, both to determine any relaxation and to
enhance the diffraction from thin layers. Let the period of the superlattice in real
space be A, and the thickness of layers of A x B1− x of composition x 1 and x 2 be
D 1 and D 2 respectively. Then
■ =D 1+D 2 (6.3)
and
D 1 x 1+D 1 x 2=x av■ (6.4)
where x 1+x 2=1. Therefore
D 2=■ (x av−x 1)/(1–2x 1) (6.5a)
D 1=■ (x av−x 2)/(1−2x 2) (6.5b)
If the total A+B component of the MQW is less than about 0.25 μ m, then the
zero-order peak does not appear and dynamical-theory simulation must be used
to determine the mean composition.
6.6.2
Periodicity
For X-ray wavelength λ , let L i and L j be two diffraction orders, (e.g. 5 and 7)
and λ i and λ j the angles at which these orders diffract; then the following
relationship holds:
■ =(L i −Lj)λ /[2(sin λ i −sin λ j )] (6.6)
A then follows from the measurement of any two satellite peak positions, or,
better, the measurement and averaging of several.
One may wish to know separately the average widths of each part of the unit
(well and barrier); change in this ratio with no change in the overall period will
not affect the satellite positions, but only their intensities. In effect, we have a
large unit cell, the superlattice or MQW unit, and are varying the structure factor
within this cell. The procedure is then to calculate the structure factors of all the
observable satellites of the superlattice unit, using the normal formula to sum
over atoms in a unit cell, and compare these with the integrated intensities
actually measured after correcting for any instrumental effects. The calculation is
iterated until the difference between the calculated and measured intensities is
minimised as a function of the well to barrier ratio. The difference is expressed
as the usual R factor
R=■ {|F 0|−|F 0|}/|F 0| (6.7)
where F o and F c are the observed (integrated intensities) and calculated

structure factors.
The procedure is shown in detail in Fewster 18 , who states that it is accurate to
about 5% of a monolayer, and that it can be used to determine the
incommensurability in well widths (i.e. widths that are not an integral number of
lattice planes but formed from fractions of a monolayer).
It is worth noting that there are two particularly useful cases with sharp
interfaces:
1 If the well to barrier ratio is 1:1 then even-numbered satellites are absent. This
corresponds to the Fourier transform of a square wave, which contains only
odd harmonics.
2 If the well to barrier ratio is 2:1 or 1:2 then third (sixth, ninth, …) harmonics
are absent.
6.6.3
Thickness or period dispersion
Fewster obtains this from the increase of satellite widths as a function of satellite
order. The true integral breadths or FWHM values are first found; with a double
or multiple crystal diffractometer this is simple measurement, and with a powder
diffractometer the instrument function must first be found (by finding the zeroth-
order integral breadth and subtracting). 19
Equation (6.6) may be simplified to
■ =(L i −L j )λ /2(cos λ )(■ λ ) (6.8)
on the assumption that the satellite peaks are near the Bragg peak and that over
that range cos λ does not change appreciably. We may differentiate equation (6.8)
to obtain
■ ■ =(L i −L j )λ d(■ λ )/(cosλ )(■ λ )2 (6.9)
The meaning of d(■ λ ) is that it is the difference in the angle within the satellite—
the satellite broadening—for a period difference of ■ ■ . The FWHM of the
Gaussian distribution of periods is given by
λ (■ )=(L i −L j )λ b s /(cos λ ) (■ λ )2 (6.10)
where b s is the statistical average for all the measured satellites.
This approach does give plausible results but the physical mechanism is
unconvincing, since the individual peak widths of both satellite and intermediate
thickness fringes must be determined by the total thickness of the superlattice
structure. Superlattice reflections can be thought of as a constructive interference
between thickness fringes from successive units of the superlattice, as is easily
seen by modelling a series of superlattices with increasing numbers of repeat
units on a simulation program. Such modelling shows that the individual
reflection widths change little with the introduction of period dispersion but that
the relative intensities of different maxima in the thickness oscillations are very
sensitive to irregularities in the period. Often, a small period dispersion transfers
intensity from the superlattice peak to the immediately adjacent thickness fringes,
giving the impression of peak width increase if the resolution is not sufficient. It
is therefore by examination and simulation of the intensities of fringes in
between the satellite reflections that we may determine period dispersion.
Figure 6.13 shows such a determination, in the a Si-Ge superlattice, showing the
details in between the −2 and +1 satellites. The simulated curve with 2%
dispersion is not intended to be a perfect match, but to be clearly more irregular
and showing a greater period dispersion than the experimental curve. Note the
regularity of the fringes in the latter and the narrowness of the satellite
reflections both in the experiment and in the simulation.
6.6.4
Interface roughness and grading
Roughness and grading cannot be distinguished in a double-axis rocking curve.
Fewster’s treatment is to assume that roughness is confined to less than one
monolayer (over the coherence width of the beam) and to take up the rest of the
combined parameter with grading. The introduction of sub-monolayer roughness
permits the introduction of an incommensurate well width, found as described
above. The roughness is introduced by adding coherently the structure factors of
two commensurate periods, with well widths straddling the average value. These
Figure 6.13 (a) Experimental and simulated curves for a Si-Ge 20-period superlattice, (b)
Enlargement showing the thickness fringes between the −2 and +1 satellites. Upper curve
is a RADS simulation of the superlattice structure with 2% period dispersion introduced,
showing irregular changes in the amplitude of the thickness fringes but little change in
width of satellite or subsidiary fringes. (Courtesy Dr A.R.Powell)
are then scaled with the previously determined Gaussian dispersion of periods
(equation (6.10)) and added to give the overall structure factor for that satellite.
The structure factor of each satellite is computed and compared with
experimental intensities, again using R factors, with the variable being the
composition grading away from the interface, altered one layer at a time. Clearly
this grading may be as complex as desired given ample computing time, but a linear
grading over a few layers usually suffices. The effect of such grading can have a
very significant effect on the relative heights of successive satellites. It can, for
example, result in very substantial asymmetry in the positive and negative
satellite peak intensities. An extreme example of this is shown in Figure 6.14. An
example which has been quantified by means of simulation is shown in
Figure 6.15 which gives a series of simulated rocking curves, due to Müller, 20
in which grading of the well walls is included. It is clear that such a change in the
lattice parameter profile is necessary in order to obtain a good fit between
simulated and experimental data.
An exemplary study of superlattice rocking curves was made by Gillespie et
al . 21 on Si/GaAs. Analysis of the data revealed that an excellent fit could be
Figure 6.13 (cont.)

obtained between simulation and experiment if it was assumed that the silicon
layers were non-integer numbers of monolayers in thickness. This non-physical
situation could be resolved by assuming that there existed a monolayer at the
interfaces which had mixed occupancy of Si and Ga or As atoms. The phase shift
introduced into the X-ray waves by such an alloy layer led to an equally good fit
between simulation and experiment.
More recently Holý et al. 22 have developed a theory which treats
roughness properly and shows that true roughness results in broadening of the
double-axis superlattice satellite peaks. The broadening varies as n 2 , where n is
the satellite order. Roughness does not therefore broaden the zero-order peak.
However, the data produced by Holý et al. suggested that the broadening was
constant with satellite order and that the broadening was due to a mosaic in the
underlying crystal structure.
6.6.5
Simulation of superlattice structures
In simulating a superlattice structure, there is a need to establish initial input
parameters for a first trial. The following procedure is not exclusive, but is that
adopted by the authors.
1 Determine the superlattice period from the spacing of adjacent satellites.

Figure 6.14 Rocking curve of a superlattice showing very substantial asymmetry between
the +n and −n satellite peaks. Alternate +n satellites are absent
2 Estimate the ratio of the well-to-barrier from the relative integrated
intensities of successive satellites. Recall that the satellites correspond to a
Fourier transform of the periodic structure. This done, we have input
parameters for both well and barrier thickness.
3 Determine the average composition x av from the position of the zero-order
peak.
4 Use these parameters to estimate the individual compositions x 1 and x 2 of
the superlattice using equations (6.5).
5 Input these parameters into a simulation program and begin iteration to a
best fit.
An example of the success of such fitting procedures is given in Figure 6.16,

which shows experimental and simulated rocking curves of a Cd1− x Mn x Te/
CdTe superlattice grown by molecular beam epitaxy on InSb. Not only is the
large-scale structure very well modelled but, subject to inclusion of a thin
interface layer of In2Te3, there is an excellent agreement in the detailed fringe
structure around the satellites. 23
6.7
Summary
Thin epitaxial layers display a rich variety of X-ray optical phenomena which
can be exploited for materials characterisation. Superlattice structures in
Figure 6.15 Experimental and simulated rocking curves with differing lattice parameter
profiles through the well-barrier wall. (Courtesy R.Müller, University of Munich)
particular display rocking curves which are sensitive to interface strains and
these provide unique tools for investigating the structure-property relationships
in an important class of electronic materials.
References
. J.E.AYERS, J.Crystal Growth, 78 , 3724 (1995).

. K.DUROSE & H.TATSUOKA, Inst. Phys. Conf. Ser., 134 , 581 (1993).
. P.F.FEWSTER & C.J.CURLING, J.Appl.Phys., 62 , 4154 (1987).
Figure 6.16 Experimental and simulated rocking curves of a Cd1− x Mn x Te/CdTe

superlattice grown on InSb The lower figures show enlargements of the region around the
−1 and +1 satellites respectively. 23 (Courtesy Dr C.R.Li)
. C.R.WIE, J.Appl.Phys., 66 , 985 (1989).

. D.K.BOWEN, N.LOXLEY, B.K.TANNER, L.COOKE & M.A.CAPANO, Mater. Res.
Soc. Symp. Proc., 208 , 113 (1991).
. M.J.HILL, B.K.TANNER, M.A.G.HaLLIWELL & M.H.LYONS, J.Appl.Cryst., 18 ,
446 (1985).
. C.D.MOORE, PhD thesis, Durham University (1997).
. M.H.LYONS, J. Crystal Growth, 96 , 339 (1989).
. U.BONSE & M.HART, Appl. Phys. Lett., 6 , 155 (1965).
. D.K.BOWEN, in: X-ray and neutron dynamical diffraction: theory and applications, eds.
A.AUTHIER, S.LAGOMARSINO & B.K.TANNER (Plenum Press, New York,
1996) p. 381 .
. H.HOLLOWAY, J. Appl. Phys., 67 , 6229 (1990).
. L.TAPFER & K.PLOOG, Phys. Rev. Phys. Rev. B, 40 , 9802 (1989); L.TAPFER,
M.OSPELT & H.VON KANEL, J. Appl. Phys., 67 , 1298 (1990).
. G.S.GREEN, B.K.TANNER & P.KIGHTLEY, Mater. Res. Soc. Symp. Proc., 208 , 315
(1991) G.S.GREEN, B.K.TANNER, S.J.BARNETT, M.EMENY, A.D.PITT, C.R.
WHITEHOUSE & G.F.CLARK, Phil. Mag. Letts., 62 , 131 (1990).
. C.R.LI, P.MöCK, B.K.TANNER, D.ASHENFORD, J.H.C.HOGG & B.LUNN, Il
Nuovo Cimento D 19 , 447 (1997).
. A.SEGMULLER, P.KRISHNA & L.ESAKI, J. Appl.Cryst, 10 , 1 (1977).
. J.KERVARAC, M.BAUDET, J.CAULET, P.AUVRAY, Y.Y.EMERY &
A.REGRENY, J. Appl. Cryst, 17 , 196 (1984).
. P.F.FEWSTER, Philips J. Research, 41 , 268 (1986).
. P.F.FEWSTER, Semicond. Sci. Tech., 8 , 1915 (1993).
. P.F.FEWSTER, Repts. Prog. Phys., 59 , 1339 (1996).
. R.MÜLLER, University of Munich, private communication.
. H.J.GILLESPIE, J.K.WADE, G.E.CROOK & R.J.MATYI, J. Appl. Phys., 73 , 95
(1992).
. V.HOLY, J.KUBENA, E.ABRAMOF, A.PRESEK & E.KOPPENSTEINER, J. Phys D:
Appl. Phys., 26 , A146 (1993).
. C.R.LI, B.K.TANNER, D.ASHENFORD, C.H.J.HOGG & B.LUNN, J. Appl. Phys., 82 ,
2281 (1997).
156
7
Triple-axis X-ray Diffractometry
In this chapter we show how restricting the angular acceptance of the

detector adds another dimension to the information available from
high resolution diffraction techniques, enabling strains and tilts in a
sample to be identified separately. We explain the reciprocal space
representation of triple-axis diffraction maps, and give several
examples of its application to materials characterisation.
7.1
Introduction
A double-axis system, with or without a monochromator, uses an open detector
and therefore integrates the scattering from the specimen over all angles within
its aperture. Whilst this is relatively quick and convenient, it loses information;
in particular, the scattering from bent or mosaic crystals occurs at different
settings of the specimen crystal for a given d spacing, and details such as
thickness fringes or narrow peaks can be lost or blurred. In some cases, such as
Figure 7.1, the width of the slit before the detector determines the apparent
material quality. This ambiguity can be removed by analysing the direction of
the scattered X-rays from the crystal in the triple-axis geometry.
In the triple-axis mode, an analyser crystal is placed after the specimen and
before the detector. This is mounted on an axis concentric with the specimen and
is scanned independently of the sample. The experimenter can then map the
intensity distribution with respect to the direction of the radiation scattered by the
specimen. This not only removes the complication of a possibly bent or mosaic
specimen, but also enables one to distinguish scattering from various sources.
For example, scattering due to defects occurs in a different direction in space
from scattering from the perfect crystal and from a map of the scattering as both
the specimen and analyser are rotated, this can be measured quantitatively.
Scattering from a rough surface can be separated from the perfect crystal
scattering and most importantly, strain or mismatch may be distinguished from
tilt or mosaic spread.
Figure 7.1 Double-axis rocking curves from a GaN epitaxial layer on (111) orientation
GaAs. The measured rocking curve width is determined by the detector aperture
7.2
Instrumentation
Figure 7.2 shows the schematic arrangement for triple-axis diffraction. The three
axes are those controlling the beam conditioner (and hence the input beam), the
specimen, and the analyser, respectively. Although the term ‘triple crystal’ is
sometimes used for this arrangement, but it is best avoided since it may
erroneously be thought of as applying to multiple-crystal beam conditioners in an
ordinary (i.e. double-axis) high resolution experiment. It is seen that the only
difference between the triple- and double-axis instruments is the provision of a
means to restrict the angular acceptance of the detector.
7.2.1
Mechanical precision
The angular precision required is substantial. The most important factor is for the
detector axis to track the specimen axis accurately and continuously, to better
than an arc second for most semiconductor work, corresponding to a reciprocal
space resolution of 5×10− 6 A− 1. Random errors in the tracking result in ‘noise’
in reciprocal space, while systematic errors give rise to systematic distortions of
the reciprocal space map. Artefacts due to backlash and eccentricity of gear
trains are noticeable, and direct axis encoders are much preferred. Absolute
TRIPLE-AXIS X-RAY DIFFRACTOMETRY 159
Figure 7.2 Diagram of a high resolution triple-axis instrument
accuracy is less important since reciprocal space maps are normally viewed,
rather than measured, and errors around 1% are unlikely to be noticed.
7.2.2
The beam conditioner
The beam conditioner has two jobs: to condition the beam in angle and in
wavelength. The angular conditioning may be improved by using channel-cut
crystals 1 thus reducing the streak in reciprocal space discussed in section 7.7.
In practice, four reflections are required to make the streak scarcely noticeable.
Variation of wave-length also affects the resolution, since all the important
parameters scale with 1/λ . Eliminating the K λ 2 component is normally

sufficient, which may be achieved with a slit after the conditioner, but any of the
monochromating arrangements described in Chapter 2 may be used if the
intensity is sufficient.
7.2.3
The analyser and the detector
The angular restriction may be performed in a number of ways: by a slit, Soller
slit, single or channel crystal or by an area detector. The channel-cut analyser
crystal gives two advantages if intensity is not at a premium. As with the beam
conditioner crystal it reduces the instrumental streak in reciprocal space. It also
preserves the direction of the beam scattered from the sample, simplifying the
alignment. If only crude resolution is required, a slit may be sufficient (for
example, to remove most of the effects due to the mosaic spread in the analysis of
poor crystals). A single germanium crystal gives a high overall intensity.
For rapid low-resolution work, the use of an area detector both provides
sufficient angular resolution and gives rapid data collection. 2 , 3 Otherwise,
any detector may in principle be used, but a detector with a wide dynamic range
and with good performance at the low-signal end is highly desirable, since
effects such as diffuse scatter are weak and the X-ray optics necessarily reduce
intensity.
7.3
Setting up a triple-axis experiment
At first sight, the complexity of yet another Bragg reflection seems appalling.
However, this is not the case with modern instruments with computer-controlled
alignment. With a little experience, triple-axis measurements will be no more
challenging than most double-axis ones. Whilst the details of alignment will
depend on the particular instrument, the principles are common for systems using
crystal analysers:
1 Align the incident beam monochromator to deliver a beam passing

accurately over the second (specimen) axis. This procedure is exactly as for
high resolution double-axis diffraction.
2 With the specimen removed drive the detector axis (which carries the
analyser) to the zero angle position. Then independently orient the analyser
crystal to its approximate Bragg angle. Rotate the analyser crystal,
independently of the detector axis, to locate the Bragg peak.
3 Use the tilt adjustment on the analyser to maximise the intensity in the
Bragg peak. This must be done iteratively, tilting and relocating the Bragg
peak. When the maximum has been reached, the analyser is correctly set
with Bragg planes parallel to those of the beam conditioner and

monochromator. Set the analyser onto the Bragg peak maximum.
4 Translate (or rotate) the analyser away from the detector so that the
instrument is in double-axis mode. It is important that this motion should be
reproducible so that the analyser may be returned within about an arc second.
5 Place the specimen on the second axis and with the detector axis rotated to
twice the specimen Bragg angle, locate the Bragg peak of the specimen as in
a double-axis experiment. Perform any necessary tilt adjustments of the
specimen exactly as for double-axis diffraction. (If you wish you may
perform a double-axis experiment at this stage.) Set everything onto the
maximum of the Bragg peak.
6 Finally translate the analyser back in front of the detector and if necessary
make a final small adjustment of the analyser rotation to maximise the
signal once more. You are now set to perform one of several types of triple-
axis experiment.
Triple-axis measurements normally take this ‘everything at maximum’ position

as the ‘zero’ setting for reciprocal space maps. Two qualifications are needed.
The position will of course not be the origin of reciprocal space but the
reciprocal lattice point corresponding to the reflection used. Secondly, in the
usual Bragg case it will be displaced by a few arc seconds to higher angles by the
refractive index effect in dynamical scattering.
7.4
Separation of lattice tilts and strains
Triple-axis scattering enables the user to distinguish between tilts and dilations.
This may be seen by considering the specimen shown in Figure 7.3, containing
regions which are tilted with respect to each other, i.e. sub-grains and those that
are strained or mismatched, e.g. ternary layers. By the Bragg law, the scattering
angle defines the d spacing that is being examined. As the specimen is rotated,
differently tilted regions will satisfy the condition for diffraction in sequence and
the scattered intensity gives a measure of the distribution of tilts. Regions of the
crystal where the Bragg plane spacing differs will never give rise to strong
scattering when only the specimen is rotated.
Suppose we now perform a scattering experiment in which the specimen and
analyser are scanned in synchronisation. Specifically, the analyser is scanned at twi
ce the rate of the specimen (a λ –2λ scan) both starting from zero. Suppose that a
region of the specimen of lattice parameter d is set to diffract; then, as the
analyser is set at twice this angle, intensity reaches the detector. If we now
perform a λ –2λ coupled scan, any region of the specimen which also has lattice
parameter d but is tilted with respect to the original region will never provide
scattering which reaches the detector; the analyser setting will never be correct
(unlike the example in the previous section). However, another region of the
Figure 7.3 Triple axis measurements; real and reciprocal space representations
sample with lattice parameter dλ may come to a position where the Bragg angle is
satisfied. Now the analyser is at twice this angle and intensity reaches the
detector. In this mode we record intensity from only parts of the crystal, but for
heavily distorted materials such as gallium arsenide on silicon or small gap II–VI
compounds, this provides a valuable measure of the range of lattice parameters
present. In the case of ternary compounds such as cadmium mercury telluride,
this provides a measure of the composition range, independent of the range of
tilts.
Note also the resolution functions in reciprocal space of the double- and triple-
axis measurements. In the double-axis rocking curves (shown) we cannot
distinguish between the tilts and the dilations, but they are very different types of
defect. The triple-axis resolution function in the dispersion plane is the
intersection of two vectors—incident and diffracted beam—spread out by the
angular width of the monochromator and analyser, respectively. A further
blurring occurs in the plane normal to the figure, due to the vertical divergence,
and the final component of the resolution function is the wavelength spread; this
changes the lengths of the incident and diffracted beam vectors.
7.5
Measurement of kinematical scattering from a defective
surface
The scattering from a distorted surface region is no longer governed by the
dynamical scattering in the bulk of the crystal. In the bulk, due to multiple
scattering processes, the refractive index is not quite unity and the result is that
the position of the Bragg peak is shifted by a few arc seconds with respect to the
value calculated by the Bragg equation. The kinematical scattering from the
defective surface is, however, not shifted, nor does it have the same symmetry
and in principle these effects can be distinguished.
In the double-axis geometry, the kinematical scattering lies within the
envelope of that from the perfect bulk crystal and cannot be separated. However,
if the specimen is set some way off the Bragg position, rotation of the analyser will
pick up this scattering. The intensity is extremely sensitive to surface preparation.
Figure 7.4 shows a broad diffuse peak due to polishing damage in gallium
arsenide, superimposed on the sharp Bragg peak.
7.6
Measurement of distorted specimens
If a specimen is heavily dislocated it may be impossible to distinguish peaks in
double-crystal rocking curves. Figure 7.5 shows a five-layer Si-Ge specimen
grown with 0.5 μ m thick layers, with the Ge content in each layer respectively
10, 20, 30, 40 and 50%; the aim being to produce a moderate amount of
relaxation at each layer so that the top layer is fully relaxed 50% Ge but without
excessive threading dislocations. This was achieved—the dislocation density was
Figure 7.4 Defect peak in GaAs due to polishing damage. 004 reflection with CuK λ .
Triple axis λ –2λ scans with displacements of specimen from Bragg condition: solid
(upper) line, 10■ displacement, next highest line, 20■ displacement, lowest line, 20■
displacement
about 106cm− 2—but development of the process was not possible using only
double-crystal diffraction even with a channel-cut collimator, as shown in the
Figure 7.5(a). The range of the broad peak covers the 0–50% Ge contents, but no
details are visible. In the triple-crystal setting, however, the separate layers are
easily visible in a λ –2λ scan (Figure 7.5(b)). A single Ge crystal was used as the
analyser for these weak signals.
Once such detail is available it is possible to scan in the perpendicular
direction in reciprocal space, using a λ scan, to measure the tilt independently of
the strain. This is shown in Figure 7.5(c), in which the analyser was set on the
prominent left-hand peak in the triple-crystal curve. The tilt variation in this
layer is seen to be about 1500■ . Similar measurements on other peaks and
correlation with crystal growth parameters can provide excellent feedback to the
crystal grower.
The specimen-only rocking curve measures mosaic tilt very efficiently and it
is intrinsically very narrow; it may be measured in no more time than the double-
axis rocking curve and is about a quarter of its width. It may be used, for
example, to characterise overall wafer quality. Figure 7.6 shows wafer maps of
the width of the triple-axis specimen-only rocking curve as a function of position
on the wafer. The distinction between the two classes of GaAs wafer is very
easily seen.
For thick layers of highly mismatched systems, the double-axis rocking curves
often turn out to be dominated by tilts. This is the case for the II–VI materials
CdTe and Hg x Mn1− x Te grown on GaAs. 4 Another example is GaN on
sapphire 5 shown in Figure 7.7. The tilt distribution can be modelled assuming
that there is a random distribution of the edge components of the misfit
dislocations normal to the surface. This results in a Gaussian distribution, which
is quite well represented in Figure 7.7(a). Note that the coupled λ –2λ scan is
very narrow in comparison with the specimen scan. This shows that there is little
strain in the epitaxial layer but a large tilt distribution.
7.7
Full reciprocal space mapping
A full map of the scattering from the specimen can be made by recording
intensity from a series of separate specimen and analyser positions, which are
coupled so as to trace out a grid in reciprocal space. Such a contour map is shown
schematically in Figure 7.8, after Iida and Kohra. 6 We may understand what is
being measured in a triple-axis experiment with the aid of this figure. The angular
positions of the incident beam and of the analyser define two vectors which
define the scattering vector and hence angle. The angular position of the
specimen defines the position of the diffracting planes whose scattering is being
measured at this scattering angle. In Figure 7.8, the central point is the reciprocal
space point of the diffracting planes, for example 004. The scattering is being
measured in this instance from the small volume surrounding the point [q y , q z ].
The scattering vector Q (not shown directly on Figure 7.8) may be considered as
the sum of the 'ideal' scattering vector from the origin to point h, plus a deviation
q; thus
Q=h+q (7.1)
The deviation vector q , with origin at the end of the reciprocal lattice vector h ,
has two components, q y horizontal, positive rightwards going, and q z vertical,
positive upwards going. For the symmetric reflection, these components are
related to the deviations of specimen (■ λ ) and analyser (■ ) from their zero
positions at the nominal Bragg angle by the equations:
(7.2)
(7.3)
Thus a scan of the specimen axis alone (■ λ ) affects only q y and provides a scan
from left to right in reciprocal space. A scan of the analyser affects both q z and
q y and in fact sweeps along the Ewald sphere. A scan of q z alone may be
achieved by setting
(7.4)
Figure 7.5 (a) Double-crystal rocking curve of the five-layer Si-Ge structure. (b) Triple-
axis λ –2λ longitudinal scan of the same structure. (c) Specimen only (rocking curve) scan
with the analyser set on the left-hand peak in the triple-axis curve. 004 reflection with
CuK λ , single-reflection Ge analyser for (b) and (c)
Figure 7.6 Wafer maps of the triple-axis rocking curve width (specimen-only scan) for
GaAs grown by different processes. (a) A LEC GaAs wafer and (b) a VGZ GaAs wafer
i.e. scanning the analyser at twice the rate of the specimen, usually called a λ −2λ
scan (which notation has to be changed for the above description to avoid
confusion between the axes).
Geometrically, we may visualise the above scans as follows. Scanning the
specimen alone is equivalent to rotating the reciprocal lattice about its origin; the
end of k h thus describes an arc about 0, which at the scale of the drawing is a
straight horizontal line. Scanning the detector alone is equivalent to changing the
angle between k 0 and k h, thus describing a scan along the Ewald sphere. The
2■ λ −■ scan moves vertically in reciprocal space arises because ■ λ and ■ are
respectively the angles standing on the centre and the circumference of the
Ewald sphere, subtended from the same arc, hence 2■ λ −■ by simple geometry of
a circle.
The general case of the asymmetric reflection is a little more complicated and
not usually given; the following derivation is due to Wormington. 7 The
deviation of the scattering vector, q , from the reciprocal lattice point, h, can be
calculated by considering the Ewald constructions shown in Figure 7.9.
Figure 7.9(a) illustrates the case when the specimen angle, λ , and the analyser
angle, λ , are set so as to satisfy the Bragg condition, 2dsin λ B=λ . Figure 7.9(b)
Figure 7.7 (a) Double- and triple-axis rocking curves (specimen scans) of a GaN epitaxial
layer on (0001) oriented sapphire. 0002 reflection CuK λ radiation. (b) Coupled
specimen-analyser scan in ratio 1:2. 022 (with 17.65° asymmetric cut) Si duMond beam
conditioners, 111 symmetric Si analyser
shows the Ewald construction after the specimen and analyser angles have been
changed by ■ λ and ■ λ respectively. From equation (7.1) we have
On rearranging equation (7.1) we have
Figure 7.8 A scattering map in reciprocal space. Equal intensity contours are shown
schematically, and the Ewald sphere is represented as a plane near reciprocal lattice points
0 and h. The dynamical diffraction from the specimen is displaced slightly from the relp
and from the centre of the diffuse scatter by the refractive index effect
q=h−Q (7.5)
From the geometry of the Ewald construction shown in Figure 7.7(b) we can
write ■ q in terms of its Cartesian components (q y , q z );
q y =1λ {cos(λ +■ λ −(λ +■ λ ))−cos(λ +■ λ )}−1/λ {cos(λ –λ )−cos
(λ )} (7.6)
and
q z =1/λ {sin(λ +■ λ )+sin(λ +■ λ −(λ +■ λ ))}−1/λ {sin(λ )+sin(λ
−λ )} (7.7)
Expanding the trigonometric terms we may rewrite equations (7.6) and (7.7) as
q y =1/λ {[cos(■ λ −λ )cos(■ λ −■ λ )−sin(λ −λ )sin(■ λ −■ λ )] −[cos(λ )cos(■ λ )−sin

(λ )sin(■ λ )]}−1/λ {cos(λ −λ )−cos(λ )} (7.8)
q z =1/λ {[sin( λ )cos(■ λ )+cos( λ ) sin(■ λ )] +[sin(λ −λ )cos(■ λ −■ λ )+cos(λ −λ )
sin(■ λ −■ λ )]} −1/λ {sin(λ )+sin(λ −λ )} (7.9)
If ■ λ and ■ λ are small, equations (7.8) and (7.9) can be simplified by using the
small angle approximations, sin(■ λ )■ ■ λ and cos(■ λ )■ 1. We may finally write
q y ■ 1/λ {sin( λ )■ λ −sin(λ −λ )(■ λ −■ λ )} (7.10)
q z ■ 1/λ {cos( λ )■ λ +cos(λ −λ )(■ λ −■ λ )} (7.11)
where for symmetric reflection:
λ =λ B , λ −λ =λ B
for asymmetric reflection with glancing incidence:
and for asymmetric reflection with glancing exit:
is the magnitude of the angle between the diffracting planes and the surface
normal.
Several points arise from Figure 7.8. The main dynamical scattering from the
specimen is a vertical streak, along the q z direction, which in a good quality
crystal has little extension in the q y direction. This scattering is displaced from
the reciprocal lattice point by the refractive index correction. However, diffuse
scattering appears as a weak but broad, approximately annular scattering region
centred on the reciprocal lattice point.
Streaks appear at ±λ B to the vertical axis. These are caused by the finite
angular resolution of the beam conditioner and analyser crystals and would be
absent if these crystals had rocking curves with no tails. They arise because at
these settings of analyser and specimen, in addition to the true signal from the
‘sampled region’ of reciprocal space, a small number of the intense signals from
the main peak pass through the tails of the analyser function. Hence the streaks
point towards the main peak, and their direction is found by imagining a small
oscillation of k 0 about the origin and of k h about the centre of the Ewald sphere.
In a system with high resolution and very wide dynamic range, similar streaks at
very low intensity levels can sometimes be seen from air scatter in the system e.g.
Figure 7.11—this has a similar effect in blurring the angular precision. 8
Finally, we note that the standard double-crystal or high resolution rocking
curve, in which we have no control over ■ is in effect a horizontal scan
through reciprocal space, integrating all intensities along the Ewald sphere. It is
thus easy to see how the triple-axis instrument can obtain much more
information.
Figure 7.9 Ewald constructions, (a) at the Bragg condition, (b) off the Bragg condition
7.8
Applications of reciprocal space mapping
We do not always need a synchrotron or a rotating anode X-ray generator to
perform these studies. The resolution required determines the reflection used
for the analyser. For low resolution measurements, a slit or a graphite crystal
analyser can be used. An asymmetrically cut 022 Si or a 022 Ge analyser gives a
modest resolution and for most systems this is a reasonable compromise between
intensity and resolution in reciprocal space. A four-reflection symmetric Si 111
or 022 analyser provides a high resolution option for III–V or IV–IV
semiconductor systems. The reciprocal space maps shown here were all taken in
a few hours at most on a sealed-tube generator system.
A convenient data collection strategy is to perform a loop scan in which the
specimen is set at a particular angle and then a λ –2λ scan is performed. The
specimen position is then incremented and another λ –2λ or longitudinal scan is
performed. In this way, one collects data in a square grid in reciprocal space.
Appropriate software is then used to convert the data to reciprocal space
coordinates using the equations given in the last section.
The first example we show is one of a defective region of a GaAs ingot grown
in space under conditions of microgravity. 9 The main region of the ingot had
good quality as indicated by 12■ wide rocking curves, but a second region,
grown after a heater failure, exhibited broad, asymmetric rocking curves as shown
in Figure 7.10(a). Quality is clearly low but the reason for this is not clear. The
reciprocal space map (Figure 7.10(b)) shows that there are two areas of scatter.
This feature is not revealed in the double-axis measurement! Within each region
there is a large distribution of tilted regions, that is a large mosaic. The two
regions appear displaced in the q z direction, and are probably associated with
lamella twins which are not quite coherent.
Growth of a thin epitaxial film of Ga As at low temperature results in the
trapping of point defects which give rise to a larger lattice parameter than in the
wafer. This comes about because excess As is incorporated, leading to an
expansion of the basic GaAs lattice. The reciprocal space map shows that the
epilayer has a differing lattice parameter (Figure 7.11) and that there is no tilt
between epilayer and substrate. This can be seen from the fact that the 004 peaks
of layer and substrate are vertically above each other in reciprocal space. There is
equal broadening in the q y direction in both substrate and epilayer peaks. This
arises from long range curvature of the wafer due to the strain from the epilayer.
(Note that the FWHM of the peaks for a longitudinal scan was 15 seconds,
whereas the double crystal rocking curves were 60 seconds wide.)
Figure 7.11(b) shows the reciprocal space map after a high temperature anneal
of the film. The effect of this has been to precipitate out the arsenic, resulting in
the lattice parameter of the now stochiometric matrix reverting to that of the
substrate. The scattering around the layer peak, which arises from the
precipitates, is circularly symmetric and much more extensive than in the
substrate.
In II–VI and III–V nitride compounds, epilayers are often rather imperfect.
We saw in Figure 7.7 that the double-axis rocking curve with an open detector
from a GaN epilayer on sapphire was almost identical to the triple-axis specimen
only scan. This indicates that the strain distribution arises predominantly from
tilts, not dilations and the λ –2λ scan is in contrast very narrow. A reciprocal
space map (Figure 7.12) of the sample shown in Figure 7.7 reveals very clearly
that the broadening of the double-crystal rocking curve comes predominantly
from tilts, not dilations. The elongation of the scatter distribution in reciprocal
space arises from the specimen-only scan being very much wider than the λ –2λ
scan. Note that the layer peak does not lie exactly in the q z direction with
respect to the substrate peak. A simple λ –2λ scan thus cuts through the
shoulder of the layer peak, giving an incorrect impression of the height and
width. Both the substrate and layer show a mosaic structure, although the latter
distribution is wider than the former.
Figure 7.13 shows a reciprocal space map of a typical HEMT structure which
contains a ~10 nm thick InGaAs layer on a GaAs substrate, together with a GaAs
cap. We note that the thickness of the fringes associated with a thin layer now
become contours of intensity, the maxima of which run along a line normal to
the specimen surface. It can be seen immediately from Figure 7.9 that the substrate
surface was offcut from the [001] direction, since the line of interference maxima
does not coincide with the q z direction. No significant broadening occurs of the
interference fringes in this direction as the deviation in q z increases. However,
there is a large region of strong diffuse scatter centred around the substrate
reciprocal lattice point which is much stronger than normally found from
substrate wafers. We would therefore ascribe this scatter to defects within the
capping layer.
The final example (Figures 7.14 and 7.15) shows the complexity of structural
detail that can be resolved through the use of reciprocal space mapping, and is
the work of Goorsky’s group (Eldredge et al .) 11 at UCLA. The specimen was a
GaAs substrate, followed by a buffer layer in which the In content of InGaAs
rose from 0 to 25% over just under two micrometres, in order to make a substrate
with a new lattice parameter, on which a superlattice with (+, −) strain was then
grown. This is a similar application to the five-layer Si-Ge structure discussed in
section 7.6, but grown with continuous grading instead of steps. The object of the
characterisation was to discover the degree of strain and relaxation in the buffer
layer to see if this end was achieved. A further complication arises in that the
substrate was tilted by 2° towards an <011> direction from the exact (001)
orientation.
The method used by the group was first to determine the direction of substrate
tilt by means of reciprocal space maps in each of the <110> directions contained
in the plane of the wafer. These are shown in Figure 7.14. A narrow slit was
used instead of an analyser crystal, so some analyser streaks are seen near the
substrate peaks (on the right of each map—the origin in this case is on the left-
hand sides). It is seen that in (a) and (b) the beam is oriented perpendicular to the
substrate tilt axis, as these maps show only strain. In (c) and (d) the effect of the
grading can be seen, since both the tilt and the strain are changing, but these
views are insufficient to make a complete analysis. For this we need an
Figure 7.10 (a) Double-axis rocking curve of a microgravity-grown GaAs crystal after
heater failure. (b) Equivalent triple-axis reciprocal space map CuK λ 004
asymmetric reflection such as 224. These are shown for 110 and 1 0 reflections
in Figure 7.15, with the 004 maps as insets, and this time the superlattice peaks
are also seen.
Figure 7.11 Reciprocal space map around 004 for a low temperature grown GaAs
epilayer on GaAs. (a) As grown. (b) After high temperature anneal 10
The interpretation is as follows. The diagonal streak in the 2 4 reflection is not
very helpful since it simply shows us that there is both tilt and strain present in
the buffer layer, and we do now know whether the tilt was caused by the offcut or
by the growth. However, the 24 map was taken in the direction insensitive to tilt
arising from the substrate offset and we can be sure of the interpretation. The
region of intensity that runs radially from the substrate peak towards the origin
must have a cubic structure of varying lattice parameter, i.e. it is fully relaxed.
The part that runs diagonally and merges with the superlattice reflections must
Figure 7.11 (cont.)

represent a tetragonally strained region, i.e. it is not relaxed. The fact that this
strained region merges with the superlattice reflections means that the
superlattice is modulating the strained region of buffer layer, not the relaxed
region. Since the superlattice can only have coherency with the surface on which
it is grown, not that further down, then the conclusion must be that the buffer
layer is first relaxed (maybe due to too high a composition gradient), up to 20%
In, and then becomes strained and coherent. The growth objective of a fully
strained buffer layer was thus not achieved, but clear pointers to the process
problems were discovered by this careful measurement.
7.9
Summary
In this chapter we have seen the new dimension that triple-axis diffractometry
and reciprocal space mapping bring to the characterisation of complex materials.
Figure 7.12 Reciprocal space map of the GaN sample shown in Figure 7.7
On the one hand they provide new ways of looking at materials that are not of
good enough quality to use for the double-axis methods, and, on the other, they
show how very complicated structural problems can be disentangled by this extra
dimension.
References
. P.ZAUMSEIL & U.WINTER, Phys. Stat. Sol. (a), 70 , 497 (1982).

. L.R.THOMPSON, G.J.COLLINS, B.L.DOYLE & J.A.KNAPP, J. Appl. Phys., 70 ,
4760 (1991).
. S.R.LEE, B.L.DOYLE, T.J.DRUMMOND, J.W.MEDERNACH & R.P.SCHNEIDER,
JR, Advances in X-ray Analysis, 38 , 201 (1995).
. T.D.HALLAM, S.K.HALDER, J.M.HUDSON, C.R.LI, M.FUNAKI, J.E.LEWIS, A.W.
BRINKMAN & B.K.TANNER, J. PHYS. D: Appl. Phys., 26 , A161 (1993).
. T.LAFFORD, N.LOXLEY & B.K.TANNER, Mater. Res. Soc. Symp. Proc. 449 , 483
(1997).
. A.IIDA & K.KOHRA, Phys. Stat. Sol. (a), 51 , 533–42 (1979).
. M.WORMINGTON, Bede Scientific Instruments, private communication.
. R.MATYI, Rev. Sci. Instrum., 63 , 5591 (1992).
Figure 7.13 Reciprocal space map of a HEMT structure which contains an approximately
10 nm thick InGaAs quantum well and is capped by a thick GaAs layer. The substrate is
(001) GaAs
. N.LOXLEY, C.D.MOORE, M.SAFA, B.K.TANNER, G.F.CLARK, F.M., HERMANN,
& G.MUELLER, Adv. X-ray Analysis, 38 , 195 (1995)
Figure 7.14 Reciprocal space maps of GaAs with graded InGaAs buffer layer around the
004 point with different directions of the incident beam. (a) Along [110], (b) along [ 0] (c)
along [ 10],(d) along [1 0] In this and the next figure the orientation is such that the origin
is on the left (not the bottom) of the maps, hence the horizontal direction represents strain
and the vertical direction, tilt
Figure 7.15 Reciprocal space maps of GaAs with graded InGaAs buffer layer around the
224 reciprocal lattice point. (a) [110] direction, (b) [ 10] direction. The 004 maps in those
directions are shown in the insets
. J.M.HUDSON, PhD thesis, Durham University (1993).

. J.W.ELDREDGE, K.M.MATNEY, M.S.GOORSKY, H.C.CHUI & J.S.HARRIS, JR, J.
Vac. Sci. Technol., B13 , 689 (1995).
180
8
Single-crystal X-ray Topography
In this chapter, we review the fundamental contrast mechanisms in X-

ray topographs and describe the Berg-Barrett, section and projection
topography techniques. Details are given of procedures for setting up
and recording topographs. We discuss limitations on real-time
topography and detector requirements. In an extensive section on the
contrast of planar and line defects under various diffraction
conditions, we provide some additional theoretical tools to aid
interpretation.
8.1
Introduction
We have so far been concerned only with the measurement of the integrated
intensity arriving at the detector as a function of angular setting of various
crystals with no consideration of the spatial distribution of that intensity. Indeed,
it was not until 20 years after the discovery of X-ray diffraction that the first
papers appeared which reported studies of the variation of the diffracted intensity
as a function of position across a crystal. In such experiments, the intensity
variation across the diffracted beam is recorded, usually photographically, and
thus a map of the scattering power is recorded as a function of position. Such X-
ray diffraction topographs are analogous to transmission electron micrographs.
Despite the name, the technique is not principally sensitive to surface
topography; it is the topography of the crystal lattice planes that is examined.
As is the case in many branches of science, it is difficult to identify exactly
when or by whom X-ray topography was invented. The early methods of Schulz
1 and Guinier and Tennevin 2 which recorded the intensity distribution across
Laue spots were not capable of resolving individual dislocations. Normal to the
incidence plane (i.e. that containing incident and diffracted beams) the spatial
resolution is determined very simply by the projected size of the X-ray source.
Owing to the large size of the X-ray sources and the need to place the film a
significant distance from the specimen in order to separate the different Laue spots,
only large-scale strains, such as are associated with scratches due to polishing
damage, were resolved. However, as we will see in Chapter 10, with the building
of synchrotron radiation sources it has now become possible to use these simple
techniques for very fast, high resolution work.
Individual dislocation images were first observed in X-ray topographs by Lang
3 using his section topography technique and almost simultaneously by Bonse
and Kappler 4 using the double-crystal method. The latter technique will be
discussed in detail in Chapter 9 and in this chapter we will concentrate on single-
crystal techniques which use the characteristic radiation from an X-ray tube and
which are capable of high resolution imaging.
8.2
Contrast mechanisms
The aim of all X-ray topographic methods is to provide a picture of the
distribution of the defects in a crystal, and the X-ray images may be thought of
as arising in two ways. Orientation contrast arises when a region of the crystal
is misoriented by an amount larger than the beam divergence (Figure 8.1). Then
for the characteristic line, no diffracted intensity is recorded for region A when
the Bragg condition is satisfied for the rest of the crystal. Thus there is an
undarkened patch on the film. The angle of misorientation, projected into the
incidence plane, can be determined by the angle that the specimen must be
rotated in order to obtain strong intensity from the region A. This is very
conveniently achieved in real-time using an electronic imaging detector but is
tedious to perform with photographic recording. Orientation contrast may arise
from the presence of twins, sub-grains and electric and magnetic domains and
can be interpreted in a simple geometric manner without need of detailed
dynamical diffraction theory calculations. Most of the contrast in double-crystal
topographs is of this kind, due to the very low angular divergence of the beam at
the specimen (see Chapter 9).
Another type of orientation contrast, discussed in Chapter 10, arises when
continuous radiation is being used. It is most common at synchrotron sources but
can sometimes be seen in single crystal topographs in the laboratory. Here the two
regions A and B satisfy the Bragg condition for different wavelengths. The beams
then make different directions in space and can either overlap or diverge,
depending on the relative misorientation of regions A and B. Simple images
occur where the crystal contains discrete mosaic blocks, but when the lattice
distortion is continuous, the resulting contrast can be very complex and difficult
to interpret.
The second major contrast mechanism is extinction contrast. Here the
distortion of the lattice around a defect gives rise to a different scattering power
from that of the surrounding matrix. In all cases, it arises from a breakdown or
change of the dynamical diffraction in the perfect crystal. In classical structure
analysis, the name extinction was used to describe the observation that the
integrated intensity was less than that predicted by the kinematical theory.
SINGLE-CRYSTAL X-RAY TOPOGRAPHY 183
Figure 8.1 Orientation contrast from a monochromatic, collimated X-ray beam.

Diffraction occurs only in region A for a particular angular setting of the specimen
Around the defect, enhanced scattering was observed and this ‘loss of extinction’
is the origin of the name. The exact nature of the images can only be explained
using dynamical diffraction theory and we will return to this in a later section.
8.3
The Berg-Barrett technique
This Bragg (reflection) orientation technique, developed in 1945 by Barrett5
following earlier work by Berg, 6 is the simplest high resolution technique to set
up in the laboratory. It uses an extended X-ray source and the specimen is
adjusted so that diffraction conditions are satisfied for the characteristic K λ
lines from a specific set of Bragg planes. Using crystals cut in such a way that
the diffracting planes are aligned so that the incident beam makes a very small
angle to the specimen surface (Figure 8.2), it is possible to place the
photographic film within about 1mm of the specimen surface. Best results are
obtained for planes where the Bragg angle for the characteristic K λ lines is
nearly 45°, so that the diffracted beam emerges almost normal to the specimen
surface. This configuration has the convenient feature that it is very insensitive to
orientation contrast. It also has two important consequences for spatial resolution.
The characteristic K λ line consists of a closely spaced doublet with a 2:1
intensity ratio between the K λ 1 and K λ 2 lines. This not a problem if a low
spatial resolution recording medium, such as a fluorescent screen and image
intensifier, is used but is a problem if high resolution nuclear emulsion plates are
used. Then individual dislocation images are doubled because the diffracted
beams from the two lines make different directions in space, giving rise to a type
Figure 8.2 Schematic diagram of the Berg-Barrett technique

of orientation contrast. However, if the specimen to film distance is kept small,
the divergence of the beams is negligible and sharp individual images can be
recorded (Figure 8.3).
The second issue concerns the spatial resolution achievable with this setting.
To a good approximation, the spatial resolution normal to the incidence plane λ
is related to the specimen to source distance D , the specimen to plate distance L
and projected height of the source H by the simple geometrical relation:
λ = HL / D .
This is illustrated in Figure 8.4. The benefit of being able to place the plate very
close to the specimen is immediately clear. For example, with a projected source
height 0.04 mm, distant 160 mm from the specimen, a spatial resolution of 1.25
μ m can be achieved for a 5 mm specimen to plate distance. Despite its age and
simplicity, the Berg-Barrett technique is still in use today, particularly for the
initial assessment of crystals of new materials where the perfection is relatively
low.
If Bragg planes cannot be chosen, or the crystal cut, so that the diffracted beam
emerges almost normally to the specimen surface, the diffracted beam will pass
through the recording photographic film at an angle. The consequent loss of
resolution can be minimised by use of a very thin emulsion (e.g.10 μ m thick),
but then a soft radiation such as CrK λ should be used to ensure that a significant
amount of the beam is stopped by the film. Use of soft radiation does often turn
out to be convenient, as the large Bragg angles then encountered make it easier to
find low-order, grazing incidence, asymmetric reflections with near 90°
scattering angle.
An important feature of the use of soft radiation is that both the extinction
distance and the absorption distance are small, and therefore the X-rays penetrate
only a very small distance into the crystal. We therefore examine only a small
slice of the crystal close to the surface. The strain fields of dislocations deeper
into the crystal do not contribute significantly to the image and, in a transmission
experiment, overlapping of the images leads to an upper limit of about 104 cm− 2
on the dislocation density for individual defect imaging. With the Berg-Barrett
technique this can be pushed to about 106 cm− 2.
The large rocking curve width associated with the extended source makes it
easy to set up the Berg-Barrett topograph. This can be done with a counter but is
most easily achieved with a phosphor screen and TV camera. The strength of the
beams is such that only modest gain is needed and the equipment can therefore
Figure 8.3 Berg-Barrett topograph showing individual dislocation images in a single

crystal of zinc. (Courtesy B.Roessler)
Figure 8.4 Schematic diagram showing the geometrical resolution limit set by the
projected source height normal to the incidence plane
be of quite low cost. Early workers used to look, under black-out conditions, for
the image on a piece of fluorescent paper attached to a hand-held stick. Such
practice is now illegal in most countries owing to the significant radiation dose
received during this method of alignment.
8.4
Lang topography
8.4.1
Section topography
This transmission method 3 provides an image of a section through the crystal
and as such enables the experimenter to study the three-dimensional distribution
of defects. The beam from the spot of a fine-focus or microfocus source is
collimated into a ribbon beam of width approximately 10 μ m before the single-
crystal specimen. This provides an incident beam of width small compared with
that of the base of the Borrmann fan formed by extremes of the diffracted and
transmitted beams with the crystal surface. [In other words, the beam width must
Figure 8.5 Schematic diagram of the projection topography configurations. For projection
topographs, the specimen and film are translated across the beam. (Note that very narrow
(~10 μ m) slits are used for section topography whereas those for projection topography
are typically 0.25 mm)
be much less than t sin 2λ B where t is the specimen thickness and λ B the Bragg
angle.] The specimen is adjusted until a strong diffracted beam from the
characteristic K λ 1 line is obtained for the diffraction planes chosen and the
photographic plate placed behind the specimen. Owing to the high collimation,
simultaneous diffraction from the K λ 1 and K λ 2 lines cannot be achieved and
there is thus no problem of image doubling. A diffracted beam slit prevents the
main beam from striking the photographic plate. The geometric arrangement is
such that the plate can be placed within about 10 mm of the specimen. For a
source to specimen distance of 1 m and projected source height of 0.3 mm the
geometrical resolution is then 3 μ m. Lang used a microfocus source to improve
this to 1 μ m.
8.4.2
Projection topography
The section topograph provides an image from only a narrow strip down the crystal
and in order to obtain an image of the whole crystal and still retain the high
spatial resolution, Lang devised a goniometer in which the crystal and film were
translated together across the beam 7 (Figure 8.5). Lang cameras, which are two-
circle goniometers equipped with a precision translation stage and adjustable (or
interchangeable) incident beam slits, are available commercially and the
technique has become the most widely used laboratory technique of X-ray
topography. Section and projection topography can be performed with the same
apparatus, although it is advisable to replace the very narrow slits necessary for
section topography with wider ones when projection topographs are taken. This
reduces exposure times very considerably.
8.5
Experimental procedures for taking Lang topographs
In experimental science there is no substitute for experience, but if certain
procedures are followed and adapted, a number of pitfalls can be avoided. The
following sections are written for the benefit of beginners in topography, as this
is a field relatively strange to practitioners of diffraction. While most of the
material here can already be found in the Iiterature 8 , 9 these works are out of print
and much exists now only in the folklore.
8.5.1
Setting up the crystal
An important feature, often overlooked, is the technique of mounting the crystal.
X-ray topographic techniques are necessarily sensitive to strain, and clumsy
mounting will lead to difficulties. When the crystal is small, it is useful to mount
it inside a suitable metal ring which is fixed to the goniometer. This enables a
large number of reflections to be taken simply by rotating the mount. In plastic
crystals many dislocations can be introduced and in brittle materials a radius of
curvature as great as 2 m will prevent Bragg reflection from occurring
simultaneously across a 1 cm wide sample. For many purposes a soft wax is a
suitable mounting material provided that only the minimum necessary to hold the
crystal is used. (Cenco Soft-Seal wax, melted by gently heating with a small
soldering iron run cool works very well.) Fixing at one point only is essential for
all cements. Quick-drying glues which contract on setting should be avoided.
Epoxy resin or strain gauge cement can be successfully used for permanent
mounting as they contract very little on setting, and a general rule is to apply the
adhesive to the sample as far away from the region under examination as
possible. At low temperatures a small quantity of varnish works well. Cyano-
acrylate glues (super-glues), used with brittle materials such as silicon, can be
removed through thermal shock by immersion in boiling water or liquid
nitrogen.
8.5.2
Setting the diffraction vector in the horizontal plane
This is equivalent to the tilt adjustment in the high resolution diffraction
experiment. It is, however, much less critical and can often be done by eye, for
example with crystals conveniently displaying facets or edges oriented in definite
crystallographic directions. (A draughtsman’s variable angle set square is a useful
tool.) For Lang topography this alignment to about 0.5° will be quite adequate,
though fine adjustment can be made once the Bragg reflection has been found.
Where the crystal has not crystallographically oriented edges, the whole
goniometer should be transferred to a back-reflection Laue camera and the
crystallographic orientation determined. Suitable rotation will then bring the

required diffracting planes into a vertical orientation.
8.5.3
Finding the Bragg reflection
When initially aligning the camera to take off the X-ray beam at about 4.5° to 6°
to the source, use of low X-ray power is advised and an absorber should be used
in the beam. Most scintillation and proportional counter detectors saturate well
below 106 cps. It is thus easy to believe that there are no X-rays present when, in
practice, one is aligned on the unconditioned beam. During operation, high
voltage—low current conditions give a good characteristic line intensity, and
care should be taken to ensure that the critical voltage is exceeded. (A Bragg
reflection will not be found with AgK λ at a tube voltage of 20 kV.)
Care should be taken to centre the crystal over the camera axis, but in any case
it is advisable to reduce the distance between specimen and scintillator for initial
setting. With a non-centred specimen, the effective 2λ angle of the counter will
not be given by 2λ B read on the counter scale. A wide angular aperture avoids
trouble but increases the background. Once the reflection is found, the angular
acceptance should be decreased and the scintillator moved further from the
specimen. This eliminates the possibility of having found a stray reflection from
non-vertical diffracting planes. Even with computer control and automatic
searching, ‘some days it is easy, some days it is not’. 8
Two peaks should be visible, the K λ 2 being half of the intensity of the K λ 1
peak. If this intensity ratio is not 2:1 it implies that the Bragg planes are not
vertical and the orientation of the crystal in the plane normal to the incident beam
should be adjusted.
When anomalous transmission studies are made of materials giving
appreciable fluorescence it is sometimes difficult to detect a Bragg reflection
from the thick crystal against the high background. In this situation the very
intense surface reflection from the edge of the crystal can be used to great
advantage. The crystal should be adjusted until it just cuts into the X-ray beam.
Once the edge reflection is found, the crystal can be traversed into the thick
region and the low visibility Bragg peak there will be seen. In the event of no
peak being observed one can conclude fairly certainly that anomalous
transmission is not occurring and the crystal is imperfect.
When the reflection has been found, the diffracted beam slits may be inserted.
With low fluorescence it may be satisfactory to use only a straight edge to
eliminate the direct beam, but in many cases the use of a slit is required. A broad
rocking curve in which it is impossible to distinguish the K λ 1 and K λ 2
components is a good indication that the crystal is highly strained and the
resulting topographs will not be works of art. They may, however, still be
scientifically informative. Similarly, a rapid change in intensity on traversing,
giving a strong signal over only a short range, is often indicative of a bent
crystal. In the last circumstance, provided the curvature is not too great and hard
radiation is used, it is possible to obtain good topographs by continuously
adjusting the angular setting of the crystal to keep it exactly on the Bragg
reflection. This is easy to implement with a simple algorithm in a computer-
controlled camera.
8.5.4
Recording the topograph
As will become apparent, it is important to place the photographic plate as close
to the specimen as possible. With rotating anode generators, care should be taken
not to allow the full power of the beam to fall on the plate when stationary as this
leads to an unsightly overexposed vertical line on the topograph. The presence of
a horizontal stripe on the recorded topograph is often due to the presence of a
second reciprocal lattice point lying on the Ewald sphere. It can be removed by a
small rotation of the crystal about the diffraction vector as if to take a stereo pair.
8.6
Topographic resolution
As the criteria differ in the horizontal and vertical directions, they will be
considered separately.
8.6.1
Photographic resolution
Unlike electron microscopy, the quality of X-ray mirrors is not sufficiently good
to magnify an X-ray topographic image. The topograph must be recorded at a
magnification of unity and subsequently enlarged optically. As dislocation
images are upwards of a micrometre in width it is essential that a recording
medium of comparable resolution is used. It is widely accepted that the Ilford L4
Nuclear Emulsions (undeveloped grain size 0.14 μ m) have excellent
characteristics for the recording of X-ray topographs. Dental or standard X-ray
film is fast and can be used to obtain a large area image, but the spatial resolution
is poor. The Agfa D2 and D7 films, with grain size 2 and 7 μ m, are often a good
compromise between expense, speed and resolution. For topographs of large
areas of semiconductor wafers they are very good. For high resolution optical
images, a very thin emulsion would be chosen to obtain high resolution, but
owing to the relatively poor absorption of X-rays there is a limit to the minimum
thickness tolerable. For harder radiations, e.g. MoK λ and AgK λ , 50 μ m thick
emulsion is used while for softer radiation 25 μ m thickness is satisfactory. These
constitute a compromise between efficiency and resolution. To obtain a
resolution of 1 μ m, the width of the projection of the ray in the emulsion must not
exceed this and we note that with a 25 μ m emulsion an alignment error of a 2°
error can be tolerated. Thin emulsions give a greater statistical fluctuation in the
number of developed grains per unit area and necessitate longer exposures.
Owing to the dense packing of grains in nuclear emulsions, it is difficult to
obtain uniform development and in order to reduce the development rate so it is
comparable with the diffusion rate of developer through the emulsion,
development for high resolution work should be performed at about ice
temperature. It is satisfactory to carry out developing in the main body of a
domestic refrigerator running as cold as possible. Prior to development, the
emulsion must be softened by soaking. The developer Kodak D19b keeps
reasonably well when concentrated but goes off rapidly when diluted and the
developer should therefore be changed daily. The stop bath and fixer can,
however, be used repeatedly. Plates should be washed in cold, filtered tap-water
for several hours, and then gently swabbed in running water with cotton wool to
remove any remaining particles. Extreme care must be taken in keeping the
solutions clean if good results are to be obtained. Plates should be dried under
cover in air and not force-dried. The processing times of Ilford L4 emulsions are
set out in Table 8.1.
Over a considerable density range, the Ilford emulsions are linear and it is
therefore possible to compensate for incorrect exposure in the development
procedure. While it is better to have an overexposed, underdeveloped plate than
vice versa, underexposure can be compensated to a fair degree, though the range
over which this is possible is limited. One should aim for a basic photographic
density of unity from the perfect diffracting crystal in the thin crystal situation
and then the dark ‘direct’ images give densities up to 2 or 3. In thick crystal
anomalous transmission topographs a basic density of about 2 from the perfect
crystal gives a good contrast from the (white) dynamical images but still enables
increases in intensity to be observed. A general rule of thumb is to develop until
a clear image is seen on the back of the plate under a red safelight.
Table 8.1 Processing times (in minutes) of llford L4 emulsions

Thickness
50 μ m 25 μ m
Soak in filtered deionised water 10 5
Develop (1:3 D19b to deionised water) 15–60 12–30
Stop (1 % glacial acetic acid in deionised water) 10 5
Fix (300 g sodium thiosulphate 30 g sodium 60 30
bisulphite in 1 litre deionised water)
Wash (filtered tap-water) upwards of 120 upwards of 120
On development, the grains of the L4 emulsion swell to about 0.25 μ m. However,

the granularity observed on the topograph when viewed under the microscope is
not a result of the grain of the film but is statistical ‘shot noise’, arising from the
statistical variation in the number of developed grains per unit area. This
granularity can be reduced by increased exposure, but this strategy is limited by

the low dynamic range of the emulsions. Lang 10 has considered the effect of
this noise on resolution in some detail.
When using hard radiations (MoK λ and AgK λ ) the resolution will be limited
also by the length of the photoelectron tracks in the emulsion which are about 2
μ m long with these radiations. Thus many factors conspire to limit topographic
resolution to between 1 and 2 μ m.
For high resolution topographs it is not usually satisfactory to magnify
topographs with a photographic enlarger. Light scattering reduces the contrast of
dislocations, and at the necessary magnification the depth of field is small.
Experience shows that the best topographs come from the best (and most
expensive) microscopes which have macro attachments for low resolution work.
Magnifications over 500× are almost never used, while 30× to 60× provides a
very useful range of enlargement. Photography should be performed on the
microscope, and topographs are conventionally printed as positives.
8.6.2
Real-time topography resolution
The photographic method, while giving excellent resolution and contrast, is
tedious and inconvenient in, for example, a semiconductor processing
environment. It is also inapplicable to dynamic experiments in which the defect
configuration varies. Real-time topography detectors fall into two categories—
those employing direct conversion of X-rays to electrical signals and those using
an X-ray to optical converter.
The former approach has been employed in the highest resolution direct
viewing systems. As developed by Chikawa and colleagues 11 the diffracted X-
ray beam passed through a beryllium window onto a thin PbO photocathode of a
sensitive vidicon TV tube. The sensitivity of the device is finally limited by the
target thickness, as a thin target absorbs only a small fraction of the X-rays, and
in thick targets only a thin layer on the X-ray entrance surface side becomes
photoconducting. Initially, Chikawa obtained a resolution of 30μ m, but
subsequent developments in the electronics of such a camera have resulted in a
resolution of 8 μ m. 12 Unfortunately, the tube is sensitive to radiation damage
and is not produced commercially in the high resolution form. Further, owing to
the limited number of TV lines, the field of view of the camera at high resolution
is necessarily small.
Indirect conversion devices have been reported with comparable spatial
resolution13 based on a thin fluorescent screen in the form of a microscope
objective lens. This again has a small field of view and poor efficiency owing to
the low stopping power of the phosphor screen. Commercial devices based on a
channel plate intensifier coupled to a CCD camera usually give a spatial
resolution of about 25 μ m, which is adequate for many industrial applications
where screening of substrates is required. However, only slow progress has been
made in the past decade in resolving the conflict between the requirements of
high spatial resolution, large field of view and high efficiency.
In order to see why these conflicts arise, we consider a flux I per unit area per
unit time of X-ray photons incident on a detector of efficiency λ . The number of
photons N detected in an integration time ■ in a square picture element of side λ
is then
N=Iλ λ 2 λ (8.1)
If two elements in the topograph have an intensity difference ■ I , then the
difference in signal ■ N is
■ N=■ Iλ λ 2 λ (8.2)
When written in terms of the contrast C, defined as C=■ I/(2I+■ I), this becomes
■ N=λ λ 2 λ 2IC/(1−C) (8.3)
Now the rms noise on the signal is (2N+■ N)1/2 and the signal-to-noise ratio R is
then given by
R=■ N(2N+■ N)−1/2−(C■ N)1/2 (8.4)
Thus
(8.5)
This is a form of the Rose-de Vries equation and we see that it sets a
fundamental limit on the spatial resolution of a quantum limited system. The
spatial resolution is improved by increase of the incident intensity I , the detector
efficiency λ and the integration time λ . The former has been greatly improved by
use of synchrotron radiation but there is a fundamental problem with improving
the efficiency. To retain the high spatial resolution, phosphors or TV target
regions must be thin and therefore stop only a small fraction of X-rays. Use of
heavy, rare earth-based phosphors or phosphors deposited onto etched fibre-optic
plates provide some ways of reconciling these fundamental contradictions. Note
that for quality assurance applications, the use of integration times of the order of
minutes may be perfectly acceptable and with PC-based image capture
electronics and software, very high quality images can be reconstructed by
digital integration.
8.6.3
Optimal contrast and speed
The intensity dI reaching the crystal due to a small area dX, dY of the X-ray
source is
dI=PdX dY/HVD 2 (8.6)
where P is the total X-ray tube power, H the dimension in the incidence plane, V
the projected (vertical) dimension of the source normal to the incidence plane,
and D the source to specimen distance. In the incidence plane only a fraction ■ λ
B of the reflecting range is diffracted and the total diffracted intensity becomes
(8.7)
Assuming a linear response of the film, the exposure time t for a stationary
topograph is therefore
t=(k■ λ B P/HD)−1 (8.8)
with k a constant of proportionality.
For a traverse topograph
T={k■ λ B (P/HD)(M/R cos λ B )}−1 (8.9)
where R is the length of traverse and M the slit width.
In order to minimise the exposure time, we chose M= H, so the whole source
is used. Then, neglecting the distance from the collimating slit to the specimen,
we require for monochromaticity
(8.10)
where is the angular separation between adjacent X-ray lines. Thus,
(8.11a)
for the stationary topograph and
(8.11b)
for the traverse topograph.
These equations give the optimum conditions for a rapid high resolution
topograph when only the resolution in the incidence plane is considered.
However we must also consider the resolution normal to the incidence plane
(Figure 8.4) and this is given geometrically by
λ =VL/D (8.12)
In order to separate direct and diffracted beams,
L>M/tan2λ B (8.13)
That is
(8.14)
Under optimum conditions the resolution normal to the incidence plane of a
Lang topograph will be
(8.15)
The importance of always placing the plate as close as possible to the specimen
is clearly demonstrated. Substitution of typical values (V=H and =2×10− 4) gives
a resolution of 2 μ m when L=1 cm. As the value of is a lower limit,
corresponding to a Bragg angle of 6° we note that it is only with extreme care
that the limits of resolution can be reached.
8.7
Contrast on X-ray topographs
8.7.1
Crystals apparently perfect
As a result of the relatively poor spatial resolution in X-ray topographs, there can
be confusion as to whether uniform contrast results from a highly perfect or a
highly imperfect crystal. There are, however, very important tests which can be
applied, none more so than for the section topograph where the contrast from a
perfect crystal is far from uniform.
Unlike the projection (or traverse) topograph, in section topographs of
parallelsided plates, interference, or thickness, fringes are observed (Figure 8.6
(a)). They arise from dynamical interference between waves propagating within
the Borrmann fan. As we saw above, the section topograph is formed when a
ribbon beam passes into the crystal and the diffracted intensity in a stripe down
the crystal is recorded on the photographic plate. As the divergence of the beam
is large compared with the perfect crystal reflecting range we cannot make the
assumption that the incident wave is a plane wave. This is unlike the situation in
transmission electron microscopy, where the strong scattering results in a very
large range of reflection and a plane wave approximation is usually valid.
Provided that the width of the incident beam is small compared with the width of
the Borrmann fan, to a first approximation the wave at the entrance surface can be
considered as a spherical wave. As a plane wave outside the crystal excites two
Bloch waves on opposite branches of the dispersion surface, the spherical wave
excites the whole of the dispersion surface and wavefields propagate in all
directions within the Borrmann fan (Figure 8.6(b)). As they are excited
coherently, they preserve their relative phase and will therefore interfere. At a
point on the exit surface R, two wavefields arrive, one from a point S on branch
2 and one from a point T on branch 1 of the dispersion surface (Figure 8.6(c)).
As these wavefields lie on different branches of the dispersion surface, they have
different wavevectors and there will be a phase difference between them at the
exit surface. The phase difference clearly changes across the base of the
Borrmann fan and across the section topograph of a parallel-sided, perfect
crystal a series of interference fringes is seen (Figure 8.6(a)).
These fringes are known as ‘Kato fringes’ 14 , 15 and they provide important
information on the dispersion surface shape. The density of the fringes is low at
the centre of the pattern, resulting from the angular amplification of the angular
deviation of the incident rays discussed by Authier. 16 The fringe density
increases towards the margins of the section topograph image, as here the
wavevector difference becomes greater for a smaller change in the normal to the
dispersion surface. With increasing absorption, the fringe contrast becomes
weaker, as only the branch 2 wavefield reaches the exit surface and in very thick
crystals, the section pattern becomes narrow, with only the branch 2 wavefield in
the centre of the dispersion surface reaching the exit and therefore no fringes are
visible.
As the effect is a ‘perfect crystal’ effect, the presence of small defects, which
may be too small or of too high a density to be imaged individually can be
detected from the destruction of the phase relationships and the resulting absence
of the Kato fringes. In crystals which have been irradiated or internally oxidised,
small point defect clusters develop, the existence of which can be inferred from
the loss of visibility in the section topograph thickness of Pendellösung fringes
(Figures 8.6(d) and 8.6(e)). The depth resolution of section topography can be
used to monitor the width of the denuded zone in silicon which has been
intrinsically gettered to remove oxygen from the near-surface region used for
device fabrication. 17 , 18
As sketched in Figure 8.6(c), the surfaces of constant phase are on hyperbolic
cylinders and thus in a parallel-sided plate the exit surface cuts these cylinders in
straight lines. Hence parallel-sided fringes are observed. When the crystal is
wedgeshaped, the exit surface cuts the hyperbolic cylinders along hyperbolae
(Figure 8.7) and thus hyperbolic fringes are seen in the topograph.
With reasonably high Bragg angle and a conventional generator, the Kato fringes
show a periodic fading and spacing shifts due to the two states of polarisation
having different wavevectors and thus different extinction distances (Figure 8.8).
Section topograph thickness fringes have been used to make extremely precise
measurements of structure factors. 20 − 22 Long-range strain is the most serious
potential cause of error as this causes contraction of the fringes.
Section topographs provide valuable information on the direction of energy
flow in the crystal. Although it may be shown that rays in the centre of the
Borrmann fan have higher intensity than those in the margins, this does not give
us the average intensity across the section topograph. The angular amplification
A is defined as
A=dλ /d(■ λ ) (8.16)
where d(■ λ ) is the angular divergence of an incident wave-packet and dλ is the
corresponding angular divergence of the wavefields inside the crystal. We recall
that the propagation direction of a wavefield is perpendicular to the dispersion
surface at the associated tie-point. Defining the dispersion surface radius of
curvature as R, it is then straightforward to show that
A=k cos λ B /R cos λ (8.17)
We see that very close to the Bragg condition, where the dispersion surface is
highly curved, R K and the crystal acts as a powerful angular amplifier. A
reaches 3.5×104 in the centre of the dispersion surface for silicon in the 220
reflection with MoK λ radiation. Far from the centre, the dispersion surface
becomes asymptotic to the spheres about the reciprocal lattice points and A
approaches unity. Thus when the whole of the dispersion surface is excited by a
spherical wave, owing to the amplification close to the Bragg condition, the
density of wavefields will be very low in the centre of the Borrmann fan and
Figure 8.6 (a) Spherical wave thickness, or Kato, fringes in a section topograph from a
perfect parallel-sided silicon crystal. (b) Schematic diagram of the section topograph
geometry showing the surfaces of constant phase and the ray path AR of wavefields
associated with tie-points S and T on the dispersion surface. (c) Dispersion surface
construction showing tie-points S and T. (d) Experimental section topograph of the same
crystal as in (a) showing loss of visibility of Kato fringes after annealing and precipitation
of oxygen. (e) Complete loss of Kato fringes after further annealing. (Courtesy J.R.Patel)
19
extremely high in the margins. Calculations of the intensity, averaged over the
spherical wave thickness oscillations, compounding the intensity and density of
wavefields, have been performed by Kato. 23
The very important feature of the low absorption curve, visible experimentally
in Figure 8.6(a), is that the diffracted beam intensity rises very sharply at the
edges of the section pattern. It arises from the very high density of wavefields in
these regions and its presence is a good indication that the crystal is highly
perfect. Point defect clusters, for example from radiation damage, will destroy
the phase relationships and destroy the margin effect. Under moderate absorption
conditions (μ t■ 2), however, the perfect crystal profile is very similar to that of a
Figure 8.7 Schematic diagram of the surfaces of constant phase in a wedge-shaped

crystal
Figure 8.8 Beating of Kato fringes due to the presence of two states of polarisation in a
section topograph of a crystal of diamond. CuK λ radiation. (Courtesy S.S.Jiang,
University of Nanjing)
crystal containing a large amount of damage and under such conditions

topographs should interpreted with care.
8.7.2
Thickness fringes in traverse topographs
Takagi 24 has shown that the intensity on a projection topograph is given by the
integrated intensity and that this is independent of the detailed shape of the
wavefront. The phase relations are lost, and we can use the integrated plane wave
intensity to calculate the intensity of the diffracted beam. For the perfect crystal
with zero absorption the integrated diffracted intensity is given by
(8.18)
where J 0 is the zero-order Bessel function. This is plotted in Chapter 4 as
Figure 4.23.
The term outside the integral shows that high-order, long extinction distance
reflections give very low intensity while the integral determines how the
intensity varies as a function of thickness. We note that the intensity oscillates
with pseudoperiod of λ g, the extinction distance. These oscillations are the
thickness or Pendellösung fringes observed in a wedge crystal in Lang and white
beam synchro-tron radiation topographs. As the thickness increases, the
amplitude of the oscillation decreases as (λ g/t)1 / 2 and thus the visibility
decreases. When absorption is included, the mean intensity decreases with
thickness.
8.7.3
Crystals containing sub-grains and dislocations
As discussed earlier, there are two fundamental mechanisms for contrast in X-ray
topographs. The first, orientation contrast, is appropriate to crystals in which the
dislocations configure to form low-angle sub-grain boundaries. In Lang and
Berg-Barrett topographs, orientation contrast occurs where part of the crystal is
misoriented in such a way that diffraction cannot occur at the same time as that
from the rest of the crystal. Zero intensity corresponds geometrically to the
misoriented region (Figure 8.1) when the misorientation exceeds the divergence
of the incident X-ray beam.
The second type of contrast is termed extinction contrast and results from the
scattering power around the defect differing from that in the perfect crystal.
There are essentially three types of extinction contrast, identified by Authier, 25
which can be seen schematically in Figure 8.9(a). The three images which are
formed are termed the direct image (1), arising from the diffraction of X-rays
which do not satisfy the Bragg condition in the perfect crystal, the dynamical
image (2) which is formed from changes in intensity in the Bloch wavefields
propagating through the perfect crystal, and the intermediary image (3), formed
by interference between these wavefields and new wavefields created at the
defect. These images are found in cases where the strain gradient is high, such as
around a dislocation, and this leads to the region around the defect behaving as if
it were the surface of the crystal. The wavefields decouple into their plane wave
components and create new Bloch wavefields when the strain field becomes
small again below the defect. A somewhat different type of dynamical image is
formed when the strain gradients are small (Figure 8.9(b)). Then the wavefields
do not decouple, but the tie-points migrate along the dispersion surface. As the
Figure 8.9 Types of extinction contrast. (a) Defect with high strain gradient. Three
images, the direct (1), the dynamical (2) and the intermediary (3) are formed. (b) Low strain
gradient where the wavefield rays paths are curved
propagation direction, that is the Poynting vector, is always normal to the

dispersion surface, the rays paths are curved in real space. These two effects lead
to redistribution of the energy between forward and diffracted beams, resulting in
contrast on the topograph.
8.8
Theoretical tools for the interpretation of X-ray topograph
contrast
8.8.1
Small distortions
When the deformation is small, the concept of a ray is retained and the crystal
deformation is described by a continuously varying reciprocal lattice vector. This
is equivalent to the situation in optics when light propagates through a medium
of slowly varying refractive index. The original modes of propagation in the
undeformed crystal retain their identity, in the X-ray case as Bloch wavefields,
but in order to accommodate the deformation each mode undergoes a gradual
transformation. Eikonal theories have been developed by Kato 26 and Penning
and Polder. 27 Here, we follow the treatment of Tanner. 8
In a distorted crystal, where the atomic displacement from the perfect crystal
is given by the vector u, we can define a local reciprocal lattice vector g by
g = g −grad (g.u) (8.19)
Let us consider two points F and G between which the local reciprocal lattice
varies from g to g + dg . If the deformation is small the shape of the
dispersion surface does not change and only a displacement of the hyperbolae
results. We can consider this as a rotation about the origin of reciprocal space,
and then the dispersion surface moves along the asymptote (Figure 8.10) by a
vector given by v, where
v . K 0=0 and v . K g= K g . dg (8.20)
No movement of the dispersion surface takes place when K g . dg =0 and thus
for a local reciprocal lattice where the vectors dg are always perpendicular to
K g, the wavefields propagate unchanged. Generally, the loci of points where dg
is perpendicular to K g form surfaces along which the dispersion surface
remains static and the associated wavefields unchanged. These surfaces divide the
crystal up into domains which can be considered as perfect and are analogous to
the surfaces of constant refractive index in the optical case. As each domain will
be misoriented or dilated with respect to its neighbour, the wavevector K 0
changes by dK 0 on crossing the domain boundary. In analogy with the optical
case this is given by
dK 0=λ grad( g.K g)=−λ grad( K g .grad( g.u )) (8.21)
As the fundamental equations of the dispersion surface are
(8.22)
(8.23)
where R=|D g|/|D 0| we derive, for λ g=

dK 0 .( K 0+R 2 K g)=−R 2 K g . dg (8.24)
If dr is the path vector between F and G
dg =( dr .grad) g (8.25)
Defining a vector P parallel to the Poynting vector given by
P = K 0+R 2 K g (8.26)
we find
λ =−R 2 dl/ |P| (8.27)
where dl is the path length along the ray trajectory given by
dr = P dl/ |P| (8.28)
Then
dR=(2R 2 dl/(k 2 Cλ g |P| ))( K 0 . grad) (g.u) (8.29)
which may be written as
(8.30)
Instead of considering the dispersion surface as a variable and the reciprocal

lattice as invariant, it is usually easier to consider the reciprocal lattice as the
variable. Then equation (8.30) determines the variation of the amplitude ratio of
the reflected and transmitted components as the wavefield propagates through
the crystal. The ratio R characterises a particular tie-point on the dispersion
surface and if R varies the tie-point must migrate along the dispersion surface
branch. This results in a change in the intensity of the transmitted and diffracted
Figure 8.10 Rotation of the local reciprocal lattice vector in a distorted crystal, giving rise
to a displacement of the dispersion surface hyperbola
beams. Further, as the direction of energy flow, i.e. the ray direction, is
everywhere normal to the dispersion surface, the ray direction varies as the tie-
point migrates. The rays therefore propagate along curved paths.
The important parameter in determining the contrast is the integral along the
ray path of the term {λ 2/λ s 0 λ s g}( g.u ). The ray path parameter λ is defined as
λ =(kCλ g)−1{λ 2/λ s 0 λ s g}( g.u ) (8.31)
From equation (8.30) we see that the sign of the change dR is independent of the
sign of R and depends only on λ . The sign of R is opposite for the two branches
of the dispersion surface but on both branches R increases with the deviation
from the exact Bragg condition. Thus, for a given deformation, the tie-points on
both branches of the dispersion surface migrate in the same direction. As has
been experimentally verified by Hart and Milne, 28 the curvature of the ray
corresponding to branch 2 is in the same sense as the curvature of the reflecting
planes while the curvature of branch 1 rays is of opposite sense to the lattice
curvature.
For the Eikonal theory to be valid, the distortions must be sufficiently small
that the concept of a ray is retained. The criterion is that the radius of curvature of
the reflecting planes does not exceed a critical value R c , approximately
equivalent to an angular rotation of the Bragg planes by half the reflecting curve
width in an extinction distance. The critical radius of curvature is thus
(8.32)
Although, Kato 29 has shown that the ray theory breaks down approximately 10
μ m from a dislocation core and we cannot apply it to compute the contrast of a
dislocation in a quantitative way, it is possible to obtain a large amount of
qualitative information.
In a thick crystal, as a result of the Borrmann effect, only the branch 2 wavefields
close to the exact Bragg condition are present and the energy flow is parallel to
the lattice planes. Let us consider first a defect such as a precipitate in the centre
of a thick crystal and look at the contrast a long way from the defect (Figure 8.11
(a)). On one side, the curvature is first positive and then negative while the
converse is true on the opposite side of the defect. Thus the tie-point migration is
in opposite sense on either side of the dislocation line but, provided the distortion
is small so that the tie-points retain their identity, the migration above and below
the defect exactly compensates on both sides of the defect. Close to the exact
Bragg condition (which corresponds to our branch 2 anomalously transmitted
wavefield) the angular amplification is extremely large (equation (8.17)) and
hence very small distortions of the Bragg planes lead to appreciable changes in
the direction of energy flow. The rays so deviated suffer additional absorption
because the effective absorption coefficient is sensitive to the deviation from the
exact Bragg condition. Thus there is a loss of intensity in the rays coming from
both sides of the dislocation, and the long-range image of the defect appears
white. Such contrast is commonly observed from dislo cations in anomalous
transmission topographs and a good example is the threading dislocation T in
Figure 8.11(b).
When the defect is close to the surface of a thick crystal, however, we may
observe opposite contrast on either side of the defect. Consider first the right-
hand side of the precipitate in Figure 8.11(a). The ray path parameter λ is first
negative and the tie-point of the branch 2 wavefield migrates to the right,
following the lattice plane curvature. At the surface the lattice planes emerge
normally, and when the defect is close to the surface the lattice planes must
rotate very rapidly. Thus the curvature is very large—too large for the Eikonal
theory to be applicable—and the rays pass out of the crystal without further
diffraction. Qualitatively, we find from equations (8.30) and (8.31) that the
migration causes an increase in the intensity in the diffracted beam. Conversely,
on the left-hand side of the defect we find decreased intensity. A precipitate close
to the surface shows black-white contrast which is enhanced on the side of
positive g when the surrounding lattice is under compression. An example of
the black-white contrast due to the differing sense of lattice curvature around
Figure 8.11 (a) Schematic diagram of tie-point migration in the long-range strain field of
a dislocation in the centre, and a precipitate near the exit surface of a thick crystal. (b)
Black-white contrast from misfit dislocations M and a precipitate P close to the exit
surface of a thick crystal. Note that the contrast of the threading dislocation T is white,
this being the dynamical image and oscillatory contrast; the intermediary image is seen only
near the exit surface
misfit dislocations and a precipitate close to the exit surface of a strongly

absorbing crystal of GaAs is shown in Figure 8.11(b).
The black-white contrast reverses with the diffraction vector and also with the
sense of the strain in the lattice. This is a useful means of determining the nature
of a precipitate with a convenient ‘rule of thumb’. On the side of positive g , if
the contrast is enhanced, the lattice is under compression, if reduced, it is under
tension.
8.8.2
Large distortions
When the distortion is high, we may obtain qualitative insights by retaining the
concepts of the perfect crystal and treating the defective region as a crystal
boundary. We then find that this leads to the concept of tie-points jumping from
one branch to another when we match the wavefields across the interface.
However, for quantitative calculations and simulation of dislocation images we

need to go to the generalised form of the dynamical theory developed by Takagi.
21
In the distorted crystal the wavevector K′ g differs from the ideal crystal
wavevector by ′ g where
′ g=−grad( g.u ) (8.33)
and u is the atomic displacement. The λ λ g parameter then is a function of
position in the crystal. The electric displacement in the crystal has a form
(8.34)
Takagi’s equations are then
(8.35)
These may be combined to give two equations of the form
(8.36)
These have hyperbolic form, and in a perfect crystal can be solved analytically.
In a distorted crystal they must be solved by numerical integration. Authier et al .
30 were the first to devise an algorithm for solution over a grid of points. If, in
Figure 8.12, M is the point at which we require the wave amplitude and P and Q
are neighbouring points at which the amplitudes are known,
(8.37)
where p = PM and q=QM as defined in Figure 8.12. It is therefore

straightforward, to derive the amplitudes at any point in the crystal by a step-by-
step calculation.
Although the simulation is substantially slower than from the equivalent
simulations of images in electron microscopy, section topographs can now be
simulated in a matter of seconds on a fast personal computer. Just as rocking
curve simulation can be used as a means of determining the microscopic strains
from the scattered intensity, so simulation of topograph images can be used to
determine the microscopic strain field around a defect and hence identify it. The
match between simulation and experiment is extremely good, although in the
section topograph the position of the defect does influence strongly the contrast
(Figure 8.13). In all cases a very simple strain field of the form
u ( r )=C r / r 3 (8.38)
has been used to model the strain field around a precipitate. This radial strain
field, based on isotropic elasticity theory is found to work very well for
precipitates in silicon.
Figure 8.12 Grid filling the inverted Borrmann fan over which numerical integration of
Takagi’s equations can be accomplished
Epelboin 31 has given an excellent review of the procedures for calculating
section topograph images and has more recently described a variable step
algorithm which enables the grid size to be adjusted as the deformation increases
in the vicinity of the defect. 32 A similar approach has been used in the
simulation of the Bragg case double-crystal topographs by Spirkl et al . 33 More
recently, Epelboin has given algorithms for parallel computation. 34 The
diffraction process takes place solely in the incidence plane and therefore very
rapid calculation can be achieved on parallel architecture machines by treating
each slice independently and in parallel.
The wave theory equations essentially have the form of a multiple scattering
theory based on plane waves. In the limit of small Bragg angles Takagi’s
equations reduce to the Howie-Whelan equations, well known to electron
microscopists.
Figure 8.13 Experimental and simulated section topography images of hydrogen

precipitates in silicon. All simulations were carried out using the same microscopic strain
field, but the different position of the defects within the crystal results in substantially
different image detail
8.9
Dislocation and stacking fault images in X-ray topographs
The contrast of defects in X-ray topographs has been recently reviewed by one
of the authors 35 to which the reader is referred f or further details and a
bibliography. In this section we will confine the discussion to dislocation and
stacking fault images, these being the most common examples of linear and two-
dimensional defects.
8.9.1
Images of dislocations
Dislocation images in single-crystal X-ray topographs differ from those in
electron micrographs primarily because the incident beam cannot be regarded as
a plane wave. Of the three types of image characterised by Authier, the most
common one seen in projection or white radiation synchrotron topographs is the
direct image. This appears dark against the perfect crystal background on the
topograph (Figure 8.14). The origin of this enhanced intensity lies in X-rays
which are outside the range of diffraction of the perfect crystal. Although not
diffracted by the perfect crystal, they will be in the deformed region around the
dislocation. They thus suffer no primary extinction and the so-called direct image
appears as an intense dark spot on the section topograph. In thick crystals the
rays far from the Bragg angle contributing to the direct image suffer ordinary
absorption and hence the direct image does not appear.
We can begin to understand the contrast mechanism by considering the crystal

to be made up of three distinct regions, namely the perfect crystal above the
defect, the perfect crystal below the defect, and the deformed region around the
defect (Figure 8.15(a)). We set the limit of the deformed region as that where the
effective misorientation λ (■ λ ) around a defect exceeds the perfect crystal
reflecting range:
λ (■ λ )=−(ksin2λ B )−1 λ (g.u)/λ s g (8.39)
The region around the defect is assumed to diffract as a small mosaic crystal of
thickness A, the X-rays diffracted from this region not being diffracted by the
perfect crystal. Any dynamical diffraction effects associated with wavefield
matching above and below the defect are neglected in this analysis. The
integrated intensity for the Laue case (Figure 8.15(b)) varies linearly with
thickness for small values of thickness, but as the thickness increases, so the
gradient decreases and, in the absence of absorption, oscillates about a constant
value. [When absorption is included, there is superimposed a gradual decay of
intensity with thickness.] As the gradient of the I versus t curve is everywhere
less than that at the origin, the intensity I due to material thickness t is always
less than the sum of I ■ and I t −■ from separate regions of thickness ■ and t−■ .
In Figure 8.15(b), where t/λ g■ 3, corresponding to the third thickness minimum,
it is very obvious that
I t <I ■ +I t −■ (8.40)
There is thus always enhanced intensity around the defect, the contrast being a
maximum when the thickness correponds to the first thickness minimum, t/λ g■ 0.
88. For the Bragg case and zero absorption (Figure 8.15(c)) the intensity is
independent of thickness beyond a crystal thickness of about an extinction
distance λ g. Once again we see that for crystal thickness greater than this value,
equation (8.40) is valid and the defect shows enhanced intensity. In crystals less
than an extinction distance in thickness, equation (8.40) becomes an equality,
and defect images are not observed. This has important implications for studies of
proteins where the extinction distances can be very long. The danger is that
single-crystal topographs can appear perfect when in reality they may contain
defects.
The width of the image can be deduced using this simple idea of contrast being
formed when the misorientation around the defect exceeds the perfect crystal
reflecting range λ λ . We consider the case of a screw dislocation running normal
to the Bragg planes, where the line direction l coincides with the diffraction
vector g . The effective misorientation λ λ at distance r from the core is
λ λ =b/2λ r (8.41)
The width of the dislocation image D, which is twice the value of r for which λ λ
= λ λ and is thus
D=b/(λ λ λ ) (8.42)
As, for a symmetric reflection, λ λ =2/g λ g , we have D=g bλ g /2λ . This result
can be generalised to:
Figure 8.14 Images of dislocations in silicon under conditions of low absorption. AgK λ
radiation. The dark images are the direct images. The white images are dynamical images
for screw dislocations D■ g.b λ g /2λ (8.43a)
for edge dislocations D■ g.b λ g /λ (8.43b)
Direct images of edge dislocations are approximately twice as wide as those for
screw dislocations of the same g . b value and this provides a useful method
of checking Burgers vector assignment. As the extinction distance is on the scale
of micrometres, dislocation images widths are of that order. The fact that this is
several orders of magnitude greater than those of transmission electron
micrographs is a fundamental limitation, set by the weak scattering, and hence
high strain sensitivity, of X-rays in crystals. In terms of the structure factor F g
have
D■ constant.[λ F g ]−1 (8.44)
and the dislocation image width goes down for increasing wavelength A and
strength of the reflection.
Equations (8.43) suggest that the image width is zero when g.b = 0. This is
of course the classic criterion, originally applied to transmission electron
Figure 8.15 (a) Diagram showing the splitting of the crystal into perfect material above
and below the defect and distorted material close to it. (b) Integrated intensity—Laue case
(c) Integrated intensity—Bragg case
microscopy, where the effective misorientation of the distortion around the

dislocation is zero. As the Bragg planes are not distorted or tilted the dislocation
is invisible in that reflection. Strictly, this criterion is that both g.b = 0 and
g.bxl = 0 and, except when the dislocation runs parallel to a high symmetry
axis, it is valid only for isotropic elasticity. Nevertheless, the contrast is often
weak for just g.b = 0 and this enables b to be determined on purely geometric
grounds by finding two reflections in which the dislocation is invisible or very
weak.
Miltat and Bowen 36 showed that direct images can be synthesised from the
cylinders of misorientation drawn around a dislocation line using continuum
elasticity theory. The image full width can be calculated from the projected width
circumscribed by the contour where λ (■ λ ) is equal to ■ times the reflecting range,
with ■ ■ 1. The intensity may be calculated from the volume of mosaic crystal
contained within the cylinder. Unfortunately, the value of ■ is a function of the
reflection used and the procedure cannot be used quantitatively to simulate the
direct image. It does, however, provide a very rapid method for defect
identification in association with other information. (Satisfactory simulation of
the direct image has not yet been achieved, even in the case of the section
topograph.) The model explains the significance of the bimodal profile of
dislocations when g.b>2 as here lobes of opposite misorientation on either side
of the defect give rise to two (diverging or converging) components of the
image. Such double images are readily seen at synchrotron radiation sources
where the divergence effect is enhanced by the large specimen-to-plate distance.
An important feature of the direct image in the section topograph
(Figure 8.16) is that the position of the direct image corresponds to the depth of
the defect within the crystal. In the section topograph it is localised
(Figure 8.16), unlike the dynamic and intermediary images which in the example
here are spread out in a ‘banana’ shape. The localisation of the direct image has
practical application in the control of intrinsic gettering of impurities in silicon.
When this process is undertaken, a high density of precipitates is formed in the
centre of the wafer, but a highly perfect region arises at the surface. As modern
devices are shallow, this is a highly satisfactory production procedure. Under
conditions of low absorption, section topography can be used non-destructively
to monitor the width of these denuded zones at the surface, as the position of the
direct image across the section topograph corresponds directly to the depth of the
precipitate in the crystal.
Dynamical images, although seen in thin crystals when the dislocation lies
close to the entrance surface and normal to the incidence plane (Figure 8.14), are
generally important in thick or high atomic number crystals where the product μ t
is significantly greater than unity (Figure 8.17). They are less intense than the
background and appear white on the plates. In thick crystals where only the
branch 2 wave at the centre of the Borrmann fan remains, the presence of the
dislocation causes scattering of intensity out of that wavefield into newly created
wavefields. These suffer higher absorption and thus the dislocation casts a
Figure 8.16 Section topograph of a dislocation in silicon. The dislocation cuts the
Borrmann fan obliquely, resulting in displacement of the direct, intermediary and
dynamical images. We note that the dynamical image, unlike the direct image, is not
localised. 1, direct image; 2, intermediary image; 3, dynamical image 37
shadow in both diffracted and transmitted beams. In general the dynamical image
is diffuse except where the dislocation is close to the exit surface of the crystal.
This broadening of the image with distance from the exit surface can be clearly
seen in Figure 8.11(b). Under extremely high absorption conditions, the
dynamical images sharpen, as only those wavefields propagating parallel to the
Bragg planes reach the exit surface. The intensity loss is then a sharp projection
of the deformed region around the defect.
In some cases interference can be observed between the newly created
wavefields leading to oscillatory contrast in the region of the dislocation close to
the exit surface (the intermediary image). Only when the extinction distance is
short are such effects usually observed, as anomalous absorption quickly damps
out the branch 2 waves within a short distance from the exit surface. Where they
intersect the exit surface, dislocations often exhibit a black-white contrast from
the surface relaxation which can be interpreted using the Penning-Polder theory
along the lines discussed in section 8.8.1. Owing to divergence of the wavefields
in the crystals, dynamical images have best contrast when the dislocation is close
to the exit surface.
Intermediary images are usually blurred out by the traversing but often
significantly affect the shape of the direct image leading to asymmetry of the
profile. In suitable circumstances the intermediary image is well separated and,
as in the section topograph, shows oscillatory contrast. The splendid
intermediary images of Authier 23 are rare, and more often the overlap of direct
and intermediary images leads to oscillatory contrast of an inclined dislocation.
32
8.9.2
Contrast of stacking faults in section topographs
Stacking fault contrast in section topographs can be most easily understood by
expanding the incident spherical wave as a Fourier expansion of plane waves
Figure 8.17 Dynamical images of dislocations under conditions of high absorption in

GaAs. Note that the images appear light due to the loss of intensity around the defect
using the method of Kato. It is also possible to use Takagi’s theory to calculate
the contrast, but the physical insight is lost.
We consider in Figure 8.18(a), a stacking fault inclined to the crystal faces. In
the plane of incidence, the fault cuts the extrema of the Borrmann fan at P and Q.
Above the fault two wavefields corresponding to the tie-points O1 and O2 in
Figure 8.18(b) propagate in the direction OR. On crossing the fault they decouple
into their plane wave components, but on re-entering the perfect crystal below
the fault they immediately excite new wavefields. The boundary conditions can
be applied to the plane wave components in a similar manner to those at a
vacuum interf ace provided that the phase difference introduced by the fault is
included. We see from Figure 8.18(b) that four waves now exist in the crystal.
Two of these correspond to the original tie-points O1 and O2 and two to newly
created Bloch wavefields corresponding to the newly excited tie-points N1 and N2.
From the symmetry implied by the continuity of the normal component of
Figure 8.18 (a) Ray paths of original and new wavefields in a crystal containing an
inclined stacking fault. (b) Tie-points O1 and O2 corresponding to original wavefields. Tie-
points N1 and N2 corresponding to new wavefields. (c) Surfaces of constant phase. (d)
Section topographs of an inclined stacking fault in diamond, (left) AgK λ radiation, (right)
CuK λ radiation showing the periodic fading due to the two polarisation states. 111
reflection. Image height 0.85 mm. (Courtesy S.S.Jiang, University of Nanjing)
Figure 8.18 (cont.)
wavevector, we see that the direction of propagation of the new wavefields N1

and N2 is the same. At a point S on the exit surface the intensity is made up of
contributions from the newly created wavefields travelling in the direction OS.
Explicitly, the intensity comprises three components, I 1, I 2 and I 3 where I 1
represents the interference between the original wavefields which gives rise to
the section topograph thickness (or Kato) fringes, I 2 corresponds to the
interference between the new wavefields and I 3 gives the contribution due to
interference between original and new wavefields. In the case of a highly
absorbing crystal, the expression for I 3 dominates as the contrast arises from the
interference of waves associated with tie-points O1 and N1. These are both on
branch 1 and are thus only weakly absorbed.
In the region OPR interference of the original wavefields occurs and the fronts
of equal phase, shown in Figure 8.18(c), are hyperbolic. In the triangle QPF the
newly created wavefields propagate and the phase fronts are also hyperbolae,
equivalent to the phase fronts in OPQ reflected about PQ. In addition, the flat,
non-hyperbolic phase fronts due to the interference of the two systems occur in
QPF. These are shown as the additional contour in Figure 8.18(c). The newly
created wavefields converge to a focus F and we note that F=OP and PF=OQ. As
the line of intersection PQ moves through the crystal, so the position of the focus
F varies. In the particular case where PQ is parallel to the crystal surface, it is
easy to see that when PQ lies exactly half-way through the crystal, F is at the exit
surface. Quite generally there is a position where no fringes from I 2 and I 3 are
observed and the resulting section topograph from an inclined fault has the
delightful ‘hour-glass’ shape illustrated in Figure 8.18(d). The flat fringes due to
interference between original and new wavefields have twice the periodicity of
the hyperbolic fringes.
In non-absorbing crystals, the hour-glass contains two sets of fringes, one of
each periodicity, but in thicker crystals the I 2 term becomes vanishingly small.
The contrast of the I 3 fringes is always higher than that of the I 2 fringes, and it
is these non-hyperbolic fringes which are usually observed experimentally. A
total of t/λ g flat fringes is found. Fringe fading arises from beating between the
two states of polarisation, an effect not seen with synchrotron radiation. 38
Under the classification of Amelinckx, stacking faults give rise to ■ fringes.
The structure in the region below the fault is derived from that above by a simple
translation r parallel to the fault. In contrast, across a twin boundary the region
below the fault is related by a similar displacement vector r which increases
linearly from the boundary and the fringes thus produced are known as λ fringes.
The term I 3 is sensitive to the sense of the phase difference, and examination
of the contrast of the first fringe from the exit surface enables the sense of λ and
hence the nature of the fault, whether extrinsic or intrinsic, to be determined. The
magnitude of r can be determined from the reflections in which the fault is
invisible. This occurs whenever g.r=0 or n where n is integral. 39 , 40
Fringes are seen in traverse topograph images of stacking faults, but in general
the contrast is weaker. Other planar defects such as magnetic domain boundaries,
ferroelectric and ferroelastic domain boundaries, growth sector boundaries and
growth bands themselves all give rise to images in X-ray topographs. The
contrast may arise either from the elastic strains at the interface or the phase
difference introduced into the wavefields on crossing the boundary. In many
cases the contrast is complex and may be difficult to simulate.
The contrast of growth bands, which arise from changes in lattice parameter
due to fluctuations in impurity levels, is particularly valuable in determining the
growth history of crystals. No contrast is observed when g.n=0, n being a vector
normal to the growth front.
8.10
Summary
X-ray topography is the X-ray analogue of transmission electron microscopy and
as such provides a map of the strain distribution in a crystal. The theory of image
formation is well established and image simulation is thus a powerful means of
defect identification. Despite a reputation for being a slow and exacting
technique, with modern detector technology and care to match spatial resolution
of detector and experiment, it can be a powerful and economical quality-control
tool for the semiconductor industry.
References
. L.G.SCHULZ, Trans. A.I.M.E., 200 , 1082 (1954).

. A.GUINIER & J.TENNEVIN, Acta Cryst, 2 , 133 (1949).
. A.R.LANG, J.Appl. Phys., 29 , 597 (1958).
. U.BONSE & E.KAPPLER, Z.Naturforschung, 13a , 348 (1958).
. C.S.BARRETT, Trans. A.I.M.E., 161 , 15 (1945).
. W.F.BERG, Naturwissenschaften, 19 , 391 (1931).
. A.R.LANG, Acta Cryst., 12 , 249 (1959).
. B.K.TANNER, X-ray diffraction topography (Pergamon, Oxford, 1976).
. D.K.BOWEN & C.R.HALL, Microscopy of materials (Macmillan, London, 1975).
. A.R.LANG, in: Modern diffraction and imaging techniques in materials science, eds. S.
AMELINCKX, R.GEVERS & J.VAN LANDUYT, Vol. 2 (North-Holland,
Amsterdam, 1970) p. 623 .
. J.-I.CHIKAWA, I.FUJIMOTO & T.ABE, Appl. Phys. Lett., 21 , 295 (1972).
. S.SUZUKI, M.ANDO, K.HAYAKAWA, O.NITTONO, H.HAHIZUME, S.KISHINO &
K. KOHRA, Nucl. Inst. Meth., 227 , 584 (1984).
. W.HARTMANN, in: X-ray optics, Topics in Applied Physics, Vol. 22 , ed. H.-
J.QUEISSER (Springer-Verlag, Berlin, 1977) p. 191.
. N.KATO, in: Crystallography and crystal perfection, ed. Ramachandran (1963), p. 153.
. N.KATO, in: Introduction to X-ray diffraction, eds. L.G.AZAROFF, R.KAPLOW, N.
KATO, R.J.WEISS, A.J.C.WILSON & R.A.YOUNG (McGraw Hill, New York,
1974) p. 425 .
. A.AUTHIER, Phys. Stat. Sol., 27 , 77 (1968).
. T.TUOMI, M.TILLI & O.ANTTILA, J. Appl. Phys., 57 , 1384 (1985).
. A.J.HOLLAND & B.K.TANNER, J. Phys. D: Appl. Phys., 28 , A27 (1995).
. J.R.PATEL, J. Appl. Phys., 44 , 3903 (1973).
. M.HART & A.D.MILNE, Acta Cryst., A25 , 134 (1969).
. N.KATO, Acta Cryst., A25 , 119 (1969).
. P.ALDRED & M.HART, Proc. R. Soc. Lond., A322 , 233 and 239 (1973).
. N.KATO, Acta Cryst., 13 , 349 (1960).
. S.TAKAGI, J. Phys. Soc. Japan, 26 , 1239 (1969).
. A.AUTHIER, Adv. X-ray Analysis, 10 , 9 (1967).
. N.KATO, J. Phys. Soc. Japan, 18 , 1785 (1963).
. P.PENNING & D.POLDER, Philips Res. Reports, 16 , 419 (1961).
. M.HART & A.D.MILNE, Acta Cryst., A31 , 425 (1971).
. N.KATO, J. Phys. Soc. Japan, 18 , 1785 (1963).
. A.AUTHIER, C.MALGRANGE & M.TOURNARIE, Acta Cryst., A24 , 126 (1968).
. Y.EPELBOIN, Mater. Sci. Eng., 73 , 1 (1985).
. C.A.M.CARVALHO & Y.EPELBOIN, Acta Cryst., A49 , 467 (1993).
. W.SPIRKL, B.K.TANNER, C.R.WHITEHOUSE, S.J.BARNETT, A.G.CULLIS, A.D.
JOHNSON, A.KEIR, B.USHER, G.F.CLARK, W.HAGSTON, C.R.HOGG &
B.LUNN, Phil. Mag. A, 70 , 531 (1994).
. Y.EPELBOIN, J. Appl. Cryst., 29 , 331 (1996).
. B.K.TANNER, in: X-ray and neutron dynamical diffraction: theory and applications,
eds. A.AUTHIER, S.LAGOMARSINO & B.K.TANNER (Plenum Press, New York,
1996).
. J.MILTAT & D.K.BOWEN, J. Appl. Cryst, 8 , 657 (1975).

. G.S.GREEN, N.LOXLEY & B.K.TANNER, J. Appl. Cryst., 24 , 304 (1991).
. G.KOWALSKI, A.R.LANG, A.P.W.MAKEPEACE & M.MOORE, J. Appl. Cryst, 22 ,
410 (1989).
. A.AUTHIER, Phys. Stat. Sol., 27 , 77 (1968).
. A.AUTHIER & J.R.PATEL, Phys. Stat. Sol (a), 27 , 213 (1975).
218
9
Double-crystal X-ray Topography
In this chapter we discuss double-crystal topography, in which we

obtain a map of the diffracting power of a crystal compared to that of
a reference. We first treat the principles and geometries, the
mechanisms of image contrast and resolution and the use of
laboratory and synchrotron radiation. We then discuss applications:
wafer inspection, strain contour mapping, topography of curved
crystals.
9.1
Introduction
As we saw in the last chapter, an X-ray topograph is a two-dimensional map of
the scattered intensity distribution across a diffracted beam. Thus simply by
placing a piece of photographic film or a direct viewing TV detector in the
diffracted beam from the specimen in a high resolution diffraction experiment,
one may take a simple X-ray topograph directly. For many purposes this simple
procedure is very informative. For example, the presence of strong bands of
intensity running at right-angles in a topograph taken from a (001)-oriented
epitaxial layer immediately shows the presence of relaxation (Figure 9.1). In
many cases this proves to be much easier than analysing symmetric and
asymmetric reflections to determine the relaxation quantitatively. At this level of
resolution, fast film, image plates or a direct read-out TV detector is satisfactory.
However, if high spatial resolution or high strain sensitivity is required, rather
more care must be taken in the design of the experiment in order for these
parameters to be optimised. In this chapter we discuss the requirements for
taking high quality double-crystal topographs and also the interpretation of the
images obtained.
Figure 9.1 Double-crystal X-ray topograph of a heavily relaxed (001)-oriented GaAsSb

layer on GaAs showing the characteristic ‘tweed’ contrast from two sets of misfit
dislocations running parallel to the (110) directions
9.2
Principles and experimental geometries
Double-crystal topography was first performed in the late 1950s 1 , 2 but never
then reached the level of usage that the single-crystal techniques achieved. The
simplest experimental scheme to understand is that of the (+ n, − n ) non-
dispersive setting shown in Figure 9.2(a). Although the wavelength diffracted
from different points across the crystal may vary, the beam emerging from the
reference crystal is of very small angular divergence. We have seen in Chapter 2
that a very small rotation of the specimen is required to go right through the
rocking curve. Thus any local distortion of the specimen around a defect will
result in loss of intensity and an image of the defect will be formed on the
photographic plate. As the rocking curve widths have been shown to be very
narrow, this implies a very high strain sensitivity of the (+ n, − n ) non-
dispersive setting. Two important things should be noted about this setting. If a
point source is used with a conventional generator, two bands of intensity are
observed across the topograph, corresponding to the K λ 1 and K λ 2
DOUBLE-CRYSTAL X-RAY TOPOGRAPHY 221
Figure 9.2 (a) Simple (+n, −n) double-crystal topography arrangement in the parallel
setting. (b) Use of asymmetric reflection to expand illuminated area and eliminate image
doubling. (c) Double-crystal topography with a standard duMond-Hart monochromator
system
wavelengths. This may be adequate, but does not give a large area topograph
with uniform illumination. If a line source is used, with the line focus being in
the dispersion plane, then a uniform intensity will be recorded across a strip
correspond ing to the width of the source. However, because the K λ 1 and K λ 2
rays make different directions in space, double images will be observed of any
defect. This may not be a problem in principle, but it does confuse interpretation.
To avoid this problem, an asymmetric reflection is often used from the
specimen as shown in Figure 9.2(b). This permits the K λ 2 beam to be removed
by a slit between the specimen and the reference crystal, but still enables a large
area of specimen to be imaged. This arrangement is also appropriate when a
duMond-style beam conditioning monochromator is used. Convenient reflections
with a high asymmetry factor for CuK λ radiation are the 224 and 044 for GaAs
and InP and 113 for Si. The advantage of this arrangement is that high resolution
topography can be performed without modification of the standard high
resolution diffraction experimental arrangement (Figure 9.2(c)). Note that by
using an asymmetric reflection to expand the beam from the reference crystal as
well as the specimen, an expansion of up to several hundred times may be
obtained. This provides a means of taking double-crystal topographs of very
large area wafers. Hung Liu at Texas Instruments in Dallas has recorded full
topographs of 5■ Si wafers using this method.
An alternative method for removing the double images is to use the dispersive
setting, where the first and second crystals are not of the same material. For
example, using the 004 reflection from a Si reference crystal with the 004 InP
from the specimen will provide enough dispersion to diffract from only the K λ 1
line. This inclusion of dispersion also reduces the extent of the defect image, as
it reduces the sensitivity to strain.
9.3
Defect contrast in double-crystal topographs
Most of the features observed in double-crystal topographs can be interpreted by
means of a simple geometrical theory first given by Bonse 3 in 1962. Let us
assume that the shape of the rocking curve is triangular and that the specimen is
oriented such that it is set on the flank of the rocking curve. Then it is very easy
to see that the fractional change in intensity λ I / I is given by
(9.1)
where K is the slope of the rocking curve, λ d / d is the fractional change in Bragg
plane spacing, λ λ is the tilt around the defect and n t and n g are unit vectors
normal to the incidence plane and parallel to the defect misorientation axis
respectively. The second term on the right-hand side is in fact the component of
tilt in the dispersion plane. The sense of the contrast change depends on the angle
that the specimen makes with the reference crystal. When the specimen is set so
the rocking curve slope is positive (position (A) in Figure 9.3) a positive dilation
results in increased intensity. When the slope K is negative (position (B) in
Figure 9.3), then there is a reduced intensity for a positive dilation. When the
specimen is set at the top of the rocking curve, the intensity is always reduced
irrespective of the sense of dilation or tilt. Figure 9.4 shows examples of
Figure 9.3 Schematic diagram of the positions of positive/negative (A) and negative/
positive (B) contrast on either flank of the rocking curve. At the rocking curve maximum
(C), the defect always results in an intensity loss, i.e. negative contrast
topographs of dislocations in InP set at points A and B respectively on the
rocking curve. The complementary contrast of the defects in the two topographs
is striking. The long sharp images of scratches, and that of the scratch A do not
exactly reverse contrast, indicating that the strains here are so large as to lead to
departure from the roughly linear region on the rocking curve flank.
A full simulation of the defect image may be achieved by solving numerically
Takagi’s equations. This was first done for the double-crystal arrangement in the
Bragg case by Riglet et al . 5 and more recently by Spirkl et al . 6 In most
cases, the full dynamical theory simulation predicts a contrast reversal in the same
way as the simple geometrical theory. However, the full dynamical theory does
show that there are situations where, as with the experiment, the contrast does not
simply reverse on going across the rocking curve. Figure 9.5 shows one such
example. This simulated ‘rocking topograph’ is a simulation of the image of a
misfit dislocation parallel to the surface of the crystal, with the offset from the
Bragg position varying along the dislocation line length. As we pass from the
bottom to the top of the image, we pass through the rocking curve. Physically
this corresponds to a twisted sample. 7 We note the reversal of contrast from the
top to the bottom of the image. The case simulated is for misfit dislocations at
the interface between a GaAs substrate and an InGaAs epilayer, but with a thick
GaAs capping layer. This capping layer results in interference contrast which
results in complex contrast at the dislocation. While the geometrical theory
Figure 9.4 Experimental Bragg case topographs taken from a bulk InP on either side of
the rocking curve: (a) low angle, (b) high angle 4
provides a valuable first step, for detailed interpretation of the microscopic strain
field, full dynamical theory simulations are necessary.
9.4
Spatial resolution
The geometrical resolution normal to this dispersion plane in a double-crystal
topograph is given in exactly the same way as for a single-crystal topograph. As
for the single-crystal topograph, the spatial resolution λ is approximately given
by
(9.2)
where h is the source size, b is the specimen to film distance and a is the source
to specimen distance. For high spatial resolution, therefore, the film or detector
must be placed as close as possible to the specimen. In the laboratory
environment this effectively limits the spatial resolution to a few micrometres.
However, this is normally perfectly adequate because the high strain sensitivity
of the double-crystal techniques leads to very wide images from isolated defects.
By reference to equation (9.1), it can be seen that when the rocking curve width
becomes very narrow, the rocking curve gradient becomes very high. A large
contrast change is then observed for a small misorientation or dilation. As the
strain field around a dislocation falls off inversely with distance, the higher the
strain sensitivity (larger value of K) the further is the distance away from the
defect at which significant contrast is formed. In highly perfect crystals of Si
Figure 9.5 Simulated ‘rocking topograph’ of misfit dislocations parallel to the surface of
a (001)-oriented GaAs wafer. Dislocation line runs up the page
containing a very small number of dislocations, the long-range nature of the

strain field has been revealed, showing that the strain field extends throughout
the whole specimen.
As seen in the last chapter, the image width is easy to quantify for a screw
dislocation, where the diffraction vector is parallel to the dislocation line. Around
a screw dislocation, the misorientation λ λ at a distance r from the core is given
by
(9.3)
The width of the image D=2r 0 where r 0 is the value of r corresponding to a
misorientation equivalent to the full width at half height maximum of the rocking
curve λ λ . Thus, again,
D={b/2λ }/λ λ (9.4)
We note that if the rocking curve is approximated to a triangle, i.e. it has a
constant slope, equation (9.1) can be used to calculate the intensity of the
dislocation image in the topograph. As for the (+ n, − n ) case,
λ λ =const.λ 2 F h /sin2λ B (9.5)
for small Bragg angles,
D■ const.[λ Fh]−1
(9.6)
Thus as for the single-crystal topograph, for a given reflection, the dislocation
image width goes up with falling wavelength and for a given wavelength, the
dislocation image width goes up with falling structure factor. Images using quasi-
forbidden or high-order reflections are therefore intrinsically wide. This
discussion shows that the higher the intrinsic sensitivity to strain, the wider are
the images of defects.
The above discussion assumed that the reference and specimen crystals were of
the same material and in the same diffraction conditions, i.e. the (+n,−n)
geometry. If the reference crystal is different from the specimen, then the
distance from the specimen at which the image forms changes. If the reflecting
range of the exit beam from the reference crystal is very large compared with the
incident reflecting range of the specimen, then the image forms at a distance
determined by the incident reflecting range width of the specimen. If, on the
other hand, the reflecting range of the exit beam of the reference crystal is small
compared with the incident reflecting range of the specimen, it is the reflecting
range of the reference crystal which determines the image width. Thus the
images of misfit dislocations observed in the 224 reflection from a (001)-oriented
InGaAs layer on GaAs are wider and sharper if the 333 reflection is used for the
reflection from a silicon reference crystal rather than the 111 reflection. In the
former case, the lower divergence results in the incident beam not satisfying the
Bragg condition over a greater volume of crystal around the defect and giving a
better signal to noise in the image.
Use of grazing incidence has a beneficial effect on the image contrast, as well
as having the effect of expanding the beam area. If the reference crystal has a
grazing incidence beam, the exit beam divergence is reduced by the square root
of the asymmetry factor. The volume of ‘defective’ material around individual
defects is therefore enhanced over the symmetric geometry. In grazing incidence
on the specimen, both the absorption depth and the extinction length become
small. Thus, both within and away from the Bragg reflecting range the X-ray
wave does not penetrate deep into the crystal. The depth of ‘defective’ material
around the defect becomes comparable with the depth of penetration of the X-ray
wave and almost 100% contrast can be obtained from the defect. A browse
through the literature will show that this argument is confirmed by experiment
and that the asymmetric settings where the reference and the specimen are both
in grazing incidence settings give the best defect contrast.
9.5
Synchrotron radiation double-crystal topography
In the laboratory, high spatial resolution, double-crystal topographs can often
take several days to expose. The intensity available at synchrotron radiation
sources makes the use of this radiation very attractive and exposures at second-
generation sources on L4 nuclear emulsions are typically about 30 minutes. This

reduces further for the third-generation sources. There are a few features of the
synchrotron radiation experiment which differ from that in the laboratory.
1 The very small source size and the large distance of the source from the
specimen means that any point on the specimen sees a beam of divergence
less than one or two arc seconds. Therefore there is very little difference
between the images in the (+n,−n), (+n, +n) and (+n, ±m) settings. Very
little dispersion is introduced by mismatching crystals and so highly perfect
silicon is recommended as a reference crystal.
2 Because both reference and specimen are a long way from the source, the
geometric resolution is good for both crystals. Thus, any defects in the
reference crystal are imaged in sharp contrast, which is not the case in the
laboratory. This provides a second reason for using silicon as a beam
conditioner.
9.6
Applications of double-axis topography
Applications of double-crystal topography with synchrotron radiation are dealt
with in the next chapter. Here, we confine ourselves to laboratory-based
applications.
9.6.1
Incoming wafer inspection
Double-crystal topography has been successfully applied by Bassignana et al. 8
to the study of incoming compound semiconductor wafers prior to growth of
epitaxial layers (Figure 9.6). Asymmetric reflections are used for both reference
and specimen crystals so that the beam expansion is about 20 times and the beam
divergence at the specimen is about 1.3 arc seconds. This provides a highly
sensitive probe of long range strain. Relatively low resolution, and hence fast,
Agfa D7 X-ray film has been used to keep exposures in the region of 20–40
minutes. This was so successful that it was incorporated as part of the routine
quality assessment of GaAs wafers purchased by Bell Northern Research and has
had significant impact on device yields. 9 Wafers from boules which had a high
density of clustered low-angle boundaries were found to be mechanically too
fragile to survive all the required steps in device processing. Wafers from boules
which showed large included mosaic crystals showed poor or erratic cleaving
characteristics, which resulted in the loss of many devices.
Figure 9.6 422 reflection double-crystal (asymmetric reflection) reflection topograph of

(a) low-angle boundaries in a 3 inch semi-insulating liquid encapsulated (LEC) GaAs
wafer; (b) dislocation cell structure in a vertical gradient freeze GaAs wafer. (Courtesy Dr
I.C. Bassignana, Nortel Inc, Ottawa)
9.6.2
Strain contour mapping
If the specimen wafer is curved, then because of the low angular divergence of
the beam, only part of the wafer will be imaged. The incidence angle across the
specimen crystal varies with position and a rocking curve is imaged! Figure 9.7
shows a fine example of the image of a curved crystal of Ga As with several
AlGaAs epitaxial layers of different composition grown on it. In addition to the
peak from the GaAs substrate and that from a thick AlGaAs buffer, a series of
interference fringes can be seen across the sample. These match extremely well
with the rocking curve taken from a single point under angle scanning. 10
If there is only the substrate, the image is a single stripe, tracing out the
contour of equal effective misorientation. Thus a series of images with different
angular separation between reference and specimen crystal provides a ‘zebra
pattern’ of stripes corresponding to different effective misorientation contours.
Figure 9.8 shows such a series of contours from a SI LEC GaAs specimen. By
rotation of the sample about the surface normal and recording a second series of
topographs, the effects of tilts and dilations can be distinguished. Tilts will appear
as contour displacements in opposite senses in the two topographs, in which the
effect of dilations is the same. Brown et al . 11 and Minato et al . 12 have given
some fine examples of use of zebra stripes for the analysis of long-range strain.
Figure 9.6 (cont.)

9.6.3
Curved crystal reference crystal topography
The problem of specimen curvature can, as we have seen, be used to advantage,
but in many cases it is extremely inconvenient. For low curvature, sharp contour
lines do not appear, but only part of the sample is imaged. In order to obtain an
image of the whole sample several topographs have to be taken at different
incidence angles, multiplying the experiment length by several times (as many
times as images are necessary). Jenichen, Köhler and colleagues in Berlin have
overcome this by the clever expedient of curving the reference crystal to match
the curvature of the sample. 14 , 15 The dramatic improvement can be seen
between the flat reference crystal topograph of Figure 9.9(a) and that when the
reference crystal is curved (Figure 9.9(b)).
The minimum radius of curvature is set at a few metres from a combination of
geometrical considerations such as primary beam divergence and breakdown of
the dynamical diffraction conditions. This does not turn out to be a severe
limitation in practice. Highly mismatched or very thick near-matched epitaxial
layers turn out to have quite uniform bending which matches well with the
mechanical bending deliberately induced on the reference crystal. In the
asymmetric geometry, the contrast is high due to the small penetration depth, and
excellent quality topographs have been reported. Figure 9.10 shows an example
of threading and misfit dislocations in the topograph shown in Figure 9.9(b) but
at greater magnification. Note that, when taking topographs of misfit
dislocations, it is usually better to use the substrate, rather than the layer
Figure 9.7 Double-crystal topograph taken of a curved GaAs wafer with several AIGaAs
epitaxial layers, showing the effective rocking curve spread out in the image 10
Figure 9.8 Composite image of a series of double-crystal topographs recorded at different

incidence angles of a Sl LEC GaAs sample. The image is a set of contours of equal
effective misorientation. (Courtesy S.J.Barnett) 13
reflection. For very thin (typically 20 nm) layers of InGaAs on GaAs this is
essential, as the intensity in the layer peak is several orders of magnitude less
than that of the substrate and exposure times become prohibitive. For thick
layers, the uniformity of the substrate images is usually better. The strain fields of
the misfit dislocations are long range and, as the elastic properties of the layer
and substrate are very similar, the images appear sharp and without loss of
contrast.
Figure 9.9 Double-crystal topographs with Ge reference crystal, asymmetric 206

reflection, glancing angle 1.7° Cu K λ 1 radiation. Sample is 2 μ m Al0.05Ga0.95As, 2 μ m
Al0.35Ga0.65As, 1 μ m GaAs on (001) GaAs substrate 206 reflection. (a) Flat reference
crystal. (b) Reference crystal curved to 12 m radius of curvature. (Courtesy of Dr
R.Köhler) 16
Figure 9.10 Magnified section of the topograph of Figure 9.9(b) showing threading and
misfit dislocations. (Courtesy Dr R.Köhler) 16
9.7
Summary
Double-axis topography can be used at the two extremes of crystal perfection.
On the one hand it can be used to probe the long-range lattice strains in very
highly perfect crystals. On the other, it can be used to reveal contours of equal
misorientation in much less perfect materials. As we will see in Chapter 10, it is
widely used at synchrotron radiation sources and with the very large power
loading in the white beam of third-generation synchrotron radiation sources. It is
rapidly becoming the standard technique of topography for most electronic

materials.
References
. W.L.BOND & J.ANDRUS, Am. Mineralogist, 37 , 622 (1952).

. U.BONSE & E.KAPPLER, Z.Naturforschung, 13a , 348 (1958).
. U.BONSE, in: Direct observation of imperfections in crystals, eds. J.B.NEWKIRK &
H.WERNICK (Wiley, New York, 1962), p. 431 .
. S.MERIAM ABDUL GANI, PhD thesis, Durham University (1982).
. P.RIGLET, M.SAUVAGE, J.F.PETROFF & Y.EPELBOIN, Phil. Mag. A, 42 , 339
(1980).
. W.SPIRKL, B.K.TANNER, C.WHITEHOUSE, S.J.BARNETT, A.G.CULLIS, A.D.
JOHNSON, A.KEIR, B.USHER, G.F.CLARK, W.HAGSTON, C.R.HOGG &
B.LUNN, Phil. Mag. A, 70 , 531 (1994).
JOHNSON, A.KEIR, B.USHER, G.F.CLARK, C.R.HOGG, B.LUNN &
W.HAGSTON, Phil. Mag. A, 69 , 221 (1994).
. I.C.BASSIGNANA, D.A.MACQUISTAN & D.A.CLARK, Adv. X-ray Analysis, 34 ,
507 (1991).
. I.C.BASSIGNANA & D.A.MACQUISTAN, 7th Int. Conf. on III-V Semi-insulating
materials, Ixtapa Mexico, (1992).
. S.COCKERTON, PhD thesis, Durham University (1991).
. G.T.BROWN, M.S.SKOLNICK, G.R.JONES, B.K.TANNER & S.J.BARNETT, in:
Semi-insulating III-V compounds, eds. D.C.LOOK & J.S.BLAKEMORE (Shiva,
Nantwich, 1984) p. 76 .
. I.MINATO, H.HASHIZUME, H.WATANABE & J.MATSUI, Japan. J.Appl. Phys., 25 ,
1485 (1986).
. S.J.BARNETT, PhD thesis, Durham University (1987).
. B.JENICHEN, R.KOHLER & W.MOHLING, Phys. Stat. Sol. (a), 89 , 12 (1985).
. B.JENICHEN, R.KOHLER & W.MOHLING, J. Phys. E: Sci. Inst., 21 , 1062 (1988).
. R.KöHLER, Appl. Phys., A58 , 149 (1994).
10
Synchrotron Radiation Topography
In this chapter, the application of synchrotron radiation for X-ray

topography is reviewed. The intensity and continuous spectrum of
synchrotron radiation is particularly important but we see that the
time structure and polarisation can also be exploited
10.1
Introduction
Despite becoming gradually accepted as a standard characterisation technique
for highly perfect crystals, X-ray topography has a reputation for being slow
when performed with conventional X-ray generators in the laboratory. The need
for high collimation sets a limit on the elastic strains which can be present in the
crystal if high resolution topographs are to be obtained and makes setting up the
experiment non-trivial. In 1974 Tuomi et al . 1 showed that synchrotron
radiation could be used for X-ray topography with a very simple experimental
configuration, which Hart 2 subsequently demonstrated was capable of high-
resolution imaging. This led to a major revolution in the type of experiment
which could be performed and the nature of the data gathered. The field of
synchrotron X-radiation topography has been recently reviewed by the present
authors. 3
10.2
Synchrotron radiation sources
In a conventional tube, X-rays are generated when energetic electrons are
stopped by impact with a metal target. The process is inefficient and the thermal
properties of solids place an upper limit on the brightness of X-ray beams that
can be achieved. [Note that brightness is flux per unit solid angle and while flux
may always be increased by increasing the area of target used for electron
impact, this does not increase the brightness. As we have seen, in a topographic
experiment the spatial resolution deteriorates as the source size increases.]
When electrons are confined to a circular orbit in a storage ring by a magnetic

field, electromagnetic radiation is emitted owing to their acceleration towards the
centre of the circle. At relativistic velocities, the emission is strongly peaked in
the forward direction, tangential to the electron orbit (Figure 10.1(a)).
Perpendicular to the orbit plane, the emission angle λ is given approximately by
λ =m 0 c 2/E, where m 0 is the electron rest mass, c is the velocity of light and E is
the electron energy. At 5 GeV, we find that the vertical emission angle is
naturally collimated to 0.1 mrad. In the orbit plane, the emission is integrated
over a range of natural collimation angles as the electron sweeps round the ring
and in practice the beamline acceptance angle limits the divergence in the
horizontal plane.
The relativistic transformation from the electron frame to the laboratory frame
results in a spectrum which is a continuum extending from the radiofrequency to
the X-ray region (Figure 10.1(b)). This is a ‘universal’ curve scaled to a
parameter called the critical wavelength λ c. The low wavelength fall-off is
determined by the energy of the electron beam and the radius of curvature at the
point of emission. As there are no target materials to melt, the brightness
achievable from synchrotron radiation sources may be 103 to 108 times higher
than those obtained from standard X-ray generators.
In a storage ring designed for production of synchrotron radiation
(Figure 10.2) the orbit is not circular but consists of dipole bending magnets
separated by straight sections. The beam is focused by quadrupole magnets,
forming a complex ‘lattice’ of magnets. After initial acceleration, to a few tens
of MeV in a linear accelerator, electrons (or sometimes positrons) are accelerated
to the full energy of 2−10GeV by a booster synchrotron, either alone or in
combination with acceleration in the storage ring itself. When the full energy is
reached the magnetic field is held steady (apart from dynamic control
adjustments to keep a stable orbit), the electron energy is maintained once or
more per revolution by a radiofrequency accelerating cavity. The electron beam
current decays exponentially by collision with residual gas molecules in the ultra-
high vacuum and lifetimes for decay to 1/e of the initial value may be as long as
70 hours.
Beamlines are constructed to extract the beam from tangent points of the
bending magnets or from insertion devices within the straight sections. In the
latter, the electron beam is deflected in a small radius of curvature by a
succession of dipole magnets of alternating sense. There are of two types:
wigglers, in which the amplitude of oscillation is large with respect to the cone
of emitted radiation, and undulators in which it is small. Recent devices all exploit
permanent magnets of neodymium iron boron to produce high fields in small
gaps which can be varied according to the experiment being performed. The
spectrum of radiation produced by a wiggler is similar to that from a bending
magnet, but with the wavelength shifted to a shorter value by virtue of the higher
magnetic fields that are used. Use of a multipole device enables the power in the
beam to be increased proportionally to the number of periods in the device. In
SYNCHROTRON RADIATION TOPOGRAPHY 235
Figure 10.1 (a) Distribution of emitted radiation from an electron travelling at relativistic
speeds in a circular orbit. (b) The universal, vertically integrated, synchrotron radiation
spectrum as a function of reduced wavelength. The axes are calibrated for the SRS at
Daresbury Laboratory
the third-generation sources such as the ESRF at Grenoble, in France, the APS at
Argonne, in the USA, and SPring-8 near Kobe in Japan, most of the beamlines
Figure 10.2 Schematic diagram of an electron storage ring showing the magnet lattice and
the form of the orbit
are served by such insertion devices where the beam is deflected and then
resumes its original orbit.
The spectrum from an undulator is very different, and numerous peaks result
from interference effects within the undulator. When the electron acceleration is
confined to the orbit plane and the emission angle very low, the radiation is
strongly elliptically polarised and, in the orbit plane itself, it is to within a few per
cent linearly polarised. Use of a sequence of permanent magnets with
magnetisation arranged in a spiral sequence enables circularly polarised radiation
to be extracted from such a helical undulator and this radiation is particularly
important for magnetic studies.
The electrons are bunched by the action of the RF accelerating field and
therefore the emission occurs in sharp pulses, only tens of picoseconds in width,
with frequency corresponding to the bunch spacing. For maximum beam current
many bunches are used, giving a pulse frequency of typically 500 MHz. When
only one bunch is injected into the ring, the period drops to a few megahertz,
while the pulse width remains at a few picoseconds. This feature has been used
extensively for fluorescence lifetime measurements and may also be exploited in
stroboscopic topography.
10.3
Source requirements for synchrotron X-ray topography
In the laboratory, X-ray topography is usually performed with characteristic
radiation and it is straightforward to show that the exposure time for a section
topograph scales with and for a traverse topograph as where P is the X-ray tube
power and S x is the source dimension in the incidence plane. The continuous
nature of the synchrotron spectrum, however, results in a very different

dependence on the source power and dimensions. Intensity dI reaching the crystal
due to area dX, dY of the source is
dI=CP ( λ ) λ λ dXdY/(S x S y D 2) (10.1)
where C is a constant, P is the power, S x is the horizontal size of the source, S y
is the vertical size of the source, and D is the source to specimen distance. We
assume that the specimen is smaller than the cone of radiation. We also assume
that the angular width of the source seen by a point on the specimen ■ λ s is
larger than the intrinsic range of the Bragg reflection λ λ . (This is valid for first-
generation sources and for high energy radiation weak Bragg reflections.)
(10.2)
where λ λ =constant.|C|λ 2 b 1 / 2 F h/sin2λ B. Thus

I=CP(λ )λ λ λ λ /(S x D) (10.3)
Consideration of the duMond diagram (Figure 10.3(a)) shows that
λ λ =λ cotλ ■ λ s (10.4)
where ■ λ s=S x /D, the angle subtended by the source. The intensity I h is given
by
I h =A|C|λ 3 P(λ )F h b 1/2/D 2sin2 λ B (10.5)
where b=λ 0/|λ h|, C=1 or cos2λ B , F h is the structure factor, A is a constant and
λ B is the Bragg angle. D is the specimen to source distance, P ( λ ) is the source
power per unit wavelength interval, λ 0=K 0 .n and λ h =K h .n , where n is a unit
vector parallel to the inward normal to the specimen surface. K h and K 0 are the
diffracted and incident wavevectors.
Now let the source angle ■ λ s be much less than the reflecting range of the
crystal λ λ . This corresponds to low order, long wavelength reflections at third-
generation sources. Then
(10.6)
Now, as seen in Figure (10.3(b))

λ λ =λ cotλ B λ λ
and again we have
I h =A|C|λ 3 P(λ )F h b 1/2/D 2 sin2 λ B (10.8)
There is thus no complicating factor involving source dimensions on going from
the first to the second criterion used above. Importantly, the intensity is not
dependent on the source size but rather on total source power. However, the
resolution R x in a topograph is given approximately by
R x =dS x /D (10.9)
where d is the source to detector distance. There is thus gain in intensity with
increased brilliance if D is reduced, or gain in resolution if D remains constant.
The limit on reduction of D is the beam height at the specimen. There is a
requirement to image an area of about 10 mm by 10 mm to half intensity across
the Gaussian beam intensity profile. Then as the intrinsic divergence m 0 c 2/

E=8×10− 5 for a 6 GeV machine, this gives D=100 m for a typical topography
beamline. The resolution is excellent and R=0.5 μ m for specimen to detector
distance of 10cm with a 0.5 mm half-width source size.
Multipole wigglers clearly have major advantages in the enhanced power
available for topographic experiments but heat load is a serious problem. Owing
to the non-uniform beam, which has a Gaussian profile in intensity there is non-
uniform heating. The crystal is locally bent and, when the angular deviation A
due to the lattice strain exceeds the rocking curve width λ λ , a strong band of
contrast is observed. Barrett et al . 4 have shown that the maximum angular
variation A from heating should be
(10.10)
where λ is the linear thermal expansion coefficient, K is the thermal

conductivity, λ is the rms width of beam profile and P abs is the power absorbed.
Ultra-short exposure times, on microsecond timescales, are unlikely to be
achieved for white beam topography owing to this heat loading problem. It is
particularly a problem for low conductivity insulators such as quartz and HgI2.
With high levels of bending, direct image contrast falls as the ‘perfect’ crystal
background rises and the defect images become invisible. Solutions to the
problem can be to use a chopper in front of the specimen to throw away up to
90% of the intensity and to filter out short wave-lengths with an absorber of
aluminium between 0.5 to 1.5 mm thick. A further reduction can be achieved by
‘cooling’ the beam by passage through about 50mm of water. Inclusion of a
solution containing the element under investigation reduces the intensity around
the absorption edge disproportionately and works very well.
A solution to the heat load problem is to use a highbandwidth monochromator
to reject most of the incoming wavelengths. However, mirrors are at present not
sufficiently uniform to be usable and the figuring error on the commercially
available synthetic multilayers is too great for topographic use. Revol et al . 5
have demonstrated the use of a vibrating silicon crystal as a broadband
monochromator and in any case a variable gap wiggler is essential. For long
wavelength work the wiggler gap should be increased.
The spectrum of an undulator which, owing to interference of the coherent X-
radiation emitted by the electron on successive excursions within the magnet,
consists of a series of peaks, looks at first sight to be ideal for synchrotron X-ray
topography. The energy width of the undulator peaks is large compared with the
perfect crystal reflecting range and quasi-white beam topography could be
performed without the heat load problems of a multipole wiggler. In addition, the
coherence of the undulator radiation results in it being many orders of magnitude
brighter than that from a multipole wiggler. The beam is of low divergence but at
400m is sufficiently wide to meet the beam-size criterion given above. However,
the spectrum from an undulator varies rather fast as a function of angle and this
Figure 10.3 (a) duMond diagram corresponding to a synchrotron radiation source where
the angular width of the source is larger than the perfect crystal reflecting range, (b)
duMond diagram where the reverse is true
means that the beam arriving at the sample has a different spectral characteristic
at any point. As seen from Figure 10.4(a), the crystal picks out a narrow band of
energy to diffract and, for a given setting, the intensity in the diffracted beam
falls much more rapidly than expected (Figure 10.4(b)). Unless a beamline of 1
km length is used, undulators do not seem to be ideally suited to X-ray
topography.
10.4
Applicability of synchrotron radiation to X-ray topography
There are five features of synchrotron radiation which make it both unique and
worth the inconvenience of travelling to a distant place to conduct experiments.
10.4.1
Intensity
In order to take high resolution X-ray topographs, slow, very small grain size
photographic emulsions must be used. Ilford L4 nuclear emulsions give the best
results, and exposure times on conventional generators range from hours or days.
Use of synchrotron radiation reduces exposure times on these emulsions to a
matter of seconds or less. Direct imaging TV detectors, which are severely
limited in resolution from quantum noise with conventional generators have
enough photons to give good quality images in real time. Depending on the area
imaged, the flux available at the detector and the type of device, resolutions
between 40 and 8 micrometres have been achieved. Figure 10.5 shows an
example of low-angle boundaries and magnetic domains in a crystal of iron-
silicon alloy imaged in real time with an X-ray sensitive TV camera at the
Daresbury SRS. Dynamic experiments, in which changes in the material are
studied as a function of variables, such as time, temperature, stress or electric or
magnetic field, may thus be performed either in real-time at moderate resolution
or in a rapid step-by-step mode at high resolution. Survey experiments in which a
very large number of samples are studied to provide statistically meaningful data
may also be undertaken.
10.4.2
Continuous spectrum
A storage ring designed for synchrotron radiation emits radiation over a very
broad spectrum (Figure 10.1(b)) which extends down to 0.1 Å at the third-
generation sources. The long wavelength cutoff is usually set at about 2.5 Å from
the absorption of the beryllium window. This wide wavelength range allows one
to optimise independently geometry, absorption conditions and strain sensitivity.
The continuous spectrum permits high resolution Laue topographs to be taken
Figure 10.4 (a) Spectrum, at 400 m distance from an ESRF undulator, for different
positions across the image with band of wavelengths diffracted by a typical crystal. (b)
Diffracted intensity as a function of position across the image
which are extremely easy to set up and are applicable to a wide range of
materials. In addition, it is possible exploit the dispersion near absorption edges.
Figure 10.5 Real-time image, with no computer processing, of magnetic domains and low-
angle boundaries in iron-silicon taken using the Bede Scientific HI-RES detector at
Daresbury Laboratory. The white radiation topograph was taken with X-rays of
fundamental wavelength 1 Å
10.4.3
Collimation
Because of the low divergence, topography experiments may be conducted at a
great distance (50–1000 m) from the source. This provides a wide beam at the
specimen and large areas may be illuminated without the need for scanning.
10.4.4
Polarisation
The high degree of X-ray polarisation in the electron orbit plane provides means
of controlling both the signal/noise ratio and the penetration of the X-rays into
the specimen. Depending on whether the incidence plane is chosen vertically or
horizontally, sigma or pi polarisation may be selected. The strain sensitivity and
the extinction distance can thus be varied while the normal photoelectric
absorption conditions remain identical.
1 0.4.5
Time structure
The pulsed time structure, typically at megahertz rates, permits stroboscopic
topography of very rapid cyclic phenomena such as the propagation of acoustic
waves.
10.5
Techniques of synchrotron radiation topography
10.5.1
White radiation topography
In a conventional Laue photograph, when ‘white’ (that is continuous) radiation is
allowed to f all on a single crystal, each set of planes selects a particular small
slice of the spectrum appropriate to the angle between the Bragg planes and the
incident beam to satisfy the Bragg condition. Many diffracted beams are
therefore observed in a pattern characteristic of the symmetry of the crystal
structure and the orientation of the crystal with respect to the beam. From this,
Guinier and Tennevin 6 and Schulz 7 had devised early continuous radiation
laboratory-based topographic techniques but the resolution was poor. Not until
synchrotron radiation became available did the power of these simple techniques
become apparent. When the source is a large distance from the specimen, and the
width of the beam at the sample large, each large Laue ‘spot’ becomes a high
resolution map of the scattering from different points across the crystal. Typical
numbers inserted into equation (10.9) show that a high spatial resolution can be
preserved in these diffraction topographs for most synchrotron radiation sources.
Figure 10.6 shows schematically how the Laue topographs arise.
Around defects, the scattering power differs from that in the perfect crystal
because X-rays which do not satisfy the Bragg condition in the perfect crystal
may be diffracted in the deformed region around the defect. Just as in the Lang
projection topograph, these regions behave as small crystals which diffract
kinematically and the net result is an increase in the intensity over that from the
perfect crystal.
For a given reflection, we can deduce from equations (10.5) and (10.8) that:
■ the polarisation factor, which is 1 for vertical dispersion and |cos 2λ B| for
horizontal dispersion, has little effect at short wavelengths and a horizontal
dispersion geometry can be used with little intensity loss
■ the intensity falls off with decreasing wavelength due both to λ and P(λ )
Defect contrast is similar to that in Lang topographs 8 but with differences due
to:
■ the presence of higher harmonics diffracted simultaneously

■ the divergence of the beams between crystal and film, and
■ the continuous nature of the radiation spectrum
The contribution of harmonics is not normally large at first- and second-

generation synchrotron sources, as the F hkl value falls rapidly with increasing
diffraction vector. A major exception is where weak quasi-forbidden reflections
Figure 10.6 White radiation topography with synchrotron radiation

such as 002 are being used in the III–V compounds. Here the 004 harmonic at λ /2
is extremely important as F 002 ( λ ) F004(λ /2). However, for third-generation
sources operating at high energy, the critical wavelength which determines the
point at which the spectrum rolls off is very short. For a fundamental at about 1.5
Å, the incident beam flux at harmonics up to fifth order will be very high. In the
(Laue) transmission geometry, the low orders are attenuated by photoelectric
absorption more than the higher harmonics and this compensates for the loss in
intensity in the higher harmonics predicted by equations (10.5) and (10.8). Thus,
at the ESRF for example, the harmonic contribution to the image is extremely
important, whereas at a second-generation source such as Daresbury, the problem
is minimal. As a result the image width as a function of wavelength may not be a
monotonic function and in some instances increases with increasing wavelength
(Figure 10.7). This is contrary to the prediction in Chapter 8 that the width of
dislocation images increases with decreasing wavelength. 9 , 10
A second difference in contrast arises due to the large specimen-to-plate
distances which are used in order to separate the various Laue topographs. Then
the angular deviation of the beams diffracted from around the defect 8 becomes
significant. The beams diffracted kinematically from either side of an edge
dislocation (Figure 10.8) diverge or converge according to the effective
misorientation around the defect. Dislocations with edge components often show
this bimodal contrast which is only seen in the laboratory for high values of the
product g.b. The separation of the lobes increases linearly with distance of the
detector from the specimen.
The third difference arises from the continuous nature of the radiation giving
rise to orientation contrast only at the boundary between misoriented regions.
Here the beams overlap or diverge depending on the sense of the misorientation
with respect to the diffraction vector (Figure 10.9(a)). Best contrast occurs when
Figure 10.7 Width of a dislocation image under conditions where harmonics are strong as
at the ESRF. (Courtesy F.Zontone)
the beams diverge, where there is a total lack of intensity associated with the
boundary region. An example is given in Figure 10.9(b) of sub-grain boundaries
in Tb0.27Dy0.73Fe2. The magnetic domain contrast in Figure 10.9 arises because
the magnetostriction leads to a tetragon al distortion of the lattice. At a 90°
boundary the lattice is coherent along a (110) plane, but the {100} planes are
tilted on crossing the boundary. This misorientation results in orientation
contrast similar to that of a sub-grain boundary.
White radiation topography has the following special features:
1 Short exposures (from milliseconds up to minutes, depending on the source

and the specimen).
2 Many reflections recorded at once on one plate.
3 The ability to image defects in crystals which contain long-range elastic
strains as each point on the specimen selects its own appropriate wavelength
band for diffraction.
4 The ability to obtain overall strain maps of specimens that are too distorted
for individual defects to be distinguished.
5 A simple experimental arrangement that is well suited for experimental
stages for in situ control of experimental variables.
It is therefore particularly well-adapted to the following types of study:

Figure 10.8 Mechanism for formation of the bimodal image seen in many synchrotron
radiation topographs
1 Dynamic experiments involving the study of changes in materials in real
time, as a function for example of temperature, stress or magnetic and
electric fields.
2 Distorted crystals where the lattice perfection is poor but information is
required of the distribution of lattice strain through the crystal.
3 Survey experiments involving the rapid examination of large numbers of
samples, or large numbers of reflections in order to obtain detailed statistical
information relating to growth technique or detailed three-dimensional strain
analysis.
Most of the early work on synchrotron radiation topography was performed with
very simple goniometers to provide basic specimen orientation. The specimen is
placed in the beam and a plate or film used to record the image. However, in
order to set up a reflection at an arbitrary wavelength one needs at least two axes
on each of the specimen and detector, and to control polarisation one must be
able to rotate, albeit perhaps indirectly, around the incident beam. To perform
these functions quickly, for alignment or to perform dynamic experiments,
remote computer control is now essential, integrated with data collection. An
excellent example is the ‘Orient Express’ program at the ESRF. A computer
program to calculate the Laue pattern for an arbitrary specimen and detector
orientation, 11 or to index a Laue pattern from the position of several images is
an important adjunct to the station control software. Real-time imaging with
acceptable dynamic range and resolution is now a possibility with the high fluxes
at the most recent synchrotron radiation sources. Without doubt, however, the
major limitation in the exploitation of X-ray topography to dynamic, in situ
Figure 10.9 (a) Schematic diagram of the deformation of orientation contrast images in
white radiation topography. (b) Low-angle boundaries in Tb0.27Dy0.73Fe2 showing
divergent (white) contrast
experiments is in the performance of current detectors. The performance of a

quantum limited detector is governed by the Rose-de Vries law,
C 2 λ 2 Iλ λ /k 2=1 (10.11)
where C is the contrast of the defect, λ is the resolution or pixel size, I is the
intensity, λ is the exposure time, λ is the detector quantum detection efficiency
and k is the system signal-to-noise ratio. While improvements in spatial
resolution are clearly desirable, it is economically unrealistic to do other than
exploit technology developed for alternative commercial purposes. Rather slow
progress has been made in this area, which remains of critical importance to the
effective use of the third-generation sources. A micron resolution CCD detector
with moderate read-out rate is now operating at the ESRF.
The main reason for requiring rapid acquisition of different images is for
detailed strain analysis. We have seen in Chapter 8 that the Burgers vector of
dislocations can be found by noting that there is zero contrast for a screw
Figure 10.10 Laue topographs taken simultaneously from a crystal of NaBaNbO3
dislocation when g.b=0 and for an edge dislocation when, in addition, g.b×l=0,
where g is the reciprocal lattice vector of the reflection, b is the Burgers vector
and l the line direction vector of the dislocation. Selection of sufficient
reflections will then permit this determination, though in practice there may be
difficulties caused by anisotropic elastic behaviour and by surface relaxation.
With white radiation topography, several reflections can be recorded
simultaneously, with enormous time-saving over laboratory investigations. In
Figure 10.10, many reflections are recorded at once, providing enough
information to characterise the twin boundaries. With a conventional generator
this would have taken about a week.
The complete strain tensor in a specimen can be obtained by sufficient
topographic images, if a narrow slit is used to define a direction, or very
sensitively, an internal marker is obtained by choosing the wavelength range on
the image so that it crosses an absorption edge. The methodology has been
worked out and applied to strains in metals by Stock and colleagues. 12
10.5.2
Example applications of white radiation topography
In our review article cited above, 3 we have given a detailed review of the work
in the field up to 1994. Therefore in this section, we will give only a few
examples to indicate the power and range of application of the technique.
Relatively little work has been done using section topography with
synchrotron radiation. Machado and colleagues 13 have used it to study thick
highly perfect diamonds where the flexibility of choosing the Bragg angle for a
particular reflection over a considerable range permits the experimenter to obtain
several different views of the crystal interior whilst still retaining the same
diffraction vector. Halfpenny has also exploited this at Daresbury Laboratory for
in situ solution growth studies.
Tuomi et al . 14 used the Stanford SSRL source for section topography
studies of processing induced defects in semiconductors. In a study of intrinsic
and extrinsic gettering of silicon they observed the loss of the ‘perfect crystal’
interference fringe pattern in the section topograph after initial oxidation and its
recovery after the well drive cycle at 1150°C for 18 hours. This indicated that the
microdefects formed during the initial oxidation were dissolved during the drive
cycle. After field oxidation at 950°C for 10 hours, a very high number of
microdefects were created which totally destroyed the interference (Kato)
fringes. In the section topographs the denuded zone was clearly visible and the
technique provides a non-destructive means of measuring the width of this
region. Correlations were found between gettering defects revealed by
synchrotron radiation section topography and device yield. The best device yield
was observed from the middle sections of wafers, with the rear surface covered
by 2 μ m of polysilicon. These discernible precipitates or stacking faults do not
develop after the complete thermal cycle.
Studies of the plastic deformation processes in ice are of very considerable
interest as they help to shed light on glacier formation and propagation. Ice is an
almost ideal material for X-ray topography in that its absorption is very low and
due to the low atomic number, the structure factors F h are low, thus giving
large dislocation widths which may be recorded very satisfactorily on low
resolution film such as Kodak R film or even a TV detector. Hondoh et al . 15
used a direct conversion X-ray TV camera 16 to measure directly the climb rate
of dislocations, thereby yielding the self-diffusion coefficient as a function of
temperature. There have been a number of detailed studies, by Hondoh’s group
15 , 17 , 18 at the Photon Factory at Tsukuba in Japan, by Whitworth 19 − 22 at
Daresbury in the UK and by Dudley and colleagues at Brookhaven in the USA.

The influence of slip on non-basal planes in ice was found to be unexpectedly
high, and on these planes a very large ratio of edge to screw dislocation velocity
was seen. 17 , 20 , 21 Image forces near crystal surfaces also played an important
role in dislocation multiplication.
In the study of the nucleation and propagation of dislocations (Figure 10.11)
from scratches in a geometry designed to produce slip on non-basal planes, edge
dislocations were found to glide on non-basal planes, but screw dislocations were
completely immobile except in the basal plane. 21 Detailed measurements of the
dislocation velocities 17 , 22 using stress pulse techniques showed that
dislocations glided on the basal plane as straight segments in both screw and 60°
orientation with velocity proportional to stress. The formation of straight glide
segments in the basal plane is evidence that dislocations glide by propagation of
kinks across a Peierls barrier. However, activation energies determined for the
two types show marked order of magnitude disagreement with current
theoretical models. The activation energy data for non-basal dislocations
suggests that proton disorder may limit dislocation mobility, though there is still
Figure 10.11 Dislocation multiplication in ice during plastic deformation. (Courtesy Dr

R.Whitworth)
a disagreement between the predicted and observed magnitudes for the

dislocation velocities.
The Stonybrook group under Dudley has studied the behaviour of ice
bicrystals and has shown that under certain conditions, grain boundaries can act
as sources of dislocations. Grain boundary facets have been shown to act as
dislocation nucleation sites and grain boundaries themselves have been observed
to act as barriers to dislocation motion.
In the 1970s there was considerable interest in antiferromagnetic domains. As
there is no net magnetization in such domains, decoration and Lorentz
microscopy techniques cannot be used to reveal them. Polarised light can only be
used to reveal the birefringence associated with the lowering of the symmetry in
the antiferromagnetic phase if the magnetostriction is large. X-ray topography is
much more sensitive to the magnetostrictive distortion and can be used to reveal
the antiferromagnetic domain structures 23 in materials such as KNiF3 where the
magnetostriction is of the order of parts in 105. Quasi-static SR studies of KNiF3
and in moderate magnetic fields generated by an iron yoke electromagnet, and at
temperatures down to 4.2 K, provided the first experimental confirmation of
Neel’s theoretical predictions concerning the behaviour of cubic
antiferromagnets in applied fields. Domain boundaries move in such a way as to
increase the volume of material where the sub-lattice magnetisation is normal to
the applied field, thus reducing the free energy of the system. The experimentally
observed displacement as a function of field was in very good agreement with
the predictions of Neel’s theory. Figure 10.12 shows an example of a series of
topographs taken at temperature of 4.2 K as the magnetic field is increased.
Recrystallisation and grain growth after deformation has been studied by in
situ SXRT experiments by Gastaldi, Jourdan and co-workers for over a decade.
25 X-ray topography will not reveal anything until new grains are more than a
Figure 10.12 White radiation topographs of 90° antiferromagnetic domain walls in KNiF3
under conditions of increasing magnetic field. (a) 0.27T, (b) 0.46T (c) 0.64T
few micrometres in size, and so it is not useful for nucleation studies, but it is
excellent for overall distribution measurement of new grains, correlation with
other microstructure, kinetic studies and determination of defect structures after
grain growth. In situ kinetic studies, for example, enable the range of grain
growth kinetics to be extended to some 10–100 times higher velocities than is
possible with interrupted annealing experiments.
The work of Gastaldi and Jordan has provided important insights into the
interaction of moving grain boundaries and formation of defects when they
collide. Their early studies were performed only under high vacuum but
subsequent redesign of the furnace has permitted work to be achieved under
ultra-high vacuum conditions. In situ measurements of the grain boundary
displacements at various temperatures have been made 26 and it has been shown
how grains which nucleate at pre-positioned surface indentations and
subsequently grow into the pre-strained matrix, transform progressively from
random to faceted configurations. Analysis of migration rates as a function of
temperature yielded activation energies a factor of five greater than those
previously reported, a feature attributed to the variation in grain boundary
mobility with grain boundary inclination as a consequence of growth selection.
Matrix screw dislocations have been seen to be left behind moving grain
boundaries and these defects are incorporated in the dislocation network
produced by reaction with dislocations originating from the grain nucleus. In
contrast to previous findings, they concluded that the density of matrix
dislocations left behind did not depend on the migration rate, 25 leading to the
proposal of a new dislocation generation mechanism.
Figure 10.12 (cont.)
The group has more recently undertaken studies of the melt-growth of alloy
systems in which the liquid-solid interface is not planar. Figure 10.13 shows a
sequence of white beam topographs taken during melt growth of an AlCu alloy.
As the interface advances, the material which solidifies is relatively perfect, with
a low dislocation density. However, at the growth rate here (4.1 μ m/s) the
interface is not stable and cellular growth results. The cell spacing is a
continuous function of growth rate. Eventually molten liquid becomes entrapped
behind the advancing interface and from these regions, dislocations are
generated. The key insight is that the defects are generated a long way behind the
advancing interface. 27
10.5.3
Stroboscopic studies
Standard optical chopper methods may be applied to stroboscopic studies of
cyclic dynamic phenomena up to frequencies of a few kilohertz. While this has
been used in multiple crystal topography f or the study of the pinning of domain
walls in Fe-Si 28 it has not been applied to white radiation topography.
However, the time structure of the synchrotron beam itself may be used in high
frequency stroboscopy. In single-bunch mode, where only one bunch of
electrons orbits in the ring, X-rays are emitted in bursts typically picoseconds in
width, spaced by a few microseconds. Using this MHz signal taken from the
beam position monitors in the ring, a high frequency device can be excited phase
locked to the synchrotron radiation emission. This high frequency stroboscope
was first utilised in the study of travelling surface acoustic waves in SAW
devices (intermediate frequency filters and delay lines), by Whatmore, Goddard,
Tanner and Clark 29 and bulk wave devices by Graeff and Gluer. 30 − 32 Until
these experiments were undertaken, it had only generally been possible to
observe the time-averaged displacement. Loss of transmitted acoustic power in
surface acoustic wave (SAW) delay lines had been a problem for some time and
was thought to be associated with lattice defects. The stroboscopic topography
experiments, together with other evidence, showed conclusively that the main
defect responsible for the SAW scattering was a region of reverse polarisation,
known as a ‘rille’. These reverse ferroelectric domains were associated with low-
angle grain boundaries, their interaction with the boundary preventing their
removal during the poling process. The defects were shown to initiate new bulk
wave modes, thus removing power from the surface wave.
An illustration of travelling surface waves revealed by stroboscopic SXRT is
shown in Figure 10.14. Here, interaction of the wave with a sub-grain boundary
can be clearly seen. A phase shift is introduced and, at the sub-grain boundary,
new modes are excited. In particular one may identify the second- and fourth-
order harmonics. Thus, although the defect does not absorb the acoustic power, it
results in a redistribution of power into unwanted modes of excitation. From an
engineering viewpoint, the power is lost.
Bulk wave devices have different tolerances and recently Capelle, Zarka and
co-workers have studied bulk waves in quartz resonators and used stroboscopy to
identify unwanted modes associated with defects. 33 They have also performed
true section topography in stroboscopic mode to identify if the interaction
between a dislocation and the acoustic wave could be described by simple linear
piezoelectric theory. Using simulation of the section topographs to analyse the
data, they concluded that a non-linear interaction was present near to the
dislocation line, linear theory working satisfactorily in the region far from the
defect. Etch channels appeared to have more influence on the acoustic wave than
individual dislocations.
Figure 10.13 In situ synchrotron radiation X-ray topographs of the interface between melt
and solid in an AlCu alloy. (Courtesy G.Grange and J.Gastaldi)
10.5.4
Multiple crystal topography
The white beam results in a high background and insensitivity to long-range
strains. Both problems are overcome by use of a crystal before the specimen,
which, owing to the small intrinsic angular divergence of the synchrotron
radiation beam, acts as a monochromator. This is illustrated in the duMond
diagram of Figure 10.3(b). The area is phase space is small and we see
immediately that the reflecting range of the monochromator sets the wavelength
dispersion, while the angular divergence is set by the source characteristics. Here,
unlike the laboratory case, we can think of the first crystal as a genuine
Figure 10.14 Stroboscopic topograph of travelling surface acoustic waves in a LiNbO3 IF

filter
monochromator. As discussed in Chapter 9, the high collimation, and hence

small dispersion, results in the (+ n, − m ) and (+ n, + m ) settings giving almost
identical results. With synchrotron radiation, it is often convenient to use a
channel-cut crystal or a two-crystal Golovshenko-Graeff device, in what is
strictly the double-crystal geometry. The pair of reflections throws the beam into
the forward direction and greatly simplifies the experiment.
If the (flat) specimen is rotated in the dispersion plane with respect to the beam
conditioner the rocking curve is extremely narrow and is the correlation function
of the two perfect crystal reflecting ranges. This is typically a few seconds of
arc. If the specimen or the beam conditioner contains a region misoriented by
more than this amount, no diffracted intensity reaches the detector from that
region. Thus, a photographic plate will record a topograph with very high strain
sensitivity. The contrast in double-crystal topographs is mainly orientation
contrast and has been treated by Bonse 34 (see Chapter 9). There are several
differences between double-crystal topography in the laboratory and at
synchrotron radiation source.
1 Owing to the distance between source and specimen, the divergence of the
beam normal to the dispersion plane is small and with SR it is unnecessary
to perform tilt adjustments to bring the Bragg planes of specimen and beam
conditioner exactly parallel.
2 Owing to the small divergence in the dispersion plane the range of
wavelengths diffracted from the beam conditioner is always small and
dispersion broadening of the SR rocking curve is small, even when the
Bragg planes of beam conditioner and specimen are significantly different.
3 As specimen and beam conditioner are both far from the SR source, defects
in both are revealed with almost equally good resolution on the topograph.
4 Owing to the very high intensity of SR, exposure times on nuclear plates are
typically ten minutes, whereas for such a slow emulsion, exposures are of
the order of days in the laboratory. This reduction in timescale is particularly
important for multiple exposure topographs of deformed crystals.
A further important consequence of the small source size and low divergence is
that it now becomes possible to undertake ‘weak beam’ topography. Weak beam
imaging is a standard technique in electron microscopy. The technique is to set

the specimen far from the Bragg peak and then only the region where the lattice
planes around the defect are rotated into the Bragg condition gives rise to
scattered intensity. As the strain field falls with increasing distance, the
diffracting volume falls rapidly and thus the image rapidly narrows. Weak beam
topography has not previously been widely practicable because of the large
source sizes of earlier synchrotron radiation machines. However, with third-
generation sources, the technique works well and it has already been
demonstrated at the ESRF. 35 Using simulation code 36 developed for use in the
Bragg reflection geometry, the variation in image width with deviation from the
Bragg position has been examined. It is seen that the image narrows rapidly out
to about twice the perfect crystal reflecting range, but after that decreases rather
slowly. The intensity continues to fall and little is gained at large deviations
(Figure 10.15).
10.5.5
Example applications of synchrotron radiation for double-
crystal topography
The earliest application was at LURE in the study of misfit dislocations in
GaAlAs epitaxial layers on GaAs 37 using an elegant, fixed wavelength, low
divergence beam conditioner designed by Hashizume. Bragg case dislocation
images were compared with simulations and excellent agreement was found. 38
Double-crystal topography has the advantage that the substrate and epilayer can
be imaged separately and defects in each identified. More recently, at Daresbury,
studies of thin InGaAs layers on GaAs have shown how the misfit dislocations
are initially nucleated in this highly strained system. This result is of immense
technological importance. Substantial relaxation of the strain leaves
optoelectronic devices unworkable and thus it is extremely important to
understand the formation mechanisms of the misfit dislocations which are
responsible for the relaxation. The initial relaxation can be extremely
anisotropic, as seen in Figure 10.16. This was believed to be due to the differing
mobilities of dislocations on the two types of glide plane in the III–V
compounds.
The in situ studies have shown, however, that all the misfit dislocations
nucleated in the initial stage are not only of 60° type with inclined Burgers
vectors, but also of the fast ■ type. 39 From these studies in very low dislocation
density vertical freeze gradient material, it has become apparent that the misfit
dislocation anisotropy arises from an anisotropy in the Burgers vector
distribution in the standard liquid encapsulated Czochralski crystals. The
measurements have given very clear support for the mechanism of misfit
dislocation generation proposed by Matthews and Blakeslee. 40 Two critical
thicknesses have been observed, one for initial nucleation of misfit dislocations
Figure 10.15 Simulated image width as a function of deviation parameter in Bragg case
weak beam topographs. Here, the specimen is set off the Bragg peak and an image of the
defect occurs only when the lattice planes are locally rotated or dilated back into the
Bragg condition. As this occurs only close to the dislocation core, the images are
narrowed from those under strong beam conditions
and another, at a substantially greater layer thickness, for misfit dislocation
multiplication.
In situ plastic deformation experiments on semiconductors have also been
undertaken at LURE by George and colleagues. Tensile loading experiments
were performed on sulphur-doped InP 41 in the temperature region of 548 to 648
K. In this range, where InP is ductile, the velocities of screw and 60° dislocations
were measured, an activation energy of 1.7 eV being obtained from the
temperature-dependence. The velocity of the fast 60° dislocations appeared to
depend signifi cantly on the orientation relative to the crystal surface, a feature
still not satisfactorily explained. Studies of GaAs and Si have been concerned
with the phenomenon of fracture, crack-tip plasticity and the brittle-to-ductile
(BDT) transition. Samples were fabricated containing a notch which was then
loaded in such a manner that the applied stress intensity factor at the crack tip
increased at a constant rate. The BTD transition was found to be extremely
sharp, 42 which can be related to the rapid change in dislocation velocities.
An important feature to note in double-axis topography experiments is that
when the beam area is large, the measured rocking curve widths are not
necessarily intrinsic. For example, mismatched epitaxial layers curve substrate
wafers by an amount which depends on the degree of mismatch and layer
thickness. Topographs of such curved wafers show bands of diffracted intensity,
Figure 10.16 Double-crystal topographs of misfit dislocations in InGaAs on GaAs: (a)

below, (b) at, and (c) above the critical thickness for relaxation
and it is often possible to image both substrate and epilayer reflections from
different parts of the specimen on one plate. These intensity bands can be used to
plot the long-range deformation in imperfect crystals by making multiple
exposures on a single plate with the specimen moved by a constant angular step
between exposures.
This technique, first developed by Renninger 43 using laboratory equipment,
has been used to study the relation between deep level, electrically active EL2
defects and lattice perfection in semi-insulating GaAs. 44 Ishikawa et al. 45
performed plane wave topography with a separate monochromator-collimator in
the (+, +) setting to produce beautiful ‘zebra patterns’ of stripes as a different part
of the crystal diffracts for successive angular steps. A two-fold symmetric strain
was observed in a (001) orientation GaAs wafer with a highly stressed region in
In-doped materials owing to impurity segregation. In order to distinguish
between tilts and dilations, two sequences of ‘zebra-patterns’ are required.
Provided that the crystal symmetry permits this, the simplest arrangement is to
rotate the crystal 180° about its surface normal. Then for one set of contours the
tilts and dilations they add, for the other they subtract. It is then possible to
determine the separate contours for tilt and dilation. Bilello et al. 46 have used a
similar technique with the double-crystal station at the NSLS, Brookhaven, to
measure the strains around cracks in fatigued zinc bicrystals. By use of several
sets of Bragg contours in the double-axis geometry, they were able to reconstruct
the full tensor deformation field around cracks. These experiments prove that
high angular sensitivity, but low spatial resolution, X-ray topography has
applications outside the confines of highly perfect crystals. Their work is an
excellent example of how X-ray topography, with a very high intrinsic strain
sensitivity can be applied to heavily deformed materials. X-ray topography is not
just useful for the study of highly perfect single crystals.
On the other hand, the importance of highly perfect crystals to the
semiconductor industry in Japan has meant that considerable effort has been
expended in experiments to increase the sensitivity to microstrains. These
experiments have all used one or more beam-conditioning crystals to produce a
pseudo plane wave, a beam of angular divergence much less than the reflecting
range of the specimen crystal. Chikaura, Imai and Ishikawa 47 used an
asymmetrically cut reference crystal to give a beam of divergence 0.59 arcsec,
less than 0.3 of the range of the specimen reflection, but subsequent experiments
have improved on this significantly. 48 As an example of the application of the
instrumentation at the Photon Factory, we cite the work of Matsui and colleagues
49 from NEC who have measured the strain fields associated with D type
microdefects. Experiments with such highly collimated and low disper sion beams
take only a very small fraction of the incident X-ray flux and thus second- or
third-generation synchrotron sources are essential if exposure times are to be
practicable. We note, however, that the extremely small source size now
achieved at the ESRF means that the divergence is always sub-arcsecond at the
topography station.
10.5.6
Triple-axis topography
In the double-crystal mode, tilts and dilations can only be distinguished by use of
more than one reflection and then reconstructing the strain field. A direct method
for producing tilt or dilation maps involves the use of an analyser crystal after
the specimen. In the true triple-axis geometry (Figure 10.17), the analyser selects
the scattered intensity from a small volume of reciprocal space. It is easy to see
Figure 10.17 Schematic diagram of the triple-axis topography arrangement
that if the analyser is left fixed and the specimen only is scanned, then rocking
curve broadening arises solely from lattice tilts. By recording a topograph while
the specimen is scanned over this rocking curve, a band of intensity is recorded
on the topograph corresponding to a contour of constant lattice plane dilation. 50
Subsequent steps of the analyser permit full contour maps of dilation to be built
up by the multiple exposure technique described above. Such experiments are
impracticable in the laboratory because of the long exposure times implicit, but are
quite feasible at synchrotron radiation sources.
Kitano et al . 51 have used the triple-axis topography method to examine the
tilts and dilations around growth dislocations in bulk GaAs crystals. A water-
cooled silicon 111 reflection was used to remove most of the heat load on the
experiment, the following two 553 asymmetric reflections being chosen because
the Bragg plane spacing is almost the same as the 008 reflection used for the
GaAs specimen. A symmetrically cut silicon crystal was used with the 553
reflection as analyser. Using this arrangement, they were able to measure the
lattice spacing variations in regions with different dislocation configurations.
Near cell walls, the interplanar spacing contracts, while where dislocations
interact with the crystal surface, the interplanar spacing expands. The reasons for
these variations are not fully established but believed to be due to different
concentrations of point defects.
10.6
Summary
X-ray topography with synchrotron radiation is now a well established technique
for the visualisation and measurement of strains in crystalline materials.
Continuing development of storage ring sources is resulting in a steady
improvement in capability. The most significant gains in the near future may now
well come from attention to detector technology.
References
. T.TUOMI, K.NAUKKARINEN & P.RABE, Phys. Stat. Sol. (a), 25 , 93 (1974).

. M.HART, J. Appl. Cryst., 8 , 436 (1975).
. D.K.BOWEN & B.K.TANNER, Materials Science Reports, 8 , 369 (1992).
. R.BARRETT, J.BARUCHEL, J.HÄRTWIG & F.ZONTONE, J. Phys. D: Appl. Phys.,
28 , A250 (1995).
. C.REVOL, J.BARUCHEL, D.BELLET, G.MAROT, P.THÉVENEAU & F.ZONTONE,
J. Phys. D: Appl. Phys., 28 , A 262 (1995).
. A.GUINIER & J.TENNEVIN, Acta Cryst., 2 , 133 (1949).
. L.G.SCHULZ, J. Metals, 6 , 1082 (1954).
. B.K.TANNER, D.MIDGLEY & M.SAFA, J. Appl. Cryst., 10 , 281 (1977).
. A.AUTHIER, Adv. X-ray Analysis, 10 , 9 (1967).
. J.MILTAT & D.K.BOWEN, J. Appl. Cryst, 8 , 657 (1975).
. J.MILTAT & M.DUDLEY, J. Appl. Cryst, 13 , 555 (1980).
. S.R.STOCK, H.CHEN & H.BIRNBAUM, Phil. Mag., A, 53 , 73 (1986).
. W.G.MACHADO, M.MOORE & A.R.LANG, J. Crystal Growth, 71 , 718 (1985).
. T.TUOMI, M.TILLI & O.ANTTILA, J. Appl. Phys., 57 , 1384 (1985).
. T.HONDOH, A.GOTO, R.HOSHI, T.ONO, H.ANZAI, R.KAWASE, P.PIMIENTA &
S.MAE, Rev. Sci. Inst., 60 , 2494 (1989).
. S.SUZUKI, M.ANDO, K.HAYAKAWA, O.NITTONO, H.HASHIZUME, S.KISHINO
& K.KOHRA, Nucl. Inst. Meths., 227 , 584 (1984).
. T.HONDOH, H.IWAMATSU & S.MAE, Phil. Mag. A, 62 , 89 (1990).
. A.HIGASHI, T.FUKUDA, K.HONDOH, K.GOTO & S.AMAKI, in: Dislocations in
solids, eds. H.T.SUZUKI, K.NINOMIYA, K.SUMINO & S.TAKEUCHI (University
of Tokyo Press, 1985) p. 511 .
. S.AHMAD, M.OHTOMO & R.WHITWORTH, Nature, 319 , 659 (1987).
. S.AHMAD & R.WHITWORTH, Phil. Mag. A, 57 , 749 (1988).
. C.SHEARWOOD & R.W.WHITWORTH, J. Glaciology, 35 , 281 (1989).
. C.SHEARWOOD & R.W.WHITWORTH, Phil. Mag. A, 65 , 85 (1992).
. B.K.TANNER, Science Progress (Oxford), 67 , 411 (1981).
. M.SAFA & B.K.TANNER, Phil. Mag. B, 37 , 739 (1978).
. J.GASTALDI, C.JOURDAN, G.GRANGE & C.L.BAUER, Phys. Stat. Sol. (a), 109 ,
403 (1988).
. J.GASTALDI, C.JOURDAN & G.GRANGE, Phil. Mag. A, 57 , 971 (1988).
. G.GRANGE, J.GASTALDI, C.JOURDAN & B.BILLIA, J. Crystal Growth, 151 , 192
(1995).
. J.MILTAT & M. KLEMAN, J. Appl. Phys., 50 , 7695 (1980).
. R.W.WHATMORE, P.A.GODDARD, B.K.TANNER & G.F.CLARK, Nature 299 , 44
(1982).
. C.C.GLUER, W.GRAEFF & H.MOLLER, Nucl. Inst. Meths., 208 , 701 (1983).
. H.CERVA & W.GRAEFF, Phys. Stat. Sol. (a), 82 , 35 (1984).
. H.CERVA & W.GRAEFF, Phys. Stat. Sol. (a), 87 , 507 (1985).
. A.ZARKA, B.CAPELLE, J.DETAINT & J.SCHWARTZEL, J. Appl. Cryst., 21 , 967
(1988).
. U.BONSE, in: Direct observations of imperfections in crystals, eds. J.B.NEWKIRK &
H.WERNICK (Wiley, New York, 1962) p. 431 .
. F.ZONTONE, Thesis, University of Grenoble (1995).

JOHNSON, A.KEIR, B.USHER, G.F.CLARK, C.R.HOGG, B.LUNN &
W.HAGSTON, Phil. Mag., A, 69 , 221 (1994).
. J.F.PETROFF, M.SAUVAGE, P.RIGLET & H.HASHIZUME, Phil. Mag. A, 42 , 319
(1980).
. P.RIGLET, M.SAUVAGE, J.F.PETROFF & Y.EPELBOIN, Phil. Mag. A, 42 , 339
(1980).
. S.J.BARNETT, A.M.KEIR, A.G.CULLIS, A.D.JOHNSON, J.JEFFERSON,
G.W.SMITH, T.MARTIN, C.R.WHITEHOUSE, G.LACEY, G.F.CLARK,
B.K.TANNER, W.SPIRKL, B.LUNN, J.C.H.HOGG, W.E.HAGSTON &
C.M.CASTELLI, J. Phys. D: Appl. Phys., 28 , A17 (1995).
. J.W.MATTHEWS & A.E.BLAKESLEE, J. Crystal Growth, 27 , 118 (1974).
. A.GEORGE, A.JACQUES & R.COQUILLE, Inst. Phys. Conf. Ser., 76 , 439 (1985).
. G.MICHOT & A.GEORGE, Inst. Phys. Conf. Ser., 104 , 385 (1989).
. M.RENNINGER, Z. Angew. Phys., 19 , 20 (1965).
. G.T.BROWN, M.S.SKOLNICK, G.R.JONES, B.K.TANNER & S.J.BARNETT, in:
Semi-insulating III-V materials, eds. D.C.LOOK & J.S.BLAKEMORE (Shiva,
Nantwich, 1984) p. 76 .
. T.ISHIKAWA, T.KITANO & J.MATSUI, Japan. J. Appl. Phys., 24 , L968 (1985).
. J.C.BILELLO, H.A.SCHMITZ & D.DEW-HUGHES, J. Appl. Phys., 65 , 2282 (1989).
. Y.CHIKAURA, M.IMAI & T.ISHIKAWA, Japan. J. Appl. Phys., 26 , L889 (1987).
. Y.CHIKAURA & M.IMAI, Japan. J. Appl. Phys., 29 , 221 (1990).
. S.KIMURA, T.ISHIKAWA & J.MATSUI, Phil. Mag. A, 69 , 1179 (1994) .
. T.ISHIKAWA, T.KITANO & J.MATSUI, J. Appl. Cryst., 20 , 344 (1987).
. T.KITANO, T.ISHIKAWA & J.MATSUI, Phil. Mag. A, 63 , 95 (1991).
Index
aberrations 34 Borrmann effect 94, 200

absorption 3, 54, 78, 90, 95, 105, 206, 242, Borrmann fan 86, 94, 185, 194, 195, 204,
243, 248 210, 212
absorption coefficient 3, 78, 95, 95, 134 boundary matching 93, 203, 212
absorption depth 95, 185, 225 Bragg case geometry 81, 101, 104, 183,
acoustic waves 242, 252 206, 224, 256
activation energy 249, 250, 256 Bragg plane tilt 36ff
adhesive 47, 187 Bragg’s law 2, 20, 76, 80, 98, 102, 132,
AlCu alloy 252, 253 136, 144, 146, 161
AlGaAs 38, 116, 125ff, 135, 137, 144, bremsstrahlung 16, 28
146, 227, 229, 255 brittle to ductile transition 258
aliasing 123 Burgers vector 209, 225, 247, 256
alignment 47, 187
automated 39, 187 CdTe 124, 140, 152, 164
analyser 155, 160, 165, 165, 168, 172, 259 CdMnTe 152
graphite 172 channel-cut crystal 22, 28ff, 158, 160, 164,
streak 168, 169 254
anomalous dispersion 78, 240 characteristic spectrum 16, 28
anomalous transmission 94, 188 coherence 239
antiferromagnetic domain 249, 250 composition mapping 49, 68, 69
asymmetric reflections 31, 47, 54, 115, Compton scattering 78
134, 147, 178, 185, 217, 221, 227, 228 correlation (mathematical) 27, 59, 117,
asymmetry factor 103, 113, 225 120, 125, 254
atomic scattering factor 6, 72, 73, 144 cosmic rays 43
autocorrelation 68, 125 counting time 49
cracks 258
Be 240 critical wavelength 235, 243
beam
conditioner 14, 17, 19, 0 ff, 41, 42, detectors 43, 158, 160, 240, 244
117, 123, 155, 158, 165, 221, 254 charge-coupled device (CCD) 247
streak 168, 169 dead-time 45
size 39, 46, 59 efficiency 190, 192
Bessel function 107, 195 deviation parameters 88, 101, 105, 114
Bloch waves 88, 95, 98, 112, 194, 198, deviation vector 76, 169
199, 212 diamond 124, 196, 211, 248
263
264 INDEX
diffuse scatter 59, 98, 116, 127, 131, 160, peak shift 136, 147
162, 169, 173 relaxation 46, 52, 56, 58, 62ff, 114,
dilation distribution 162 136, 142, 147, 162, 173, 175, 217, 255,
direct image 189, 198, 206, 242 257
direction cosine 31, 66, 104, 114 superlattice 124, 142ff, 175
dislocation 52, 53, 101, 0 , 164, 184ff, tetragonal distortion 62, 115, 175
206ff, 248ff thickness 53, 65, 123, 132ff, 143
bimodal image 244 threading dislocations 162, 202, 230,
climb 248 229
glide 248, 256 tilt 58, 172, 173, 178
image width 209, 225, 243, 244, 248, zero-order peak 124, 143, 146, 147,
255 151
velocity 248, 249, 256, 258 Ewald sphere 80ff, 168ff, 188
dispersion 41 evanescent wave 92
dispersion (incidence) plane 36, 37, 220, extinction contrast 183, 198
221, 255 extinction distance 54, 95, 98, 185, 196,
dispersion surface 89ff, 112, 194, 195, 198, 200, 207, 209, 225, 242
199ff
duMond diagram 20ff, 42, 237, 238, 254 fatigue 258
duMond monochromator 30, 31, 221, 221 ferroelectric domains 215, 252
dynamical diffraction 168, 169, 206, 207, Fe–Si 240, 240, 252
230 forbidden reflections 75
dynamical image 189, 198, 199, 207, 210, four-reflection monochromators 30
211, 213 Fourier analysis 66, 87, 120, 124, 213
dynamical theory 70, 81ff, 112ff, 132, 136, fourier transform 146, 148, 151
183, 224
GaAs 3, 10, 28, 38, 42, 43, 54, 59, 66, 75,
edge dislocation 209, 243, 247, 248 116, 118, 125ff, 158, 164, 166, 172, 175–
effective misorientation 206, 209, 243 70, 211, 220, 221, 224–20, 255, 257–7
Eikonal theory 199, 200, 203 liquid encapsulated (LEC) 166, 227,
electric susceptibility 87, 90, 113, 114 230, 256
electron diffraction 6, 88 vertical gradient freeze (VGF) 166,
electron microscopy ix, 116, 188, 194, 227, 256
204, 206, 209, 255 GaN 158, 164, 165, 172, 173
epitaxial layers GaP 12
buffer layers 173, 227 Gaussian function 59, 132, 148, 149, 164,
capping layers 123, 173 237
composition 57, 123, 145, 149, 227 gettering 194, 248
critical thickness 256, 257 Ge 3, 30, 136, 162–6, 172, 173, 229
graded layers 137, 149, 173 glacier 248
HEMT structures 66, 127ff grazing incidence scattering 7, 143, 146,
interface grading 145, 149 225
interface roughness 145, 149 grain boundary 250, 252
misfit dislocations 62, 65, 164, 202, growth bands 215
224, 230, 229, 255 growth sector boundaries 215
mismatch 52, 56, 114, 131, 136, 140,
142, 143, 147, 172, 173, 258 harmonic reflections 243
INDEX 265
HgCdTe 124, 138, 162 relaxed 56

HgI2 239 see also epitaxial layers
HgMnTe 138, 164 Mosaic 39, 41, 52, 53, 75, 98ff, 151, 155,
Howie-Whelan equations 112, 205 160, 164, 172, 183, 206, 209, 228
multiple diffraction 116
ice 248, 249 multi-quantum well 142, 145
incidence (dispersion) plane 36, 37, 184,
192, 205, 210, 212, 236 NaBaNbO3 247
InGaAs 10, 66, 68, 127, 139, 142, 143, neutron diffraction 6
145, 175–70, 225, 225, 231, 255–5 non-dispersive geometry 22, 25ff, 118
InP 0 , 36, 41, 54, 55, 57, 121, 124, noise 118, 144, 158, 190,192
139, 143, 221, 223, 256 nuclear emulsion 183, 188, 239, 255
InSb 152
instrument function 117, 122 Orient Express 247
integrated intensity 66, 67, 105, 124, 125, orientation contrast 182, 183, 198, 244,
132–9, 151, 183, 195 244, 254
interference fringes 137, 146, 227
interferometer 140 Pendellösung 194, 198
intermediary image 198,202, 210, 211, Penning-Polder theory 199ff, 213
213 photographic film 189, 228, 239, 248
plane wave rocking curve 3, 22, 50, 115,
Kato fringes 194, 195, 212, 248 132
kinematical scattering 162, 242, 243 plastic deformation 248, 256
kinematical theory 70, 72ff, 107, 132, 135, Poisson ratio 57, 63, 114–11, 0 , 131
137, 183 polarisation beating 196, 211
KCoF3 249 polarisation (circular) 236
KNiF3 249, 250 polarisation factor 73, 88, 112, 114, 242
Kramers–Krönig relation 78 polishing damage 162, 164, 179
powder diffraction ix, 81, 84, 122
Laue case geometry 81, 103, 104, 207, 243 Poynting vector 91, 95, 199, 200
Laue condition 88, 95 precipitate 172, 194, 195, 200–5, 210, 248
Laue method (white radiation) 81, 85, 179, primary extinction 98
187, 242, 247 proportional counters 43
Laue point 90 pseudoplane wave 258
LiNbO3 254
Lorentzian function 59, 117, 0 , 131 quartz 239
low angle boundaries 228, 240, 252 quasi-forbidden reflections 75, 145, 225,
low temperature GaAs 172 243
low temperature topography 249
R factor 147, 149
magnetic domains 215, 240, 244, 249, 250, ray optics 91
252 reciprocal lattice point (relp) 81, 143, 168
magnetostriction 249 reciprocal lattice rod 146
Maxwell’s equations 87ff reciprocal lattice vector 80, 112, 199, 247
microgravity 172, 173 reciprocal space 79ff, 110, 143, 158, 159,
mismatch 170ff, 199, 259
experimental (X-ray) 56, 115 maps 143, 158, 166ff, 170ff
266 INDEX
recrystallisation 250 theta-two-theta scan 162–6, 169, 172

reflectivity 27, 105, 117 thickness fringes 66, 124–20, 135, 148,
refractive index 66, 87, 112, 162, 168, 169, 155, 175, 194–7
199 Thompson scattering 72
rocking curve 120, 217, 221, 223, 227 tie point 91, 194, 195, 199ff, 211, 212
profile 105, 122, 137, 221 jumping 203
width 102, 104, 122, 148, 172, 185, migration 200, 203
225, 237, 254, 258 tilt distribution 161ff, 259
rocking topograph 224 triple axis scattering 9, 19, 155ff
Rose-de Vries equation 192, 247 alignment 160
slit-collimated 42, 160, 178
sapphire 164, 165, 172 topography 259
satellite peaks 143, 146ff twin boundary 212, 247
scattering vector 73, 74, 169 two-beam approximation 88, 112
scintillation counters 43
scratches 179, 221 Undulator 236, 239, 240
screw dislocation 209, 225, 247, 248, 252, Uniqueness Theorem 122
256 unit cell 3, 72, 114
secondary extinction 98
simulation 123ff, 204, 224, 255, 256 Vegard’s law 57, 116, 117
single axis diffraction 14, 21
solid state detectors 43 wafer inspection 228
soller slit 160 wiggler 235, 237, 239
specimen curvature 28, 39, 59, 120, 122,
155, 186, 227, 230, 258 X-ray mirrors 31, 188, 239
spherical waves 98, 194, 195, 213 X-ray sources 16, 179
step size 47, 123 microfocus 34, 185
stacking fault 206, 211ff, 248 rotating anode 170, 188
extrinsic 215 X-ray topography 11, 20, 121, 179ff
intrinsic 215 Berg-Barrett method 183, 198
storage ring (electron) 232 ff contrast 193
strain contour mapping 227, 247 double crystal 179, 205, 217ff, 254
strained layers 36 curved crystal 230
structure factor 3, 21, 72, 74, 87, 114, 135, Lang method 11, 185ff, 242, 243
146, 196, 209, 225, 237, 243, 248 photographic processing 189
sub-grains 198, 244, 252 projection topography 185, 242
substrate misorientation 58, 173 real-time 190, 217, 240, 240, 247
synchrotron radiation 12, 17, 115, 116, section topography 179, 194ff, 204,
170, 179, 192, 210, 232ff 205, 210, 211, 213
sources 232 spatial resolution 183ff, 188ff, 217,
225–15, 232, 247, 255
Takagi’s equations 203, 205, 224 synchrotron radiation 232ff
Takagi-Taupin theory 86, 112ff, 136, 203, double crystal 225
205, 213 section topography 248
(TbDy)Fe2 244 stroboscopic 236, 242, 252, 254
thermal diffuse scatter 98 white beam 198, 240, 242ff
thermal vibration 98 triple axis 259
INDEX 267
weak beam 255
Zebra patterns 227, 258

Zn 184, 258

High Resolution X-Ray Diffractometry and Topography 1998

Uploaded by

Copyright:

Available Formats

High Resolution X-Ray Diffractometry and Topography 1998

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

High Resolution X-Ray Diffractometry and Topography 1998

Uploaded by

Copyright:

Available Formats

High Resolution X-ray Diffractometry and

UK Taylor & Francis Ltd, 1 Gunpowder Square, London EC4A 3DE

Copyright © D.Keith Bowen and Brian K.Tanner 1998

ISBN Master e-book ISBN

ISBN (Adobe eReader Format)

Library of Congress Cataloging-Publication-Data are available

1 Introduction: Diffraction Studies of Crystal Perfection 1

3.2 Plane wave diffraction from heteroepilayers 52

9.5 Synchrotron radiation double-crystal topography 226

We are pleased to acknowledge the contributions and the stimulation of our

In this chapter we introduce high resolution diffraction studies of

The experimental beginning of the study of materials by diffraction methods took

of silicon from the electronics industry have permitted the development of

■ the scattering strength of an electron or nucleus

Figure 1.2 (cont.)

1 The techniques are normally non-destructive, since adequate penetration and

Table 1.1 summarises the characteristics. Electron microscopy and diffraction

advantages of X-ray techniques are their convenience and non-destructive

Table 1.1 Characteristics of electron, neutron and X-radiations

1 For dynamic experiments, such as modelling of crystal growth or circuit

. M.H.LORETTO & R.E.SMALLMAN, Defect analysis in electron microscopy

We first define the geometry and instrumental parameters common

Figure 2.1 The divergence of a single-axis system

■ continuous radiation spectrum

Figure 2.4 Schematic of high resolution double-axis instrument

Figure 2.5 Schematic of high resolution triple-axis instrument

Next we consider the input to the crystal. A plane monochromatic wave is

■ The first reflection is given the notation + n.

• If the second reflection is – n then the plots superimpose and no rotation is

■ The intensity of the combined reflection is proportional to the area of overlap

The denominator (normalising constant) is the integrated reflectivity of the first

The double-crystal rocking curve is symmetric, though the plane wave

1 If the specimen is changed to a diff erent material or reflection then the

Accordingly, a number of beam conditioners have been developed which both

CCC to give very good angular resolution, plus a Ge 220 monochromating

for glancing or grazing incidence, and

High intensities can be achieved with conventional sources as well as with

■ the rocking curve is broadened

This is shown experimentally in Figure 2.22. The specimen comprises a GaAs

Figure 2.22 A three-dimensional plot of the diffracted intensity as a function of specimen

■ passes along the axis of the instrument, to a precision, 0.1°

3 Clamp the diffractometer if appropriate so that these alignments are

1 Mount the specimen on the holder, so that it is securely held without

. H.MARK, in Fifty years of X-ray diffraction, ed. P.P.EWALD (Int. Union of

In this chapter we discuss the measurement and analysis of simple

Table 3.2 Extinction distances in GaAs with CuK λ radiation

very useful to perform topography to see the dislocations directly, as discussed in

Table 3.3 The effect of curvature on rocking curve broadening

and the Bragg law, we obtain

The accuracy of c l may be checked from measurement of the symmetric 001

where b is the magnitude of the Burgers vector of the interface dislocations

Table 3.4 Comparison of best Fourier analysis with specimen structure 7

5 Perform the Fourier transform.

■ thin layers, say below 0.5 μ m

Computer simulation is required to analyse such cases, as will be discussed in

application in the semiconductor industry, since the procedures discussed may

. L.D.LANDAU & E.M.LIFSCHITZ, Elasticity (Pergamon, Oxford, 1972).

In this chapter we first extend the kinematical theory, discussing the

Conditions (2) and (3) are equivalent to the Fraunhofer or far-field

been divided by V to make direct comparison between the materials. A number