Modeling of diffraction patterns based on

microstructural properties

Gábor Ribárik

Physics Doctorate School

School Leader: Prof. Zalán Horváth

Material Science and Solid State Physics Program

Program Leader: Prof. János Lendvai

Eötvös Loránd University

Institute of Physics
Department of Materials Physics

Adviser:
Prof. Tamás Ungár, DSc

Ph.D. Thesis
2008.
Contents

1 Introduction 1

2 Theoretical aspects of X-ray line broadening 4

2.1 Line profile parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2 Size broadening of diffraction profiles . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2.1 Size parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2.2 Size distribution functions . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2.3 Determination of the size profile [S1, S4, S6] . . . . . . . . . . . . . . . . . 10
2.2.4 Anisotropic size broadening [S1, S4, S6] . . . . . . . . . . . . . . . . . . . 15
2.3 Strain broadening . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.3.1 The Krivoglaz–Wilkens model of dislocations . . . . . . . . . . . . . . . . . 20
2.3.2 Strain anisotropy: the concept of contrast factors . . . . . . . . . . . . . . . 25
2.3.3 The Groma-Csikor strain function . . . . . . . . . . . . . . . . . . . . . . . 27
2.4 Planar faults [S18] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

3 The classical methods of X-ray line profile analysis 31

3.1 Breadth methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.2 Fourier methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.3 Full profile fitting methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.4 Pattern fitting methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.5 Methods analysing the moments of the profiles . . . . . . . . . . . . . . . . . . . . 39

4 Determining microstructure by the Multiple Whole Profile fitting method 42

4.1 The MWP method [S4, S6] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.1.1 The principles of the method . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.1.2 Data preparation: profile separation using the MKDAT program . . . . . . . 42
4.1.3 Data preparation: instrumental deconvolution using the MKDAT program . . 43
4.1.4 Evaluation of X-ray diffraction profiles using the MWP program . . . . . . . 44
4.1.5 The steps of the MWP evaluation procedure . . . . . . . . . . . . . . . . . . 45
4.1.6 The MWP frontend . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.2 MWP application to deformed Cu [S4, S6] and ball milled PbS [S9] . . . . . . . . . 48
4.3 MWP application to Al-Mg alloys [S13, S14] . . . . . . . . . . . . . . . . . . . . . 51
 4.4 MWP application to Ti [S12, S21] and Si3 N4 [S1, S6] . . . . . . . . . . . . . . . . . 54
4.5 MWP evaluation of carbon black samples by assuming ellipsoidal crystallite shape
[S4, S5, S6, S8] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

5 Determining microstructure by the Convolutional Multiple Whole Profile fitting method 58

5.1 The CMWP method [S14] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
5.1.1 The principles of the method . . . . . . . . . . . . . . . . . . . . . . . . . . 58
5.1.2 The JAVA frontend of CMWP . . . . . . . . . . . . . . . . . . . . . . . . . 60
5.1.3 The WEB frontend of CMWP . . . . . . . . . . . . . . . . . . . . . . . . . 63
5.1.4 The CMWP evaluation procedure . . . . . . . . . . . . . . . . . . . . . . . 64
5.1.5 Determining the background . . . . . . . . . . . . . . . . . . . . . . . . . . 65
5.1.6 Sampling of the Fourier transforms . . . . . . . . . . . . . . . . . . . . . . 67
5.1.7 Sampling of the intensity patterns . . . . . . . . . . . . . . . . . . . . . . . 68
5.1.8 The profile cutting parameter . . . . . . . . . . . . . . . . . . . . . . . . . . 69
5.1.9 Instrumental profiles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
5.1.10 Weighting algorithms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
5.1.11 Interpretation of the errors of the fitting parameters . . . . . . . . . . . . . . 71
5.1.12 Convergence and numerical stability . . . . . . . . . . . . . . . . . . . . . . 72
5.2 Modeling asymmetric diffraction profiles . . . . . . . . . . . . . . . . . . . . . . . 75
5.3 CMWP application to Al-Mg alloys [S14, S16] . . . . . . . . . . . . . . . . . . . . 77
5.4 CMWP application to nanocrystalline Cu containing planar faults [S18] . . . . . . . 79
5.5 CMWP application to ball milled fluorides [S17] . . . . . . . . . . . . . . . . . . . 79
5.5.1 The interference effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
5.5.2 Correcting for the interference effect . . . . . . . . . . . . . . . . . . . . . . 86
5.5.3 The microstructure of the fluorides . . . . . . . . . . . . . . . . . . . . . . . 89

6 Summary and Conclusions 93

Own Publications 96

Bibliography 99
Chapter 1

Introduction

Macroscopic properties of materials are strongly affected by the microstructure. Microstructure, on

the other hand, is a complex feature with very different aspects which often depend even on the
method of investigation. The most common and most effective microstructure testing is based on
electron probing. Transmission electron microscopy (TEM) provides direct images of the microstruc-
ture of almost any kind of materials including hard crystalline, hard amorphous or soft materials.
Scanning electron microscopy (SEM) can test bulk materials, however, only the surface or surface
near regions. The focussed ion beam (FIB) technology combined with SEM and TEM offers new
perspectives for extremely sophisticated microstructure testing opportunities. Still, there are numer-
ous important features, especially residual internal stresses, microstresses, or different kinds of size
distributions which easily escape the observation facilities of electron microscopy methods. X-ray
line broadening was one of the first promising methods of microstructure testing, long before electron
microscopy, emerged in the early twenties. However, the first promising results faded away when it
turned out that experimental uncertainties and interpretation difficulties did not allow straightforward
conclusions. As a result, in the sixties and seventies electron microscopy was thought to be the only
reliable method for characterising the microstructure and X-ray line profile analysis was almost for-
gotten. From the seventies onward, the appearance of dedicated synchrotron X-ray sources, better
laboratory X-ray generators, improved X-ray detectors and new, more appropriate fundamental the-
ories of diffraction by crystals containing defects, gave a new impetus to X-ray line profile analysis.
Today, the fast computer technology and greatly enhanced experimental possibilities brought back
X-ray line profile analysis as one of the most effective alternative methods to electron microscopy for
the characterisation of the microstructure of crystalline materials. The aim of the present dissertation
is to provide a comprehensive summary of the theoretical considerations, experimental and numerical
methods which were developed for the evaluation of X-ray diffraction patterns in terms of disloca-
tion density, dislocation character, crystalite size and size distributions, and planar defects (especially
stacking fault and twin boundary frequency). A computer software package has been developed which
allows to evaluate different diffraction patterns measured on different crystalline materials: powders,
bulk polycrystals, single crystalline samples or textured specimens. The different parts of the soft-
ware package offer different ways of the evaluation which can be optimally adapted to the particular

1
problem. For example, in the case of single crystals, the individual diffraction profiles of different
Bragg reflections can be analysed in terms of individual dislocation contrast (or orientation) factors.
Or, in the case of an ideally random powder specimen, the whole measured powder pattern can be
evaluated with the same philosophy as the structure refinement by the Rietveld method. In this case
the whole measured pattern is fitted by a theoretically constructed diffraction pattern which is ex-
pressed in terms of dislocation structures by average dislocation contrast factors and a crystallite size
distribution function. In the entire software package the evaluation is based on microstructural mod-
els in which strain is assumed to be caused by dislocations, size broadening is caused by coherently
scattering domains and planar defects are included on the basis of simulated diffraction profiles.
The software package was applied successfully to more than a dozen cases of completely different
materials which have been evaluated for the microstructure, cf. [S1-S23]. The most important facili-
ties provided by the software package are the following.

(i) It allows the fitting of the separated diffraction profiles or their Fourier transforms by theoreti-
cal intensity profiles or Fourier coefficients calculated for the concomitant size and strain effect
(where strain is considered to be produced by dislocations). This part of the package is the Mul-
tiple Whole Profile (MWP) fitting procedure [S4, S6]. The dislocation densities, the character
of dislocations and the subgrain size distribution in copper deformed by equal channel angular
pressing (ECAP) [S4, S6] and ball milled PbS (galena) [S9] was evaluated by the MWP proce-
dure. Shape anisotropy in carbon-black [S4, S5, S6, S8] and crystallite size distribution and the
dislocation density in hexagonal Si3 N4 were also evaluated by the MWP method [S1].

(ii) Measured diffraction patterns can be fitted by numerically constructed diffraction patterns in
which the size and strain effects are theoretically calculated and can be convoluted with the
measured instrumental profiles. In this procedure all the different profile functions, i.e. the
size, the strain and the instrumental profiles are convoluted, therefore the method is called:
Convolutional Multiple Whole Profile (CMWP) fitting procedure [S14].

(iii) In both, the MWP and CMWP methods, the strain profiles can either be scaled by average
dislocation contrast factors, C, or by individual dislocation contrast factors, Chkl . Former cor-
responds to the case when averaging over the permutations of the hkl indices is appropriate,
whereas latter applies when the specimen is either a single crystal or strongly textured. When
strain anisotropy is treated by using the individual dislocation contrast factors, it means that the
strain part of each diffraction profile is scaled by a separate individual scaling parameter.

(iv) Balogh and coworkers have systematically analysed the effect of planar defects in cubic [S18]
and hexagonal crystals (Balogh et al., 2008). It was found that the planar faults profile can be
given as the linear combination of a delta function and one or more Lorentzian functions. The
systematic analysis provides the correlation between the planar fault densities and the different
parameters of the Lorentzian functions. The profile functions and the parameter files were
incorporated into the CMWP method [S18].

2
In a mechanically alloyed Al base Al-Mg alloys the subgrain size distribution, or coherently scattering
domain size distribution and the dislocation densities and character were determined by the CMWP
method [S14]. These microstructural parameters were evaluated and analysed as a function of the
Mg concentration. The diffraction patterns were also evaluated by the MWP method. The results
provided by the MWP and CMWP methods are critically discussed (see in section 5.3).
The diffraction patterns of different ball milled states of the CaF2 , SrF2 , BaF2 and CdF2 fluorides have
been analysed by the CMWP procedure in terms of dislocation densities and types, and crystallite size
and size distributions [S17]. An unusual X-ray optical interference effect of line broadening, similar
to what was observed earlier by Rafaja et al. (2004) is found in particular states of the ball milled
fluorides. The interference effect has been successfully corrected (see in section 5.5).

3
Chapter 2

Theoretical aspects of X-ray line broadening

In this chapter the different fundamental theoretical models of size and strain broadening and planar
faults are reviewed. The size profile and its Fourier transform corresponding to spherical domains or
crystallites with lognormal size distribution is calculated. The same is also determined for anisotropic
crystallite shape. The model of strain broadening is briefly presented for different dislocation dis-
tributions. Strain anisotropy is discussed and described in terms of the anisotropic strain effect of
dislocations. The concept of dislocation contrast factors is presented for different crystal systems.
The broadening caused by planar defects is briefly discussed. The models described here are the the-
oretical background of the two microstructural methods of line profile analysis described in detail in
Chapters 4 and 5.

2.1 Line profile parameters

The X-ray diffraction measurements usually provide the I(2θ) intensity profiles of the different reflec-
tions. In order to compare this with theory, it is practical to convert the variable 2θ to g, the variable
2 sin θ
of the reciprocal space, where g = is the absolute value of the diffraction vector. The value
λ
2 sin θB
of g at 2θB , the exact Bragg position, is denoted by gB and it is expressed as: gB = . In the
λ
following often g will be denoted also by K.
The variable of the intensity profiles can be expressed as:

2 cos θB
s = g − gB ≈ ∆θ. (2.1)
λ
The most important characteristic parameters of an I(s) intensity function corresponding to the Bragg
peak at 2θB are:

• the maximum intensity:

I0 = max{I(s) | s ∈ R}. (2.2)

• the Full Width at Half Maximum (FWHM):

4
 I0
FWHM{I(s)} = s2 − s1 , where s1 < s2 and I(s1 ) = I(s2) = . (2.3)
2
• the integral breadth (equivalent to the area of the normalised intensity curve):

R∞
I(s)ds
−∞
β= . (2.4)
I0

In X-ray diffraction the relation FWHM < β is usually satisfied.

2.2 Size broadening of diffraction profiles

With decreasing scattering volume the diffraction profiles broaden. This is called size broadening.
The X-ray measurements provide the coherently scattering domain size (crystallite size). This is
often smaller than the grain or subgrain size obtained by transmission electron microscopy (TEM),
however, Hansen and coworkers claim that when TEM micrographs are evaluated correctly then the
X-ray and TEM sizes correlate well (Winther et al., 2004). In Ungár et al. (2005) it was shown that
the coherently scattering domain size given by X-ray line profile analysis provides the subgrain or
cell size bounded by small angle grain boundaries or dipolar walls. In the present section the effect
of the coherently scattering domain size on the diffraction profiles is discussed.

2.2.1 Size parameters

The definition of the commonly used different size parameters (2.12), (2.13) and (2.16) is presented
through a simple case study (Warren, 1969). Let’s assume an infinite plane crystallite with the thick-
ness of N atoms. According to the theory of kinematical X-ray scattering, the line profile of this
special crystallite (Warren, 1969):
sin2 (N x)
I(s) ∼ , (2.5)
sin2 (x)
where x = πGa, G = g + ∆g, g is the diffraction vector, ∆g is a small vector, and a is the unit cell
sin2 (Nx)
vector chosen to be perpendicular to the plane of the crystallite. The function describes the
sin2 (x)
shape and position of the peaks in this special case. This function is plotted for different values of N
in Fig. 2.1. It has a maximum value at positions: x = nπ, n ∈ Z. This condition is equivalent to the
sin2 (Nx)
Laue equations. The maximum value of this function is lim = N2.
x→0 sin2 (x)

For large values of N this profile function can be approximated by the following simple function:
!2
sin2 (Nx) sin(Nx)
2
= N2 . (2.6)
sin x Nx

5
The FWHM value of this function is given by:

sin(Nx) 1
=√ . (2.7)
Nx 2

This transcendent equation can be numerically solved for Nx: the solution is Nx = 1.39, conse-
quently, the FWHM (2x) is reciprocally proportional to the number of lattice points perpendicular to
the diffracting plane:
1
FWHM = 2.78 . (2.8)
N
This means that the profile function becomes narrower as the crystallite becomes thicker. The integral
breadth of this curve is:

R∞ sin2 (Nx)
dx 1
−∞ x2
β= =π . (2.9)
N2 N

400
N=10
N=15
N=20
350

300 N=20
sin2(Nx)/sin2(x)

250

N=15
200

150

100 N=10

50

0
-0.6 -0.4 -0.2 0 0.2 0.4 0.6
x
sin2 (Nx)
Figure 2.1: The function plotted close to its first maximum for different values of N. As N
sin2 (x)
tends to infinity, the curve becomes a delta function.

A conventional θ - 2θ diffractometer measures the intensity parallel to the direction of the g diffraction
vector as a function of s = |∆g|. The FWHM and integral breadth value of the I(s) intensity function

6
as a function of s can be expressed as:

2.78 1 0.9
FWHM = = , (2.10)
π N a cos((G, a)∡) LG,a

1 1
β= = , (2.11)
N a cos((G, a)∡) LG,a
where LG,a = N a cos((G, a)∡) is the projection of the crystallite width in the direction of the diffrac-
tion vector. This means that by measuring the widths of the peak profiles, only the crystallite size
parallel to the diffraction vector can be determined.
0.9 1
For the reflection h00, LG,a is equal to the thickness of the crystal, i.e. either to or to . The
FWHM β
following two size parameters can be defined generally for an arbitrary I(s) intensity profile:

0.9
D= , (2.12)
FWHM

1
d= . (2.13)
β
Eq. (2.12) is the Scherrer (1918) equation. The theoretical description requires the Fourier transform
sin2 (Nx)
of the intensity profile too. The Fourier transform of the function is:
x2
N

 π(N − π|L|), if |L| ≤
π


(2.14)
N


 0, if |L| > .
π

Fig. 2.2 shows this function.

sin2 (Nx)
Figure 2.2: The Fourier transform of the function .
x2

7
For the infinite plane crystallite with the thickness of N atoms, the Fourier transform of the I(s) size
function is:   
N N
 LG,a N − LG,a |L| , if |L| ≤ LG,a



(2.15)



 0, if |L| > LG,a .
This means that LG,a can be determined from the initial slope of the Fourier transform. The size
parameter L0 is generally defined for an arbitrary I(s) intensity profile as the initial slope of the AS (L)
Fourier transform (Warren & Averbach, 1952):

AS (0) d 
− = AS (L) . (2.16)

L0 dL 
L=0

The definition of L0 is illustrated in Fig. 2.6 for the case of spherical crystallites with lognormal size
distribution.

In general the different size parameters satisfy the following relation: D ≥ d ≥ L0 (Langford &
Wilson, 1978).

For spherical crystallites d and L0 are proportional to the volume and area weighted average crystallite
size, respectively (Langford & Wilson, 1978):

∑ Vi di 4
i
< x >vol = = d, (2.17)
∑ Vi 3
i

∑ Ai d i 3
i
< x >area = = L0 . (2.18)
∑ Ai 2
i

2.2.2 Size distribution functions

In the previous section the size broadening of a single crystallite has been discussed. A polycrystalline
or fine powder sample consists of many crystallites with different sizes which can be characterised
by a size distribution function. By selecting the proper size distribution and assuming a realistic crys-
tallite shape, the size broadened profile can be calculated on a theoretical basis. Several distribution
functions can be used to describe the size distribution of crystallites (Langford et al., 2000; Scardi
& Leoni, 2002). Among these, one of the most flexible is the lognormal size distribution (Aitchison
& Brown, 1957), which was confirmed by several observations and successful applications (Valiev
et al., 1994; Terwilliger & Chiang, 1995; Krill & Birringer, 1998, Ungár et al., 1999; Langford et
al., 2000). The Gamma distribution (Arley & Buch, 1950) is also suitable to describe the experimen-

8
tal size distributions. York (1999) proposed another distribution. These distributions are discussed
below.

i) The lognormal distribution: this is the most commonly used size distribution of particle size in
a fine powder according to TEM observations (Aitchison & Brown, 1957). It can be shown that
a milling procedure leads to a lognormal size distribution (Hinds, 1982), therefore the lognor-
mal distribution is widely used in microstructural investigations. It is obtained by substituting
the variable of a normal distribution with its logarithm. When one applies the lognormal dis-
tribution to describe the size distribution of crystallites, this means that the logarithm of the
crystallite size follows a normal distribution. The density function of the lognormal size distri-
bution has the following form:
   2 
x
1 1 log
m
f (x) = √ exp − . (2.19)
 
2πσ x 2σ2

where m and σ are the parameters of the distribution, log m is the median and σ is the variance
of the normal distribution.
The parameters m and σ are called “median” and “variance” of the lognormal size distribution.
√
In the MWP and CMWP procedures [S4, S6, S14] the b = log m and c = 2 σ parameters are
used for the fitting procedure.

ii) The Gamma distribution:

The Gamma distribution (Arley & Buch, 1950) is also flexible and can be widely used for
describing observed crystallite size distributions. Its density function has the form:

a 
ax
a−1 
ax

f (x) = exp − , (2.20)
bΓ(a) b b

where a and b are the parameters of the distribution and Γ(x) is the Gamma function.

ii) York’s distribution: The York distribution (York, 1999) was obtained by assuming a normal
growth phenomena. The density function of the York distribution has a form similar to the
Gamma distribution:
1 
ax
a 
ax

f (x) = exp − , (2.21)
bΓ(a) b b
where a and b are the parameters of the distribution and Γ(x) is the Gamma function.

It should be noted here that Leoni & Scardi(2004) proposed a bar-diagram for the size distribution
density function and each individual column height of this diagram is fitted independently in their
pattern-refining procedure. Since this is an ad-hoc, experimental distribution its discussion is not
subject of this theoretical section. Langford et al. (2000) have shown that most of the above discussed

9
size distribution functions correlate well with experimental X-ray diffraction profiles and, that it is
difficult, if not impossible, to distinguish between size distribution functions on an experimental basis.

2.2.3 Determination of the size profile [S1, S4, S6]

By assuming a particular crystallite shape and crystallite size distribution, one can determine the
theoretical size profile. Bertaut (1949) and Guinier (1963) have shown, that the size profile of a
powder specimen consisting of crystallites with arbitrary size and shape can be determined as follows:

i) the crystallites should be divided into columns parallel to the diffraction vector g,

ii) the “size intensity profile” is obtained as the volume-weighted sum of the intensity profiles
normalized by their integral intensities corresponding to each column.

The intensity profile normalized by its integral intensity of a column with area Ai and height Mi is:

sin2 (Mi πs)

. (2.22)
Mi (πs)2

By summing up the contributions from all columns of all crystallites using the volume of the column
as weight, the intensity distribution becomes:

sin2 (Mi πs)

I(s) ∼ ∑ A i Mi . (2.23)
i Mi (πs)2

Let us introduce g(M) dM as the sum of the volumes of the columns with height between M and
M + dM from all crystallites:
g(M) dM = ∑ dV j (M, dM). (2.24)
j

Using this quantity, the intensity distribution can be expressed as:

Z∞ sin2 (M πs)
I(s) ∼ g(M) dM, (2.25)
M (πs)2
0

In this way, the size profile can be obtained by determining g(M) dM, which depends on the crystallite
shape and the size distribution of the crystallites. In the following, the size profile is calculated
according to the lognormal size distribution, and for spherical (i) and ellipsoidal (ii) crystallite shapes
[S4, S6]. Latter is a simple and general description of deviations from spherical shape. It also accounts
for anisotropic size broadening as a function of hkl.
For a particular crystallite shape g(M) dM is determined first for one crystallite. This calculation is
based on the geometrical properties of the crystallite shape. The size profile is obtained by summing
up for all crystallites using the crystallite size distribution function. In the case of spherical crystallites
and lognormal size distribution, g(M) dM is obtained as follows:
Using the notations of Fig. 2.3 the following geometrical equation can be written:

10
 2
M

2 2
x =y + , (2.26)
2
For one crystallite g(M) dM is equal to the volume of the part of the sphere with column length
between M and M + dM:

g(M) dM ≈ −2πy dy M. (2.27)

M dM
By differentiating eq. (2.26) the following is obtained: 2y dy = − . Therefore for one crystallite:
2

g(M) dM ∼ M 2 dM. (2.28)

Since f (x) dx is proportional to the number of the crystallites with diameter between x and x + dx and
all the crystallites with diameter x ≥ M contain the column length M:

Z∞
 

g(M) dM ∼  f (x) dx M 2 dM. (2.29)

M

Using the distribution density function in (2.19), this integral can be expressed as:
 
Z∞ M

1  log m 
f (x) dx = erfc  √ , (2.30)
2 2σ
M

where erfc is the complementary error function, defined as:

∞
2 Z −t 2
erfc(x) = √ e dt. (2.31)
π
x

Thus for all crystallites g(M) dM can be written as:


M

 log m 
g(M) dM ∼ M 2 erfc  √  dM. (2.32)
2σ

Using eq. (2.25) the following size function is obtained [S1]:

  
Z∞ M
sin (M πs)
2 log
m 
I S (s) = M erfc  √  dM. (2.33)

2
(πs) 2σ
0

11
 x x
x
y+dy
M M+dM
y

Figure 2.3: Determination of g(M) dM in the case of a spherical crystallite with radius x. The goal is
to calculate the volume of the part of the sphere with column length between M and M + dM. This
part of the sphere is approximated by an annulus based prism and it is expressed with M, x and y,
where y is the radius of the annulus.

1
m=1 nm, σ=0.71
m=1.65 nm, σ=0.71
0.9 m=2.72 nm, σ=0.71

0.8

0.7

0.6
I(s)

0.5

0.4

0.3

0.2

0.1

0
-0.6 -0.4 -0.2 0 0.2 0.4 0.6
s
Figure 2.4: The size function for spherical crystallites with lognormal distribution with fixed value of
σ = 0.71, as a function of s. The value of m varies for the different curves. The value of m is indicated
in the upper right corner of the figure.

12
 1
m=2.72 nm, σ=0.14
m=2.72 nm, σ=0.35
0.9 m=2.72 nm, σ=0.5

0.8

0.7

0.6
I(s)

0.5

0.4

0.3

0.2

0.1

0
-0.6 -0.4 -0.2 0 0.2 0.4 0.6
s
Figure 2.5: The size function for spherical crystallites with lognormal distribution with fixed value
of m = 2.72 nm, as a function of s. The value of σ varies for the different curves. The value of σ is
indicated in the upper right corner of the figure.

This size function is plotted for different values of m and σ in Figs. 2.4 and 2.5.

In order to speed-up the numerical calculations in the MWP and CMWP methods [S4, S6], the explicit
form of the Fourier transform of the size profile is required. In the following, the Fourier transform
of the theoretical size profile is determined [S4, S6].
sin2 (M πs)
According to eq. (2.14), the Fourier transform of the function is:
(πs)2

 M − |L|,
 if |L| ≤ M
(2.34)

0, if |L| > M.


13
Using this equation, the Fourier transform of the size function (2.33) can be expressed as:

R∞
AS (L) = 2 I S (s) cos(2πsL) ds =
 0
M
   
∞ ∞ log
m
R R sin2 (M πs)
=2  M 2 erfc  √2σ  dM  cos(2πsL) ds =
(πs)
0 0
M
  
R∞
 ∞  log (2.35)
R sin2 (M πs) m 
= M 2 2 cos(2πsL) ds erfc  √ dM =
(πs) 2σ
0 0
M
  
R∞ log
m 
M 2 − |L|M


= erfc  √ dM.
2σ
|L|

By using substitutions and partial integration this integral can be further simplified:
   
9 √ 2
  |L|
S
m3 exp ( 2σ)
4  log m 3√ 
A (L) = √ −
3
erfc 
 2σ 2 −
2σ

   
√ |L|
 log m
2
m2 exp ( 2σ) √ 
|L| erfc  √ − 2σ
2  2σ + (2.36)

  
|L|
|L|3  log m 
erfc  √ .
6  2σ 

Dividing AS (L) by the maximum value, the normalised size Fourier transform is obtained. The maxi-
mum value of AS (L) is:
9 √ 2
 
2 m3 exp ( 2σ)
4
AS (0) = . (2.37)
3
An example for the plot of the size Fourier transform is shown in Fig. 2.6. It is noted that instead
of calculating the Fourier transform of eq. 2.23, Dr. Gubicza calculated the size Fourier transform
(Ungár et al., 2001) according to the following direct equation for AS (L) given in Guinier (1963):

1Z
AS (L) = σ(r)σ(r + L) d3 r, (2.38)
V

14
where σ(r)=1 if r ∈ “scattering object” otherwise σ(r)=0. The two different calculations are leading
exactly to the same result for A(L) [S4, S6].

The formulae of I S (s) and its Fourier transform AS (L) enable to express the size parameters as a
function of the parameters of the distribution. The derivation of AS (L) gives the L0 size parameter:

5 √ 2
 
AS (0) 2 m exp ( 2σ)
4
L0 = − ! = . (2.39)
d AS (L) 3
dL
L=0

The maximum value of the size profile is:

 
Z∞ M

 log m 
I S (0) = M 3 erfc  √  dM. (2.40)
2σ
0

The integral of the normalized size profile leads to the size parameter d:

7 √ 2
 
1 I S (0) I S (0) 3 m exp ( 2σ)
4
d= = R∞ = = . (2.41)
β I S (s) ds AS (0) 4
−∞

According to eqs. (2.17) and (2.18), the following is obtained for the volume and area averaged
crystallite size:

< x >vol = m exp 3.5σ2

(2.42)

< x >area = m exp 2.5σ2

(2.43)

It is noted that these expressions for the volume and area averaged mean crystallite size are identical
to those given by Hinds (1982) and Langford et al. (2000).

2.2.4 Anisotropic size broadening [S1, S4, S6]

If the crystallite shape is spherical, the size function is isotropic and thus independent of the hkl
indices. If a non-spherical crystallite shape is supposed, the size function becomes anisotropic and
it depends on the hkl indices. In the following the size function is calculated for crystallites with
spherically ellipsoidal shape and lognormal size distribution [S4, S6].

15
 FT(sf(x, 4.10502, 0.378924))
AS(0) AS(0)-(AS(0)/L0)*x
100000

80000
Amplitude

60000

40000

20000

0
L0
0 20 40 60 80 100
Frequency [nm]
Figure 2.6: The theoretical Fourier transform of a size profile for spherical crystallites with lognormal
size distribution. The initial slope defining L0 is also indicated in the figure.

g z
z’

c
y
a α
a

y’

x=x’
Figure 2.7: The determination of the ellipsoidal size function. The radii of the spherical ellipsoid are:
′ ′ ′
a and c. There are two coordinate systems: x, y, z and x , y , z . The latter is the eigensystem. The two
coordinate systems are rotated by an angle of α around the axis x. The diffraction vector g is also
indicated. The goal is to determine g(M) dM based on the geometrical properties of the part of the
ellipsoid with column length between M and M + dM parallel to g.

16
The determination of g(M) dM consists of the following steps (see Fig. 2.7 for the notations).
c
1. construction of the equation of the ellipsoid with ellipticity ε = in the eigensystem denoted
′ ′ ′
a
by x , y , z .

2. transformation with rotation of angle α around the axis x into the system x, y, z (α is the angle
between the diffraction vector and the axis of revolution of the ellipsoid and z is parallel to g).

3. expression of the column length M: M(x, y) = z2 (x, y) −z1 (x, y), where z1 and z2 are determined
using the equation of the ellipsoid.

4. determination of the area, T (M), of the plane-curve determined by equation M(x, y) = M (which
is in fact an ellipse).

5. For one crystallite the following is obtained:

π
g(M) dM = M[T (M) − T (M + dM)] = h(ε, α) M 2 dM, (2.44)
2

where h(ε, α) is determined using the equation of M(x, y). It is noted, that in the case of a
sphere: h(1, α) = 1.

6. calculation of the maximum column length Mmax :

2a
Mmax = s   (2.45)
1
1 + 2 − 1 cos2 α
ε

7. summing up for all crystallites with the column length M and using the lognormal size distri-
bution density function f (x):
  s   
1
 M 1 + 2 − 1 cos2 α  
 
 log  ε 

 
  m 

 
 
g(M) dM ∼ h(ε, α)M erfc 
2
√  dM. (2.46)
 

 2σ 

 
 
 
 
 

17
The size function obtained in this way is identical to that corresponding to spherical crystallites, the
only difference is that the parameter m becomes hkl-dependent and the parameter m of the distribution
has to be substituted for the following hkl dependent expression:
mA
mhkl = s   , (2.47)
1
1 + 2 − 1 cos2 αhkl
ε

where mA is the parameter m of the size distribution.

If the relative orientation of the unit cell vectors to the axis of revolution of the ellipsoid are known,
cos αhkl can be expressed with the hkl indices.

For cubic crystals, if the axis of revolution is parallel to the unit cell vector a:

l
cos αhkl = √ (2.48)
h2 + k 2 + l 2

For hexagonal crystals, if the axis of revolution is parallel to the unit cell vector c:

l
cos αhkl = r . (2.49)
4 c2 2
(h + hk + k2 ) + l 2
3 a2

It is noted that Scardi & Leoni (2002) calculated the hkl dependent size broadening for different
polyhedra, which is another possibility to introduce size anisotropy.

2.3 Strain broadening

In a real crystal, due to the lattice defects, the atoms are displaced relative to their ideal position. For
this reason the reciprocal space may also be distorted and the condition of diffraction is satisfied not
only in the ideal positions of the reciprocal lattice points, but in a finite volume in their proximity.
This effect is called strain broadening. Warren & Averbach (1952) gave the Fourier transform of the
X-ray line profile if size and strain effects are present simultaneously:

A(L) = AS (L)AD (L), (2.50)

where the strain Fourier coefficients can be expressed in the following form:

AD (L) = exp −2π2 g2 L2 hε2g,L i ,

(2.51)

18
where g is the absolute value of the diffraction vector, hε2g,L i is the mean square strain, depending
on the displacement of the atoms relative to their ideal position, and the brackets indicate spatial
averaging.
According to the continuum theory of elasticity, the longitudinal strain parallel to the direction of the
g diffraction vector is defined as (Wilkens, 1970):

g L g g L g
u(r + ) − u(r − )
|g| 2 |g| |g| 2 |g|
εg (L, r) = , (2.52)
L
L g L g
where u is the displacement field and L is the distance of the points r + and r − . Thus the
2 |g| 2 |g|
mean square strain is obtained as:

(εg (L, r))2 d3 r

RRR
V
hε2g,L i = (2.53)
d3 r
RRR
V

For discrete atoms, the strain, εg,L is defined as:

g g
′ u(rj ) − u(rj′ )
( j, j ) |g| |g|
εg,n = , (2.54)
nl0
′ ′
where j and j are the indices of the atoms in the direction of g, n = j − j , L = nl0 and l0 is the
distance between two atoms in the direction of g (Warren & Averbach, 1952). The mean square strain
is obtained as the square of the strain averaged over all atomic pairs with the distance of nl0 :
′
( j, j )
∑ (εg,n )2
′
j− j =n
hε2g,L i = (2.55)
∑ 1
′
j− j =n

Several authors worked on the determination of the mean square strain, including Warren & Averbach
(1952), Krivoglaz & Ryaboshapka (1963) and Wilkens (1970). Warren and coworkers (1952, 1959)
assumed either random atomic displacements and/or stacking faults. Krivoglaz (1969) and Wilkens
(1970) assumed dislocations as the main source of peak broadening close to the fundamental Bragg
positions. Krivoglaz and Ryaboshapka (1963) assumed a totally random distribution of dislocations
in the entire crystal and for small L values obtained:
 2  
b D
hε2g,L i = πρC log , (2.56)
2π L

where D is the crystal size. The problem with this formula is that (i) it diverges as D tends to infinity
and (ii) using this strain function, the strain Fourier transform does not decay as L tends to infinity.
However, for small L values this logarithmic expression enables to estimate the dislocation density,

19
for more details see the description of the modified Warren-Averbach method in section 3.2.
The logarithmic behaviour of the mean square strain for small L values, as a general property of any
dislocation configuration was derived by a fairly exact method by Groma (1998, 2003).
A numerical code has recently been developed in [S23] to calculate the mean square strain and the
strain profile for an arbitrary sized (e.g. 100x100 nm) crystal box containing up to a few hundred
randomly distributed straight parallel edge dislocations. The numerical evaluation of eq. (2.53) has
proved the logarithmic L dependence of hε2g,L i for small L values in correlation with Krivoglaz (1969),
Wilkens (1970) and Groma (1998). Fig. 2.8 shows a typical example for 10 dislocations in a 50x50
nm crystal box.

0.0006
"strain.dat" using 1:3
-0.000135*log(x/37)

0.0005

0.0004
<ε2>

0.0003

0.0002

0.0001

0
10 100
L [nm]
Figure 2.8: The mean square strain calculated for 10 randomly distributed edge dislocations in a
50x50 nm crystal box. The x scale is logarithmic. The dashed line indicates that the calculated mean
square strain can be well described by a logarithmic function between approx. 10 and 25 nm.

Wilkens (1970) calculated the mean square strain in the entire L range assuming the so called restrict-
edly random distribution of dislocations. This calculation is discussed in somewhat more detail in the
next paragraph.

2.3.1 The Krivoglaz–Wilkens model of dislocations

Wilkens improved the model of Krivoglaz by introducing a length parameter, the effective outer cut
off radius of dislocations (R∗e ), instead of the crystal diameter, by this eliminating the logarithmic
singularity in the expression of the mean square strain in eq. (2.56). He assumed that the crystal
contains separate regions with diameter of R∗e , in which parallel screw dislocations are randomly

20
distributed. Within each of these regions the distribution of the dislocations is completely random
and they have a density of exactly ρ. There is no interaction between dislocations outside of these
regions. He called this special dislocation configuration restrictedly random distribution and the mean
square strain was derived in the following form for the entire L range (Wilkens, 1970):
 2  
b L
hε2g,L i = πρC f , (2.57)
2π R∗e

where b is the absolute value of the Burgers-vector, ρ is the dislocation density, C is the contrast factor
of the dislocations and f is the strain function. In the following f will be called the Wilkens function.
f has the following explicit form, see eqs. A6-A8 in Appendix A in Wilkens (1970):

7 512 1
 
f ∗ (η) = − log η + − log 2 + +
4 90π η

1 Rη arcsinV
 
2
1− 2 dV −
π 4η 0 V

1 769 1 41 2
h ip
+ η + η3 1 − η2 −
π 180 η 90 90 (2.58)
 
1 11 1 7 1 2 1
+ + η arcsin η + η2 , if η ≤ 1,
π 12 η2 2 3 6

512 1 11 1 1
h i
f ∗ (η) = − + log 2η 2 , if η ≥ 1,
90π η 24 4 η
 
L ∗ (η) and η = 1 exp − 1 L .
 
where f = f
R∗e 2 4 R∗e
The Wilkens function and its approximations for small and large values of L are plotted in Fig. 2.9.
In the Wilkens function the same logarithmic term is present as in the Krivoglaz model, but it does
not diverge with the crystallite size, since it depends on the correlation length parameter, R∗e , which
is a finite parameter of the distribution. A common property of the Krivoglaz and Wilkens models is
that the mean square strain has a singularity for small values of L. This singularity does not affect the
shape of the line profiles considerably, since in the Fourier transform it is multiplied by L2 and this
multiplier strongly compensates the divergence: L2 log L → 0, if L → 0.
It is noted that in previous papers Re = exp(2) R∗e = 7.4 R∗e was used as the effective outer cut–off
radius of dislocations (Ungár et al. 1984; Wilkens, 1988; Hecker et al., 1997).
Inserting eq. (2.57) into (2.51), one obtains the strain Fourier–transform:

πb2 2
  
D 2 L
A (L) = exp − (g C)ρL f . (2.59)
2 R∗e

21
 7
f*(η)
f1(η)=7/4-log(2)-log(η)
f2(η)=(512/(90*pi))*(1/η)
6

3
f*(η)

-1

-2
0 2 4 6 8 10
η
1 1 L
 
Figure 2.9: f ∗ (η), the Wilkens function (where η = exp − ) and two of its approximations:
2 4 R∗e
7
f1 (η) = − log 2 − log(η), the function describing its logarithmic singularity at η → 0 and f2 (η) =
4
512 1
, the hyperbolic function describing the tails of the function.
90π η

22
Kamminga and Delhez (2000) have shown using numerical simulations that the line profile calculated
by the Wilkens model is also valid for edge dislocations and curved dislocations.
Wilkens (1978) introduced M ∗ , the dislocation arrangement parameter, which is a dimensionless pa-
rameter:
√
M ∗ = R∗e ρ (2.60)

The M ∗ parameter characterizes the dislocation arrangement:

• if the value of M ∗ is small, the correlation between the dislocations is strong

• if the value of M ∗ is large, the dislocations are distributed randomly in the crystallite

Fig. 2.10 shows two dislocation configurations: in the first the dislocations are strongly correlated
and the value of M ∗ is small and in the second the correlation is weak and M ∗ is large. Fig. 2.11
presents the strain profile for fixed ρ and variable M ∗ values.
It is noted that the Fourier transform given by the Wilkens model is real, this means that its inverse
Fourier transform, the strain intensity profile is symmetric, which is not always fulfilled in case of
real measurements, e.g. in single crystals with dislocation cell structure, the line profiles can be
asymmetric. Several different dislocation configuration can produce asymmetric broadening of the
profiles (Ungár et al., 1984; Mughrabi et al., 1986; Gaál, 1973, 1976, 1984). In Ungár et al. (1984)
and Mughrabi et al. (1986) authors propose a special cell-wall structure for dislocations: a matrix
structure where the dislocation density in the dislocation walls is significantly larger (so the material
is harder) than in the cell interiors. This model is called “Mughrabi’s composite model”. In section 5.2
an implementation of this model for X-ray line profile analysis is presented. Gaál developed a model
for a far more general distribution of dislocations, in which a configuration of polarised dislocation
dipoles leads to asymmetric strain broadening (Gaál, 1973, 1976, 1984). It should also be noted that
not only strain broadening causes asymmetric broadening of the line profiles, but planar faults also
introduce asymmetry in the profile shape, for more details see section 2.4.

23
 R∗e << √1ρ R∗e >> √1ρ
M ∗ << 1 M ∗ >> 1

Figure 2.10: Schematic representation of dislocation configurations and the corresponding strain pro-
file for small and large values of the M ∗ parameter. The y scale of the strain profiles is logarithmic.

1
M*=0.1
M*=1.0
M*=10.0

0.8

0.6
Intensity

0.4

0.2

0
0 0.5 1 1.5 2 2.5
s [1/nm]

Figure 2.11: The shape of the strain profile for fixed ρ and variable M ∗ values. It is noted that for
each profile the s variable is normalized by the FWHM value of the profile.

24
2.3.2 Strain anisotropy: the concept of contrast factors
Strain anisotropy means that the broadening of the profiles show an anisotropic behaviour as a func-
tion of the hkl indices: the width of the profiles is not a monotonous function of the length of the
diffraction vector or its square, see for example the Williamson-Hall plot (Williamson & Hall, 1953)
in Fig. 3.1. This is explained by the anisotropy of the mean square strain: hε2g,L i depends on the hkl
indices. This dependence is described by the contrast (or orientation) factors C. The value of the
contrast factors depend on the elastic constants of the material and on the relative orientation of the
diffraction vector (g), the Burgers vector (b), the line vector (l) and the normal vector of the slip plane
(n) of the dislocations. The contrast factors of dislocations characterize the “visibility” of disloca-
tions in the diffraction experiments. For example if bg = 0 for a dislocation, it has no, or almost no
broadening effect on the line profile. While the contrast factors can be determined experimentally for
single crystals, in a polycrystalline sample only the averages of C can be observed. For these type of
materials, strain anisotropy can be well accounted for by the average contrast factors of dislocations.
See for example the modified Williamson-Hall plot (Ungár & Borbély, 1996) in Fig. 3.2.
It has been shown by Ungár & Tichy (1999) that if the specimen is either untextured or if all possible
slip systems are equally populated, the average contrast factors can be expressed by the fourth order
polynomials of the hkl indices. For cubic crystals (Ungár & Tichy, 1999):

C = Ch00 (1 − qH 2), (2.61)

where
h2 k 2 + h2 l 2 + k 2 l 2
H2 = 2
. (2.62)
(h2 + k2 + l 2 )
For hexagonal crystals (Ungár & Tichy, 1999):

C = Chk0 (1 + a1 H12 + a2 H22 ), (2.63)

where
[h2 + k2 + (h + k)2 ] l 2
H12 = ,
[h2 + k2 + (h + k)2 + 32 ( ac )2 l 2 ]2

(2.64)
l4
H22 = 2 ,
[h + k2 + (h + k)2 + 32 ( ac )2 l 2 ]2

and ac is the ratio of the two lattice constants.

For orthorhombic crystals (Ungár & Tichy, 1999):

Chkl = Ch00 H02 + a1 H12 + a2 H22 + a3 H32 + a4 H42 + a5 H52 ,

(2.65)

25
where
h4
4
H02 =  2 a2 2 2 ,
h +k +l
a2 b2 c2

k4
b4
H12 =  2 2 2 2 ,
h
+k +l
a2 b2 c2

l4
c4
H22 =  2 ,
h2 k2 l 2
+ +
a2 b2 c2
(2.66)
h2 k2
a2 b2
H32 =  2 2 2 2 ,
h +k +l
a2 b2 c2

l 2 h2
c2 a2
H42 =  2 2 2 2 ,
h +k +l
a2 b2 c2

k2 l 2
b2 c2
H52 =  2
h2 k2 l 2
+ + .
a2 b2 c2

Here it is noted that a formally similar equation to (2.61) has been derived for random displacement
of atoms in elastically anisotropic cubic crystals by Stokes & Wilson (1944). However, it has not
been evaluated until Ungár and Tichy (1999) have rediscovered and further evaluated the correlation
between dislocations and strain anisotropy.
The constants Ch00 and Chk0 are calculated on the basis of the crystallography of dislocations and
from the elastic constants of the crystal (see: Ungár et al, 1999). The parameters q, a1 , a2 , . . ., a5
are the same for all reflections. These parameters are related to the edge or screw character of the
dislocations. In Ungár et al. (1999) the value of these contrast factor parameters were calculated for
different materials and for different type of dislocations. For example the q parameter were calculated
and plotted as a function of the AZ Zener constant for pure edge and pure screw dislocations. The
Williamson-Hall, MWP or CMWP procedure provide a measured value of the q parameter, which
usually falls between the values of the q parameter calculated for edge or screw dislocations. If the
value of the measured q parameter is close to the q value calculated for edge dislocations, then the
character of the dislocations is edge, if it is close to the q value calculated for screw dislocations, then
the character is screw and if its value is close to the average of the edge and screw q values, then the
character of dislocations is mixed. One can see from eq. (2.59) that the diffraction order dependence
of the strain Fourier–transform is given by g2C. A numerical code (the ANIZC program) has been
developed by Dr. Borbély (2003) for the calculation of the individual or average contrast factors taking
into account the elastic constants of the material, the lattice parameters and the relative orientations
of the l, b, n and g vectors. In addition to X-ray line profile analysis, the theory of dislocation
contrast factors can also be applied successfully for evaluating neutron diffraction experiments (see
e.g. Somogyvári et al., 2001).

26
2.3.3 The Groma-Csikor strain function
The logarithmic approximation of the strain function for small L values given by eq. (2.56) was
determined by several authors: Krivoglaz (1969) and Wilkens (1970) derived it for special dislocation
distributions and Groma (1998) proved its validity for a general distribution of dislocations. This
1
means an intensity profile which decays asymptotically as 3 (Groma, 1998).
s
Gaál (1984) determined the (complex) strain Fourier transform of a random configuration of polarised
screw dislocation dipoles. Groma & Monnet (2002) calculated A(L) corresponding to a distribution
of infinitesimal polarised dislocation dipoles and derived a negative exponential function for the strain
Fourier transform. The strain Fourier transform calculated by Wilkens (1970) has the same asymptotic
behaviour for large L values. This means a Lorentzian intensity profile.
Groma and Csikor proposed a simple interpolation function which connects smoothly these two func-
tions (Groma, 2003; Csikor & Groma, 2004):
 2
b
hε2g,L i = π Chkl f GR−CS (L) , (2.67)
2π

where:
!
X Y
f GR−CS (L) = W log 1 + + , (2.68)
|L| 1 + ZL2
and W , X , Y and Z are parameters of the function. Actually the authors have derived this formula for
the P(τ) distribution function of the internal stresses caused by dislocations, which is analogous to
the I(s) intensity distribution used in X-ray line profile analysis.
X
For L → 0 this function tends to: W log +Y and by comparing the expression of hε2g,L i to that given
|L|
by Krivoglaz (2.57) the following equation is obtained for ρdisl. , the dislocation density, and Re , the
effective outer cut-off radius of dislocations:
X Re
W log ρ
disl. disl.
+Y = Ch00 log (2.69)
|L| |L|

For L → ∞ the function takes the form:

!
1 WX2 Y 1
WX + − + . (2.70)
|L| 2 Z L2

Since the profile function is area normalised, in other words A(0) = 1, the second term should disap-
pear: !
WX2 Y
− + = 0. (2.71)
2 Z
This means, that the strain function depends only on three independent parameters, therefore one of
2Y
the four parameters can be eliminated, e.g.: Z = .
WX2

27
The first term in eq. (2.70) can be compared to the following expression depending on adip. , the
ρdisl.
dipole distance and on ρdip. = , the density of dislocation dipoles:
2
1 dip. 1
WX = Ch00 adip. ρdip. (2.72)
|L| |L|

From these equations, the following expressions are obtained for the values of the physical parameters
disl. and C dip. are the dislocation and dipole contrast factors, respectively):
(Ch00 h00

W
ρdisl. = , (2.73)
disl.
Ch00

Y
 
Re = X exp , (2.74)
W
disl.
Ch00
dip.
a = 2X , (2.75)
dip.
Ch00
W
ρdip. = . (2.76)
disl.
2Ch00
It is noted that the Groma-Csikor strain function is implemented is the CMWP program package
[S14], however it is not always possible to determine so many strain parameters from the measure-
ments, especially if size broadening is also present. In some cases, e.g. if the quality of the measured
data is low, or if the tails of the profiles cannot be measured because of the strong overlapping between
peaks, it is even not possible to determine ρ and R∗e accurately.
Eq. (2.68) provides a simple and flexible strain function, which can also be used to approximate
the Wilkens function  if less independent
 parameters are used, e.g. by writing eq. (2.68) in the form
X Y 1 1 L
 
f GR−CS∗ (η) = W log 1 + + , where η = exp − , the parameters W, X, Y and
|η| 1 + Zη2 2 4 R∗e
Z can be determined by fitting these parameters to the data points of the f ∗ (η) Wilkens function.
The resulting function gives a perfect description of the Wilkens function, see the graphs of the
fitted f GR−CS∗ (η) function and the Wilkens function in Fig. 2.12 (also cf. Fig. 2.9). The fitting
provided the following parameters: W = 1.14908, X = 1.4125, Y = 0.212226 and the value of Z
2Y
is = 0.18514. By using these parameters a good approximation of the Wilkens function is
WX2
obtained which is even simpler than the following approximate formula given by van Berkum (1994):

7 η2 32η3
 
f ∗ (η) = − log η + − log 2 + − if η ≤ 1,
4 6 225π
(2.77)
256 1 11 1 1
h i
f ∗ (η) = − + log 2η 2 , if η ≥ 1.
45π η 24 4 η

28
 7
f*(η)
fGr-Cs*(η)
6

3
f(η)

-1

-2
0 2 4 6 8 10
η

1 1 L
 
Figure 2.12: the Wilkens function (where η = exp −
f ∗ (η), ) and f GR−CS∗ , the Groma-
2 4 R∗e
Csikor strain function fitted to the data points of the Wilkens function. The fitting provided the
following parameters of the Groma-Csikor strain function: W = 1.14908, X = 1.4125, Y = 0.212226
2Y
and Z = = 0.18514. This strain function gives a perfect approximation of the Wilkens function.
WX2

2.4 Planar faults [S18]

According to the numerical calculations of Mr. Levente Balogh, based on the theoretical work of
Warren (1969) and the numerical code developed by Treacy et al. (1991), the peak broadening due to
stacking faults for an hkl reflection in cubic fcc crystals can be expressed in the following form:

phkl
1 phkl
2 phkl
3
0 δ(s) +
st
I (s) = phkl 2 + 2 + 2 , (2.78)
s − shkl s − shkl s − shkl
  
1+ 1 1+ 2 1+ 3
whkl
1 whkl
2 whkl
3
where the first term is given by a Dirac delta function, the others are broadened and shifted Lorentzian
functions, whkl
i are the FWHM values of the Lorentzian functions and shkl i are the shifts of the
Lorentzian functions. Both wi and si can be expressed as fifth order polynomials of the α j proba-
hkl hkl

bility of planar faults, where j stands for intrinsic or extrinsic stacking faults or twin boundaries:

1 = W1,1 α j +W1,2 α j +W1,3 α j +W1,4 α j +W1,5 α j ,

whkl hkl hkl 2 hkl 3 hkl 4 hkl 5
(2.79)

2 = W2,1 α j +W2,2 α j +W2,3 α j +W2,4 α j +W2,5 α j ,

whkl hkl hkl 2 hkl 3 hkl 4 hkl 5
(2.80)

29
 3 = W3,1 α j +W3,2 α j +W3,3 α j +W3,4 α j +W3,5 α j ,
whkl hkl hkl 2 hkl 3 hkl 4 hkl 5
(2.81)

1 = S1,1 α j + S1,2 α j + S1,3 α j + S1,4 α j + S1,5 α j ,

shkl hkl hkl 2 hkl 3 hkl 4 hkl 5
(2.82)

2 = S2,1 α j + S2,2 α j + S2,3 α j + S2,4 α j + S2,5 α j ,

shkl hkl hkl 2 hkl 3 hkl 4 hkl 5
(2.83)

3 = S3,1 α j + S3,2 α j + S3,3 α j + S3,4 α j + S3,5 α j .

shkl hkl hkl 2 hkl 3 hkl 4 hkl 5
(2.84)

The real and imaginary parts of the Ast (L) Fourier transform of I st (s) can be expressed as:

ℜAst (L) = phkl hkl hkl hkl hkl hkl hkl

0 + p1 cos(2πs1 L) exp(−πw1 L) + p2 cos(2πs2 L) exp(−πw2 L)+

+phkl hkl hkl

3 cos(2πs3 L) exp(−πw3 L) (2.85)

ℑAst (L) = phkl hkl hkl hkl hkl hkl

1 sin(2πs1 L) exp(−πw1 L) + p2 sin(2πs2 L) exp(−πw2 L)+

+phkl hkl hkl

3 sin(2πs3 L) exp(−πw3 L). (2.86)

It is noted, that since this Fourier transform is complex, the intensity profile (obtained by inverse
Fourier transforming this expression) is asymmetric. For each type of materials and stacking faults,
the values of phkl hkl hkl
i , Si, j and Wi, j are determined for different hkl values by using the program DIFFAX
developed by Treacy et al. (1991). The phkl hkl hkl
i , Si, j and Wi, j parameter values have been evaluated for
fcc crystals for the first 15 reflections and for the three fundamental planar fault types by Mr. Levente
Balogh [S18].

30
Chapter 3

The classical methods of X-ray line profile

analysis

This chapter is a summary of the methods of X-ray line profile analysis developed by other au-
thors. The information about the microstructure is in the width and the shape of the diffraction
profiles. Therefore the methods of X-ray line profile analysis can be divided into two main groups:
(i) breadth methods and (ii) whole profile or pattern fitting methods. A brief summary is given for
the Williamson-Hall breadth method and the Warren-Averbach Fourier procedure. The so called full-
profile and whole-pattern methods of other authors and the moment-methods are shortly discussed.

3.1 Breadth methods

The Williamson-Hall procedures are based on the different order dependence of the different physical
effects causing the broadening of diffraction profiles. In the case of spherical crystallites the size
broadening of profiles is constant and isotropic in the whole reciprocal space, while strain broaden-
ing increases with K, the reciprocal space coordinate. If strain is caused by dislocations, the strain
broadening is usually anisotropic and can be interpreted by the concept of dislocation contrast (or
orientation) factors (Ungár & Borbély, 1996; Ungár & Tichy, 1999). If planar or stacking faults are
present, the broadening is also anisotropic, however it does not increase with K. Here it is noted that
anisotropic crystallite shape can also introduce anisotropy in the broadening ([S4, S6, S5]; Scardi &
Leoni, 2002).

In the Williamson-Hall procedure (Williamson & Hall, 1953), the ∆K values defined as the FWHMs
(a) or integral breadths (b) of the measured profiles are plotted as a function of K. This is called the
Williamson-Hall plot. Such a plot gives a qualitative idea about the causes of peak broadening: if the
sample consists of small crystallites, the ∆K value at K = 0 is high, while if the crystallites are large,
this value is negligible or small. Since at K = 0 no strain broadening is present, the corresponding ∆K
value is due only to size broadening and according to eqs. (2.12) and (2.13) this ∆K value is equal to
0.9 1
(a) or (b), respectively. Here it is noted that this value can only be determined by extrapolation
D d

31
from the measured data points, since there is no Bragg peak at K = 0.
The slope of the curve gives qualitative information on the strength of the strain effect. If the effect is
strong, ∆K increases strongly with K, while if the strain is weak, the ∆K values are almost constant
or the slope is small. In order to estimate the dislocation density, the slope can also be interpreted
quantitatively, however it is better to use the Warren-Averbach (Warren & Averbach, 1952) method
(or a more advanced full profile fitting procedure, see e.g. the methods described in Chapters 4 and
5) for this purpose.
If strain is caused by dislocations, the widths of peaks show a typical anisotropic behaviour (see for
example for deformed polycrystalline copper sample Fig. 3.1) and the modified Williamson-Hall
procedure (Ungár & Borbély, 1996) can be used to interpret strain in terms of dislocation contrast
√
factors. In this procedure the widths are plotted as a function of K C or K 2C leading to a consider-
ably smoother curve for materials containing dislocations, see for the same deformed polycrystalline
copper sample Fig. 3.2. The strain anisotropy parameters can be determined by using a simple linear
regression procedure. In previous papers (e.g. Ungár & Borbély, 1996), it is suggested that the widths
√
depend quasi-linearly on K C. In recent papers [S4, S6], it is suggested that the widths depend
quasi-linearly on K 2C. This is based on the Wilkens model (Wilkens, 1970), which is a very powerful
model for describing the strain properties of materials containing dislocations. The Wilkens strain
√
profile depend only on the powers of K 2C, so it is an even function of K C, therefore the widths
√
(FWHM or integral breadth values) of the peak profiles are also even function of K C. As long as
the basic assumptions of the Wilkens model are suitable for the investigated materials, the widths
√
can depend only on K 2C and not on K C. The modified Williamson-Hall plots have been plotted
as a function of K 2C for several materials, and it has been found that the data can be very well fitted
linearly as a function of K 2C. For a demonstration, see for example Fig. 3.3 for the same deformed
polycrystalline copper sample as plotted in Figs. 3.1 and 3.2. However there is no guarantee that
the assumptions of the Wilkens model are always satisfied and for some particular sample a linear
√
dependence on K C is also conceivable. The Williamson-Hall methods use only the widths of the
profiles which is a very limited portion of the information in the profiles. However, these simple and
powerful methods are well suitable to qualitatively investigate the materials.

32
 0.06

0.05

0.04
∆K [1/nm]

0.03

0.02

0.01

0
0 2 4 6 8 10 12
K [1/nm]

Figure 3.1: Classical Williamson-Hall plot for the integral breadths (deformed polycrystalline copper
sample). In this figure the β values are plotted as a function of K.

0.06
0.0163661+0.000308631K2C

0.05

0.04
∆K [1/nm]

0.03

0.02

∆KS

0.01

0
0 2 4 6 8 10 12
¬
K√C [1/nm]
Figure 3.2: Modified Williamson-Hall plot for the integral breadths√(deformed polycrystalline copper
sample). In this figure the β values are plotted as a function of K C. The data points are fitted by
a parabolic curve, the fitted parameters of the parabola are also indicated. The value of the fitted
function at K = 0, ∆K S is the breadth of the pure size function.

33
 0.06
0.016366+0.000308633K2C

0.05

0.04
∆K [1/nm]

0.03

0.02

∆KS

0.01

0
0 20 40 60 80 100 120
2 2
K C [1/nm ]
Figure 3.3: Modified Williamson-Hall plot for the integral breadths (deformed polycrystalline copper
sample). In this figure the β values are plotted as a function of K 2C. The data points are fitted by
a straight line, the fitted parameters of the line are also indicated. The value of the fitted function at
K = 0, ∆K S is the breadth of the pure size function.

34
3.2 Fourier methods
In the Warren-Averbach procedure, the Fourier transforms of the profiles are analysed. In the clas-
sical Warren-Averbach plot (Warren & Averbach, 1952) the normalized Fourier transforms of the
measured peak profiles are calculated for equidistantly sampled values of L in a range [0, Lmax ] and
log A(L) is plotted as a function of K 2 for each L value. The Warren-Averbach plot shows usually
a similar anisotropic behaviour as the Williamson-Hall plot. In the modified Warren-Averbach plot,
the measured data points can be plotted smoothly as a function of the scaling parameter K 2C. It is
demonstrated in Figs. 3.4 and 3.5.
The values of the physical parameters can be determined as follows. Using eqs. (2.50), (2.51) and
(2.56), for small values of L, the logarithm of the Fourier transform of the profiles can be approxi-
mated as:  
Re
log A(L) ≈ log A (L) − ρBL log
S 2
(K 2C), (3.1)
L
πb2
where B = . In this method, the parameters in the contrast factors are not fitted, but are fixed to
2
particular values (obtained for example by the modified Williamson-Hall procedure).

For each value of Li , log A(Li ) is fitted by the parabolic curve ai + bi K 2C + ci K 4C2 and by plot-
ting the exponential of the ai values as a function of L, the size Fourier transform is obtained. By
fitting the initial slope of the size Fourier transform (using linear regression) one can get the value of
L0 , see Fig. 3.6.

The initial slopes bi divided by L2i are plotted as a function of log Li , as shown in Fig. 3.7. By
using linear regression on the linear part of this data, the initial slope is equal to ρB, from which
the dislocation density, ρ, can be determined. The other parameter of the regression line is equal to
−ρB log Re from which the value of Re can be determined.
b
The applicability of this method is limited by the fact that in some cases when plotting 2i as a
Li
function of log Li , the curve has no linear part (the logarithmic formula is not valid for larger L values
and even for small values of L for materials containing dislocations a “hook-effect”, a deviation from
the logarithmic function can be observed, see e.g. Wilkens, 1970), so the determination of ρ and Re
becomes very unstable.

When fitting log A(Li ), the term ci K 4C2 indicates the presence of a higher order term in hε2g,L i.
This term is used only for describing the curve more accurately, the value of ci is not used at all.

35
 0

-1

-2
log(A(L))

-3
L=0
L=2.84864
L=5.69728
-4 L=8.54592
L=11.3946
L=14.2432
L=17.0918
L=19.9405
L=22.7891
-5 L=25.6378
L=28.4864
L=31.335
L=34.1837
L=37.0323
-6
0 5 10 15 20 25 30 35
K2 [1/(nm)2]

Figure 3.4: Warren-Averbach plot (deformed polycrystalline copper sample). In this figure log A(L)
is plotted as a function of K 2 for each L value. The anisotropic behaviour of the data points can also
be observed.

-1

-2
log(A(L))

-3
L=0
L=2.84864
L=5.69728
-4 L=8.54592
L=11.3946
L=14.2432
L=17.0918
L=19.9405
L=22.7891
-5 L=25.6378
L=28.4864
L=31.335
L=34.1837
L=37.0323
-6
0 5 10 15 20 25 30 35
K2C [1/(nm)2]
Figure 3.5: Modified Warren-Averbach plot (deformed polycrystalline copper sample). In this figure
log A(L) is plotted as a function of K 2C for each L value. The data points are fitted by parabolic curves,
which are also plotted. The log A(L) values at L = 0 correspond to the pure size Fourier transform.

36
 1.2

0.8
AS(L)

0.6

0.4

0.2

L0
0
0 10 20 30 40 50 60
L [nm]
Figure 3.6: The size Fourier Transform obtained by the modified Warren-Averbach method (deformed
polycrystalline copper sample). In this figure the AS (L) values, the values of the size Fourier trans-
form, are plotted as a function of L. The data points are fitted by a line and the L value of this line
corresponding to AS (L) = 0 is the parameter L0 .

1e-04

-0.0001

-0.0002
bi/L2i

-0.0003

ρB
-0.0004

-0.0005
- ρB log Re
-0.0006
0 0.5 1 1.5 2 2.5 3 3.5 4
log(Li)
Figure 3.7: The plot for determining ρ and Re by using the modified Warren-Averbach method (de-
b
formed polycrystalline copper sample). In this figure the 2i values are plotted as a function of log(Li ).
Li
The linear part of the data points is fitted by a line and the slope of this line gives the value of ρ and
b
the 2 value of this line at L = 0 gives the parameter Re .
L

37
3.3 Full profile fitting methods
Instead of the breadth methods, whole profile methods have been developed recently for the determi-
nation of microstructural parameters. In Chapter 4 the Multiple Whole Profile (MWP) fitting method
[S4, S6] will be presented in detail, which is in fact a Fourier method working on the whole (Fourier)
profiles. In the MWP method the Fourier transforms of the measured profiles for all reflections are
fitted simultaneously with ab-initio theoretical functions. Its theoretical basis for including the effect
of size and strain simultaneously is the convolutional equation (2.50) given by Warren & Averbach
(1952). According to the literature, some of the full profile methods other than the MWP method are
briefly presented here:

1. Nusinovici & Rehfeldt-Osierski (1990) developed the program PROFILE which uses a method
called pattern decomposition. By using this program one can determine interactively line profile
parameters by fitting the measured peaks or clusters of peaks with analytical functions. The
following analytical functions have been incorporated in their program: Voigt, pseudo-Voigt
and Pearson VII.

2. Louer & Audebrand (1999) developed the program ProfFOU. In their method all reflections are
fitted simultaneously by using pseudo-Voigt functions. The microstructural parameters can be
determined by analysing the parameters of the pseudo-Voigt functions.

3. Dong & Scardi (1999) developed a similar program, called MarqX which also provides individ-
ual profile parameters.

4. A similar approach is the method of Whole Powder Pattern Fitting (WPPF) developed by Scardi
and Leoni (1999). In this method an analytical function (e.g. a Voigtian) is adopted to fit the
experimental profiles and the profile parameters are connected by suitable conditions (Gaus-
sian and Lorentzian widths in the case of a Voigtian profile function) to the microstructural
parameters.

It is noted, that unlike these methods, the MWP method does not provide individual profile param-
eters, the microstructural parameters for size and strain effect are obtained directly from the fitting
procedure which is using ab-initio theoretical profile functions depending only on the microstructural
parameters.

3.4 Pattern fitting methods

As discussed in the previous section, most of the earlier whole profile fitting methods are based
on fitting the measured profiles by analytical profile functions. The most common fitting functions
are the Voigt, the pseudo-Voigt and the Pearson VII profile functions which are scaled to the hkl
anisotropy by ad-hoc scaling parameters, as suggested by Caglioti et al. (1958). These procedures
have several deficiencies: (i) the analytical profile functions usually do not describe the physical

38
profiles correctly over the entire intensity range, they either fit around the maxima or in the tail regions,
but usually do not fit well the two regions simultaneously, (ii) there is no unique correlation between
the microstructural parameters of the materials and the profile parameters of the analytical functions.
Therefore the method of Convolutional Multiple Whole Profile (CMWP) fitting [S14] has been devel-
oped in which the profile functions are based and constructed by using fundamental physical princi-
ples applied to the different types and kinds of lattice defects. The CMWP method works directly on
the measured pattern instead of the Fourier transform of the separated individual profiles (as it was the
case in the MWP method). The theoretical basis for including the effect of size and strain simultane-
ously is the convolutional equation (2.50) given by Warren & Averbach (1952), just like in the case of
the MWP method. In the CMWP method, a model based pattern is compared directly to the measured
pattern using a nonlinear least squares procedure. As the MWP method, the CMWP method does not
need individual peak parameters either, the microstructural parameters are determined directly from
the fitting procedure. For the details of the CMWP procedure, see Chapter 5.
In this section a few introductory comments are made about the literature, regarding some of the
pattern fitting methods other than the CMWP procedure.

1. Langford et al. (2000) proposed a pattern fitting method for nanocrystalline distortion-free
materials. They applied it successfully to the powder pattern of nanocrystalline CeO2 and the
size distribution function has been determined by fitting.

2. Scardi & Leoni (2002) developed a similar procedure which is called Whole Powder Pattern
Modeling (WPPM) which uses also ab-initio theoretical functions, but does not account only for
the size effect, but also includes theoretical model for several sources of strain. Its theoretical
basis for including the size and strain effect simultaneously is also the convolutional equation
(2.50) of Warren & Averbach (1952).

3.5 Methods analysing the moments of the profiles

These single profile methods are based on the different asymptotic behaviour for the different sources
of broadening of the higher order moments of the scattered intensity. The first model was the variance
method proposed by Wilson (1962). Borbély and Groma (2001) proposed recently a very powerful
momentum method which deals correctly with measurements when both strain and size broadening
are present and also the level of the background can be determined in a physically correct way. It
uses a general theory for the asymptotic behaviour of dislocation induced X-ray peak broadening, it
does not depend on a particular model of dislocation arrangement like the Wilkens model, only the
general s−3 dependence of the tails of the intensity profile is used. The size broadening is accounted
for by the s−2 dependence of the size profile for the case of small crystallites. Unlike the multi-profile
fitting methods, this method can be applied to single profiles, especially for single crystals and the
dislocation density can be estimated in this case too. However, in order to evaluate the tails of the

39
profiles accurately, this method requires the peaks to be measured with good statistics which is not
always available.
The k’th moment of an intensity profile I(s) is defined as (Groma, 1998):

Rs k
S I(S)dS
−s
Mk (s) = R∞ (3.2)
Sk I(S)dS
−∞

Borbély & Groma (2001) gave the following asymptotic expansion for the second and fourth order
moments:
1 L Λ<ρ> s
 
M2 (s) = s− + log , (3.3)
π2 εF 4π2 K 2 εF 2π2 s0
2 s
 
M4 (s) 1 Λ<ρ> 3Λ2 < ρ2 > log
s1
= s+ + , (3.4)
s2 3π2 εF 4π2 4π2 s 2

π
where Λ = g2 b2C and εF is the area-weighted average crystallite size, for spherical crystallites:
2
3
εF = L0 .
2

By plotting the moments as a function of s one can readily see the type of broadening present in
the experiment and verify if the assumptions of small particle size or the presence of dislocations
apply. So, using this method one can easily visualize qualitatively the different sources of broaden-
ing as well as the size parameter εF and the strain parameters < ρ > or < ρ2 > can be determined
quantitatively by a fitting procedure.
An example is presented for ECAP deformed copper sample, the second order moment is plotted in
Fig. 3.8 and the fourth order moment divided by s2 is shown in Fig. 3.9. The moments (calculated by
using the measured profile of the 111 reflection) and the fitted theoretical moment functions according
to eqs. (3.3) and (3.4) are plotted.

40
Figure 3.8: The second order moment and the fitted theoretical function (ECAP deformed copper
sample). In the figure s is denoted by q.

Figure 3.9: The fourth order moment divided by s2 and the theoretical function fitted between the two
markers (ECAP deformed copper sample). In the figure s is denoted by q.

41
Chapter 4

Determining microstructure by the Multiple

Whole Profile fitting method

The essential parts of this dissertation are the two methods of X-ray line profile analysis: the Multiple
Whole Profile (MWP) [S4, S6] and the Convolutional Multiple Whole Profile (CMWP) [S14] fitting
procedures. The MWP procedure and a couple of problems solved by this method are described and
discussed in detail. It is shown that the MWP method enables to evaluate X-ray diffraction profiles in
terms of crystallite size and size distribution and dislocation densities and dislocation character.

4.1 The MWP method [S4, S6]

4.1.1 The principles of the method

The aim of the MWP procedure [S4, S6] is to describe the shape and the width of measured profiles
on the basis of the microstructure. The microstructural parameters, especially the crystallite size and
size distribution, the dislocation density and arrangement parameter and strain anisotropy, are refined
by a nonlinear least squares procedure, the Marquardt-Levenberg algorithm (Levenberg, 1944 and
Marquardt, 1963). Since it is working on individual profiles, the measured pattern first has to be
separated into single diffraction peaks. A special program, called MKDAT, was developed for this
separation procedure [S4].

4.1.2 Data preparation: profile separation using the MKDAT program

In the MKDAT procedure the selected overlapping region of the measured intensity pattern is fitted by
the sum of a background polynomial and simple analytical functions (Gauss, Lorentz, Pearson VII,
Pseudo-Voigt or Voigt). The peak positions, the peak intensities and the shape parameters are fitted
by using the nonlinear Marquardt-Levenberg least squares algorithm. For each reflection the sum of
the background polynomial and the analytical functions corresponding to the unrequired reflections is
subtracted from the measured data and only the peak profile corresponding to the required reflection
remains. The parameters of these analytical functions are not interpreted physically, these functions

42
are used only to describe the shape of the overlapping region. This procedure is correct if the peak
shapes are close to the shape of the analytical functions used in the fitting. This may depend on the
microstructural properties of the materials, however practice has shown that in most cases these sim-
ple analytical functions are adequate to describe the shape of the unrequested measured peaks. This
separation procedure should be used only for data preparation and this is not the procedure intended
for evaluating the microstructure. If the overlapping between the peaks is weak, the separation proce-
dure is simpler: in this case only the background (represented by a polynomial) is subtracted from the
measured data and there is no need for the above mentioned analytical functions, so the data and the
results of the MWP evaluation procedure are not influenced by the shape of any analytical function.
Fig. 4.1 is an example for the separation of overlapping peaks.

1 M3377.dat
bg(x)+I(x)
bg(x)
p0+p1x
I1(x)
I2(x)
0.8 I3(x)

0.6
Intenzitas
’

0.4
100

101
0.2
004

-2 -1.5 -1 -0.5 0 0.5 1

∆K [1/nm]

Figure 4.1: Typical example for the separation of strongly overlapping peaks in the case of carbon
black sample. The background polynomial is given as p0 + p1 x and the measured data are fitted by
the function p0 + p1 x + I1 (x) + I2 (x) + I3 (x), where: I1 (x), I2 (x) and I3 (x) are Pearson VII functions.
For the 100 peak, the peak profile is determined by subtracting bg(x), the sum of the background
polynomial and the other Pearson VII functions (004 and 101) from the measured data. After the
subtraction, the remaining data correspond to the pure, separated peak. The same procedure can be
applied to obtain the other peaks.

4.1.3 Data preparation: instrumental deconvolution using the MKDAT pro-

gram
If instrumental broadening is present, the peaks should be corrected for the instrumental effect too.
In the MKDAT program a simple deconvolution (based on the method of Stokes, 1948) is used for

43
this purpose. In this procedure the Fourier transforms of the measured peak, Am (L), and the Fourier
transforms of the corresponding instrumental peak, Ai (L), are calculated. By using complex division,
Am (L)
A ph (L) = i , the Fourier transform of the physical and pure microstructural peak profile is calcu-
A (L)
lated. The instrumental free intensity profile can also be determined by calculating the inverse Fourier
transform of A ph (L).

4.1.4 Evaluation of X-ray diffraction profiles using the MWP program

Since the MWP method is a microstructural method, the peak intensities are not interpreted physi-
cally, so the measured intensity profiles and their Fourier–transforms, as well as the fitting theoretical
functions are normalized by their maximum values in the fitting procedure.
The theoretical basis of the microstructural analysis is the ab-initio theoretical Fourier–transform
given by equations (2.50), (2.58), (2.59), (2.61), (2.64), (2.36) and (2.47)-(2.49). There are two
possible approaches:

(i) Multiple Whole Profile fitting of the Fourier–transforms. In this procedure first the mea-
sured intensity profiles are Fourier–transformed and normalized. Then all of them are fitted
simultaneously by the normalized theoretical Fourier–transform:

AS (L) πb2 2
  
2 L
A(L) = S exp − (g C)ρL f , (4.1)
A (0) 2 R∗e

where AS (L) and AS (0) are given by eqs. (2.36) and (2.37), respectively.

(ii) Multiple Whole Profile fitting of the intensity profiles. In this procedure first the measured
intensity profiles are normalized. Then all of them are fitted simultaneously by the normalized
theoretical intensity function:
Fc (s)
I(s) = , (4.2)
Fc (0)
where Fc is the Cosine Fourier–transform of (5.9), which is equivalent to the inverse Fourier
transform of A(L), since A(L) is a real and even function:

Z∞
Fc (s) = 2 A(L) cos(2πLs) dL. (4.3)
0

In the MWP fitting procedure all profiles are fitted simultaneously using the nonlinear Marquardt-
Levenberg least-squares algorithm, in which the Weighted Sum of Squared Residuals (WSSR) is
minimised. In this procedure the profiles are weighted uniformly. Here it is noted that the profiles
which correspond to the same g value but their hkl indices are not permutations of each other (e.g.
the 333 and 511 reflections in the case of fcc cubic materials) should be omitted from the evaluation
procedure, because their contrast factors are different and the theoretical Fourier–transform (2.59)
corresponds to only one contrast factor. If the intensity ratio of these coincident reflections is known,

44
they can be included in the CMWP procedure, see Chapter 5 for more details. The theoretical func-
tions depend on the hkl indices and on the microstructural parameters: m and σ, the parameters of the
crystallite size distribution, ρ, the dislocation density, R∗e the dislocation arrangement parameter and
the strain anisotropy parameters q (or a1 , a2 for hexagonal crystals). These microstructural parameters
are the common parameters which connect the different reflections during the simultaneous fitting.
The difference between the theoretical profiles is in their hkl dependence which is scaled by g2C de-
pending on the q (or a1 , a2 ) anisotropy parameters. The values of the microstructural parameters are
refined during the fitting procedure. If the ellipsoidal size function is used, the value of the ellipticity
parameter, ε is refined too. The other parameters, the lattice constants, a (and c in the hexagonal
case), the absolute value of the Burgers-vector (b) and Ch00 or Chk0 are input parameters of the fitting
procedure. Since the strain Fourier–transform depends on the product b2 ρCh00 (or b2 ρChk0 in the
hexagonal case), the value of b and Ch00 (or Chk0 ) should be known in order to determine the value
of ρ. The value of Ch00 (or Chk0 ) depend on the elastic constants of the materials and on the edge or
screw character of the dislocations. However, in most cases the type of dislocations is only known af-
ter the fitting procedure, by analysing the contrast factor parameters (q or a1 , a2 ) which are the result
of the fitting procedure. In this case an estimated value of Ch00 (or Chk0 ) should be given during the
fitting procedure, and the value of Ch00 (or Chk0 ) should be determined according to the resulting q or
a1 , a2 parameters. After this, the procedure should be repeated in order to get the correct value for ρ.
For the method of determining the Ch00 (or Chk0 ) constants, see Ungár et al. (1999). If the value of b
or Ch00 (Chk0 ) cannot be determined, the value of the microstructural parameter ρ becomes uncertain
up to a scaling factor.

4.1.5 The steps of the MWP evaluation procedure

The function of the program evaluate is to evaluate the previously prepared data for the microstruc-
tural parameters using the method of Multiple Whole Profile fitting. The steps of the evaluation
procedure are:

1. Selection of the crystal system. The possible selections are: cubic or hexagonal. In the following
the parentheses indicate the case of hexagonal system.

2. Setting the value of the input parameters. The program prompts for the value of the lattice
constant(s), the absolute value of the Burgers-vector and Ch00 (or Chk0 ).

3. Selection of the size function. The possible selections are: no size effect, spherical size function
(default) or ellipsoidal size function.

4. Specification of the sampling of the Fourier–transform of the measured data. The number of
samples (default: 256) and the upper limit of sampling (the default is: five times the maximal
FWHM value of the Fourier profiles) may be specified.

5. Computing the normalized Fourier–transform of the measured data. This step is skipped if the
Fourier–transforms are available from the instrumental correction (see the description of the

45
 program MKDAT, sections 4.1.2 and 4.1.3) or from previous runs. The Fourier–transforms are
saved after their computation.

6. Specification of the initial values of the parameters and the limit of stopping. The initial values
of the fitting parameters are always saved for subsequent runs. The limit of stopping is the con-
vergence criteria of the Marquardt-Levenberg algorithm: the fitting is stopped if the specified
maximal number of iterations (default: 5000) is reached or if the relative change of the WSSR
between two iteration steps is less than the specified limit (default: 10−9 ).

7. Selection of the method of fitting. The procedure of Multiple Whole Profile fitting can be carried
out by using the Fourier–transforms or the intensity profiles.

8. Fitting. In this step the values of the parameters are refined using the gnuplot program ex-
tensively modified for the MWP and CMWP procedures. The measured profiles and the fitted
theoretical functions are plotted side by side in order of g and are replotted in each step of it-
eration, so one can trace continuously how the theoretical profiles approach the measured data.
The figure is saved after the fit converges. A typical plot for fitting the Fourier–transforms is
shown in Fig. 4.3.

9. Printing of solutions. In addition to the resulting parameters m, σ, ε, ρ, R∗e , q (or a1 , a2 ) the

√
dislocation arrangement parameter M ∗ = R∗e ρ (introduced by Wilkens) and the size parame-
ters D, d and L0 are printed too. These size parameters are defined by equations (2.12), (2.13)
and (2.16), and they are calculated from the pure theoretical size profile using the fitted m and
σ values.

10. Preparation of figures. If the Fourier–transforms were fitted, the program plots the intensity
functions too. If the intensity functions were fitted, the program plots the Fourier–transforms
too.

11. Saving the results and removal of temporary files.

It is noted that the actual fitting parameters used by the MWP and CMWP methods are denoted by a,
b, c, d, e and these parameters are related to the microstructural parameters by the following equation:

m = exp(b),
c
σ = √ ,
2
q = a,
2
ρ = ,
π(bBURGERS d)2
exp(− 41 )
R∗e = ,
2e
(4.4)

46
4.1.6 The MWP frontend
The program MKDAT and a WWW frontend for the evaluate program of the MWP procedure is available
on the web: http://www.renyi.hu/mwp. The program MKDAT was developed for UNIX systems,
and its goal is to prepare the measured data for the MWP evaluation procedure including the steps
of peak separation and instrumental correction as described in sections 4.1.2 and 4.1.3. The MWP
frontend provides a platform-independent interface of the procedure: after uploading the previously
prepared data, the options of the fitting procedure can be selected (as described in section 4.1.5) and
the evaluation procedure can be started.
Fig. 4.2 shows the control panel of the MWP frontend. A detailed description of the frontend can be
found at the MWP documentation page:
http://www.renyi.hu/mwp/doc

Figure 4.2: The control panel of the MWP frontend: http://www.renyi.hu/mwp

47
4.2 MWP application to deformed Cu [S4, S6] and ball milled
PbS [S9]
The MWP method was applied to polycrystalline copper samples of 99.98% purity deformed by ECA
(Equal Channel Angular) pressing and by removing a surface layer of 100µm with chemical etching
[S4, S6]. The results of the MWP fit can be seen in Fig. 4.3. The procedure provided the following
1
values for the microstructural parameters: m = 56 nm, σ = 0.34, q = 1.64, ρ = 1.5 · 1015 2 and
m
Re = 5.1nm [S4, S6].

Figure 4.3: Example for the MWP fit (copper sample). In this figure the Fourier transforms of the
measured and fitted profiles are plotted as a function of L. The peaks corresponding to the different
hkl indices are plotted one after the other. The L ranges corresponding to the different peaks are
independent, and extend from 0 to 200 nm for each peak.

The MWP procedure can be used to determine the microstructure of systematically prepared samples.
Such an example is the fcc galena: a series of natural PbS (galena) samples were ball-milled and in
some cases heat treated. These samples were studied by X-ray diffraction together with samples from
ancient Egyptian make-ups fabricated from galena (Walter, 1999; Martinetto et al., 2000). The diffrac-
tion patterns were measured by the high resolution powder diffraction beam line, BM16 of ESRF in
Grenoble, France (Martinetto et al., 2000). By evaluating the X-ray patterns of the archeological and
systematically prepared samples a map of the microstructure was obtained. The microstructural pa-
rameters are considered as fingerprints of the state of the materials. By comparing the microstructural
parameters of the archeological samples to that of the ball-milled and/or heat treated galena samples,
the ancient manufacturing practices were traced back [S9]. It was found that the ancient specimens

48
have been gently crushed and no heat treatment was used or the heat treatment was carried out at
temperatures not higher than about 300 ◦ C.
Fig. 4.4 shows the MWP fit of the archeological sample E23105. In this case the procedure provided
the following values for the microstructural parameters: m = 87 nm, σ = 0.72, q = −4.5, ρ = 1.7 ·
1
1016 2 and Re = 10nm.
m

Figure 4.4: The MWP fit of the archeological galena sample E23105. The A(L) values of the measured
and fitted profiles are plotted as a function of L.

PbS, galena is an ionic crystal, where it is not trivial if dislocations are created by plastic deformation
(Sprackling et al., 1976). For this purpose, Martinetto and coworkers (2002) have carried out detailed
transmission electron microscopy (TEM) on ball milled PbS crystals. A typical TEM micrograph in
Fig. 4.5 shows the extended dislocation network in one of the deformed PbS specimens (Martinetto
et al., 2002).

49
Figure 4.5: PbS coarse particle bright-field TEM micrograph of a dislocation network densely con-
nected.

50
4.3 MWP application to Al-Mg alloys [S13, S14]
In [S13, S14] the effect of the nominal Mg content and the milling time on the microstructure and the
hardness of mechanically alloyed Al-rich Al-Mg solid solutions were studied.
A series of Al-Mg samples were prepared from high purity aluminium (99.9%) powder and high
purity magnesium chips (less than 2 mm thick and 5 mm long). The mechanical alloying was carried
out using a Spex8000 shaker miller at room temperature.
Aluminium with 6 wt.% magnesium alloys were milled for periods of 0.5, 1, 3 and 6 h. Additionally,
a series of Al - x wt.% Mg (x=0, 3, 6) were milled for 3 h at the same conditions to study the effect
of the nominal Mg content. The milled powders were compacted at a pressure of 1 GPa in air at
room temperature without any lubricant. The microstructure of both the ball-milled powders and
the compacted specimens were studied by X-ray line profile analysis. The diffraction profiles were
recorded by a Philips X’pert powder diffractometer using Cu anode and pyrolithic graphite secondary
monochromator. Step size and step time were 0.030 and 22 s/step, respectively. The X-ray diffraction
peak profiles were evaluated for the crystallite size distribution and the dislocation structure by the
MWP fitting procedure. The measured and the fitted Fourier transforms are shown in Fig. 4.6 for the
Al - 6 wt.% Mg specimen after ball milling for 0.5 h.

Figure 4.6: The MWP fit of the Al - 6 wt.% Mg specimen after ball milling for 0.5 h. The A(L) values
of the measured and fitted profiles are plotted as a function of L. The difference plot is also given at
the bottom of the figure.

51
The volume-weighted crystallite size, the dislocation density, the hardness and the Mg concentration
as a function of the milling time and the nominal Mg content are plotted in Figs. 4.7 and 4.8, re-
spectively. For the Al - 6 wt.% Mg samples the Mg concentration increases with milling time up to
3 h. After 3 h there is no change in the Mg concentration. The dislocation density increases while
the crystallite size decreases rapidly with milling time up to 1 h. Between 1 and 6 h milling time the
dislocation density and the crystallite size do not change significantly, as shown in Fig. 4.7. In the
case where the nominal Mg content changes and the milling time period is fixed to 3 h, the Mg con-
centration in the solid solution and the dislocation density increase while the crystallite size decreases
and the crystallite size distribution becomes wider with increasing nominal Mg content, as shown in
Fig. 4.8.
The hardness of the compacted samples was also measured by depth-sensing Vickers indentation
tests. For hardness measurements the ball milled powders were compacted. To check the effect of
compaction on the microstructure, the parameters of the crystallite size distribution and the dislocation
structure of the compacted specimens were determined by X-ray diffraction peak profile analysis.
Comparing the microstructural parameters obtained before and after compaction it is concluded that
this procedure has only slight effects on the microstructure of the ball milled specimens. This means
that the hardness measured on the compacted samples characterizes also the mechanical properties of
the milled Al-Mg alloys.
By analysing the hardness values for the compacted specimens it was found that with increasing
milling time, the hardness increases up to 3 h due to the increasing Mg concentration, the increase of
the dislocation density and the decrease of crystallite size. Upon increasing milling time from 3 to 6 h
the hardness does not change significantly since neither the Mg concentration nor the microstructure
changes. It is found that the Mg concentration in Al-Mg alloys has a considerable effect on the
defect structure. With increasing nominal Mg content of the powder or increasing milling time, the
dislocation density increases and the character of dislocations is shifted toward edge type. These
changes can be attributed to the increase of the solute Mg concentration which increases the pinning
effect of Mg atoms on edge dislocations thus hindering their annihilation. After 3 h milling the Mg
concentration in solid solution is much higher than the equilibrium solubility limit. The hardness
obtained for the mechanically alloyed Al 3 wt.% Mg specimen is in good correlation with those
predicted from the Hall-Petch relationship (Hall, 1951; Petch, 1953) determined for bulk samples in
other papers.

52
Figure 4.7: The volume-weighted mean crystallite size in nanometers (open squares), the dislocation
density in 1014 m−2 (solid squares), the Mg concentration in wt.% (open triangles) and the hardness
in 108 Pa (open circles) as a function of the milling period for the specimen of Al-6wt.% Mg nominal
composition.

Figure 4.8: The volume-weighted mean crystallite size in nanometers (open squares), the dislocation
density in 1014 m−2 (solid squares), the Mg concentration in wt.% (open triangles) and the hardness
in 108 Pa (open circles) as a function of the nominal Mg content for the specimens ball milled for 3h.

53
4.4 MWP application to Ti [S12, S21] and Si3N4 [S1, S6]
Hexagonal Ti samples were investigated in [S12]. Nanocrystalline titanium specimens were produced
by severe plastic deformation (ECA pressing) and X-ray diffractograms were measured by a Philips
X’pert diffractometer. The profiles were evaluated using the MWP method and the procedure pro-
vided the following values for the microstructural parameters: m = 38 nm, σ = 0.49, q1 = −0.05,
1
q2 = 0.18, ρ = 8.6 · 1014 2 and M = 6.5. The results of the MWP fit can be seen in Fig. 4.10.
m
The large dislocation density of ρ ≃ 9 × 1014 m−2 is confirmed by detailed high resolution TEM as
shown in the micrograph in Fig. 4.9 (Zhu et al., 2003).

Figure 4.9: (a) High resolution TEM image of a low-angle grain boundary with a misorientation
of 6.5◦ , (b) Fourier-filtered image from the white frame in (a), showing the dislocation arrangement
in the grain boundary.

(m) (m)
From the analysis of the measured q parameter values, q1 and q2 , which are easily derived from the
(m) (m) a1
a1 and a2 fitting parameters according to equations q1 = a1 and q2 = − 2  2 + a2 (Dragomir
c
3 a
& Ungár, 2002), the active slip system types can be concluded [S21]. In the ECAP deformed Ti
specimen it was found that the dominant active slip system types are hai and hc + ai [S11, S22], in
good correlation with TEM observations (Paton et al., 1970).

Another hexagonal samples were investigated in [S1, S6]. Commercial Si3 N4 samples (powder LC12
from Starck Ltd, Germany) were investigated by X-ray diffractometry. The results of the MWP fit
can be seen in Fig. 4.12.
Fig. 4.11 shows the TEM micrograph of the Si3 N4 particles and the particle size distributions mea-
sured either by TEM or X-ray line profile analysis. A good correlation between the two size distribu-
tions determined by two completely different methods can be seen.
The microstructural parameter values determined by X-ray line profile analysis are listed in Table 4.1.

54
Figure 4.10: The MWP fit of the hexagonal Ti sample. The A(L) values of the measured and fitted
profiles are plotted as a function of L.

Table 4.1: Microstructural parameters obtained for nanocrystalline Si3 N4 determined by the MWP
method

m [nm] σ ρ [1014 m−2 ] M a1 a2

20 (3) 0.65 (5) 7.5 (8) 2.1 3.54 -1.93

Figure 4.11: (a) TEM micrograph of the Si3 N4 particles, (b) particle size distributions measured by
TEM and X-ray line profile analysis.

55
Figure 4.12: The MWP fit of the hexagonal Si3 N4 sample. The A(L) values of the measured and fitted
profiles are plotted as a function of L. The difference plot is also given at the bottom of the figure.

56
4.5 MWP evaluation of carbon black samples by assuming ellip-
soidal crystallite shape [S4, S5, S6, S8]
The microstructure of hexagonal carbon blacks were investigated in [S5, S8]. Carbon black has the
turbostratic layer structure, therefore the evaluation could only be carried out by assuming that the
coherently scattering domains are of ellipsoidal shape flat crystallites (Warren, 1965). The measured
X-ray profiles were evaluated using the MWP procedure with the additional assumption that the crys-
tallites are rotational ellipsoids (see section 2.2.4 for more details about the ellipsoidal size function).
Untreated, heat treated and compressed at 2.5 GPa carbon black specimens were investigated. The
measured X-ray profiles were evaluated using the MWP procedure and assuming ellipsoidal crystal-
lite shape. The value of ε, the ellipticity parameter is found between 0.5 and 1.2.

Figure 4.13: The MWP fit of carbon black sample. The A(L) values of the measured and fitted profiles
are plotted as a function of L. The difference plot is also given at the bottom of the figure.

57
Chapter 5

Determining microstructure by the

Convolutional Multiple Whole Profile fitting
method

The CMWP procedure and a couple of problems solved by this method are described and discussed
in detail. It is shown that the CMWP method enables to evaluate X-ray diffraction patterns in terms
of crystallite size and size distribution, dislocation densities and dislocation character, without the
necessity to separate overlapping diffraction peaks or to deconvolute instrumental effects. It is shown
that the CMWP method makes it also possible to obtain the density and character of stacking faults
or twin boundaries. A special application of the CMWP procedure to solve the problem of unusual
narrowing of the first few diffraction peaks in ball milled fluorides is also presented.

5.1 The CMWP method [S14]

The Convolutional Multiple Whole Profile (CMWP) fitting procedure works as a computer program
developed for the determination of microstructural parameters from diffraction profiles of materials
with cubic, hexagonal or orthorhombic crystal lattices [S14], but works in principle, for all crystal
systems. Unlike the MWP method, which uses the Fourier transform of the separated profiles, the
CMWP method works directly on the measured intensity pattern and the separation of the profiles is
not needed. Another difference is, that there is no need to correct the measured data for the instru-
mental effect. The instrumental effect is added to the theoretical (physical) pattern by convolution,
thus avoiding numerical division of small numbers.

5.1.1 The principles of the method

The whole measured powder diffraction pattern is fitted directly by the sum of a background function
and profile funtions obtained as the convolution of ab-initio theoretical functions for size, strain and
planar faults and the measured instrumental profiles [S14]:

58
 Itheoretical = BG(2θ) + ∑ IMAX
hkl hkl
I (2θ − 2θhkl
0 ), (5.1)
hkl

where BG(2θ) represents the background, IMAXhkl is the peak intensity, 2θhkl is the 2θ value at the peak
0
hkl
center and I is the theoretical profile for the hkl reflection, which can be expressed as the convo-
hkl , the measured instrumental profile and ab-initio profile functions: I hkl , the theoretical
lution of Iinstr. size
hkl hkl
size profile, Idisl. , the theoretical strain profile for dislocations and I pl. f aults , the theoretical profile
function for planar faults:

I hkl = Iinstr.
hkl hkl
∗ Isize hkl
∗ Idisl. hkl
∗ I pl. f aults (5.2)

This convolutional equation is based on equation (2.50) given by Warren & Averbach (1952). The
equation can be easily extended to include other physical effects as further sources of broadening. In
order to get realistic results, not all the effects should be included without consideration in the eval-
uation procedure, it is better to have some initial knowledge about the nature of the effects causing
the broadening and only the physically relevant effects should be included in the evaluation. There
are several possibilities to obtain this preliminary knowledge about the nature of the microstructural
effects. For example TEM observations can be used or the qualitative information provided by the
Williamson-Hall plot or the momentum method on one of the reflections can be used. In the calcula-
tion of the theoretical functions it is assumed that the crystallites have spherical or ellipsoidal shape
with lognormal size distribution and strain is caused by dislocations. Strain and size anisotropy are
taken into account by the dislocation contrast factors and the ellipticity of crystallites, respectively.
The mathematical formulae together with the dependence on the microstructural parameters of these
theoretical functions are given in Chapter 2. Note that in most cases the Fourier transform of the
profiles is expressed in an analytical form, which is favourable, since in the CMWP method, the
convolution is carried out in the Fourier-space and the I hkl profiles are obtained by inverse Fourier-
transforming the product of the theoretical Fourier transforms and the (complex) Fourier transform
of the corresponding measured instrumental profiles. Another possibility would be the usage of the
definition of the convolution:
Z∞
( f ∗ g)(x) = f (t) g(x − t) dt, (5.3)
−∞

however, it is simpler and considerably faster for obtaining the theoretical intensity profile to directly
use the Fourier transforms provided by theory and performing an inverse Fourier transformation once
after the multiplication of the Fourier transforms, than inverse Fourier transforming all the theoretical
Fourier transforms and using the convolutional equation (5.3). Depending on the effects included,
the fitting procedure can provide the following microstructural parameters: median and the variance
of the size distribution, the ellipticity of the crystallites, the density and arrangement of dislocations,
the strain anisotropy parameters and the probability of planar faults. Since it is a microstructural

59
 hkl and 2θhkl values are not modeled on the basis of the crystal structure, they are
method, the IMAX 0
hkl
treated simply as free fitting parameters. The fit should be carried out in two steps: in (i) the IMAX
and 2θhkl
0 parameters are fitted together with the microstructural parameters (a few steps of iteration
hkl and 2θhkl parameters which is the goal of this step; in this
is usually enough to determine the IMAX 0
hkl and
step the values of the physical parameters are not interpreted) and in (ii) the previously fitted IMAX
2θhkl
0 parameters are fixed and only the physical parameters are refined.
The CMWP procedure treats the reflections which correspond to the same g value but their hkl indices
are not permutations of each other (e.g. the 333 and 511 reflections in the case of fcc cubic materials).
In this case both profiles are added to the model based, generated pattern and an intensity ratio is
introduced for this purpose (the total intensity at maximum is equal to IMAX hkl in this case too). By

avoiding the need for the profile separation and instrumental correction the CMWP method is free
from the uncertainties due to these steps and it is much more adequate and superiour in a sense than the
MWP method for evaluating full powder patterns with overlapping peaks. However, if the overlapping
of the profiles is weak and they are measured separately on an instrument with negligible instrumental
effect, like a Nonius FR 591 special double crystal high resolution diffractometer (see e.g. Ungár et
al., 1998), it is more convenient to apply the MWP method. In the case of separately measured
profiles, before applying the CMWP method the profiles should be linked into a pattern, the empty 2θ
regions between the peaks should be filled with artificial data and in this case the background of the
generated pattern usually has larger steps because of the different level of background of the different
peaks. This causes uncertainties when determining the background which is not present when using
the MWP procedure in these cases. The situation is the same for the evaluation of the measured
profiles of single crystals, where obviously no powder pattern can be measured. Another case when
the MWP method is more suitable than the CMWP when only the Fourier transforms of the profiles
are available. In the CMWP procedure the measured and model-generated pattern given by eq. (5.1)
are compared using the Marquardt-Levenberg nonlinear least squares algorithm (Levenberg, 1944
and Marquardt, 1963) and the fitting procedure provides the parameters of the ab-initio theoretical
functions as well as the refined theoretical powder pattern. Figs. 5.1 and 5.2 demonstrate the results
of the CMWP fitting procedure for Al-6Mg sample ball milled for 6 hours, the measured and the fitted
pattern are shown in both linear and logarithmic scale and their difference is shown in linear scale.
The results for the microstructural parameters are also indicated.

5.1.2 The JAVA frontend of CMWP

There are two main options to use the CMWP method: the first is to run directly the program
evaluate (which is part of the CMWP package) and the second is to use it via its WWW frontend.
Latter will be described in the next subsection. The evaluate program is a frontend written in the

60
 50000
measured data
theoretical curve
difference

40000
111

30000
Intensity

20000

200

10000
220 311

222 331 420

422
400
0

40 60 80 100 120 140

o
2θ [ ]

Figure 5.1: The measured (solid lines) and theoretical fitted (dashed lines) intensity patterns for Al-6Mg
sample ball milled for 6 hours as a function of 2θ. The difference plot is also given at the bottom of the figure.

100000
measured data
theoretical curve
Results of the
111 CMWP fit:
m = 21nm
200 σ = 0.36
ρ = 1016 m−2
√
10000 220 311 M = Re ρ = 1.3
q = 1.3
Intensity

222 331
420
422

400
1000

100
40 60 80 100 120 140
2θ [o]

Figure 5.2: The measured (solid lines) and theoretical fitted (dashed lines) intensity patterns for Al-6Mg
sample ball milled for 6 hours as a function of 2θ (plotted with logarithmic intensity scale). The results for the
microstructural parameters are also indicated.

61
shell script language zsh and it is enhanced with graphical functions written in the JAVA program-
ming language. Fig. 5.3 shows the JAVA panel of the evaluate frontend. All the parameters can be
adjusted in this panel and all the functions of the fitting can be reached by pressing the appropriate
buttons. Another JAVA tool, the mkspline program can be used for the interactive determination
of the base points of the background spline. See Fig. 5.4 for an example of the background spline
determined by mkspline.

Figure 5.3: The JAVA panel of the CMWP frontend

Figure 5.4: The mkspline program of the CMWP package

62
5.1.3 The WEB frontend of CMWP
The CMWP procedure can also be used via its WWW frontend:
http://www.renyi.hu/cmwp .
This latter option has many advantages compared to the previous one:

1. it is platform-independent, only an internet connection and a machine with a working web

browser is needed,

2. there is no need for installation,

3. the frontend is served by well configured and relatively fast machines,

4. the program and the code behind the frontend is continuously maintained.

Fig. 5.5 shows the control panel of the CMWP frontend, a more detailed description of the frontend
can be found at the CMWP documentation page: http://www.renyi.hu/cmwp/doc

Figure 5.5: The control panel of the CMWP frontend

63
5.1.4 The CMWP evaluation procedure
Using the evaluation program (by running evaluate or by using the WWW frontend) one can evalu-
ate the previously prepared and uploaded data for the microstructural parameters using the CMWP
method. The steps of the evaluation are the following:

1. Specification of the sample name.

2. Selection of the crystal system. The possible selections are: cubic (default), hexagonal or or-
thorhombic.

3. Setting the values of the input parameters. The value of the lattice constants, the absolute value
of the Burgers-vector, Ch00 (or Chk0 in the case of hexagonal crystal system) and the wavelength
of the measuring instrument have to be specified.

4. Specification of the instrumental profiles. If instrumental profiles are available, the inclusion of
the instrumental effect can be selected and the name of the instrumental profiles directory can
be specified.

5. Determining the background. The base points of the background spline should be given inter-
actively by using the mkspline program.

6. Peak searching. In this step the indices, the centers and the maximal intensities of all peaks
should be determined.

7. Specification of the interval used for fitting and plotting. Here can be specified the lower and
upper limit of the interval used for fitting and plotting (in 2θ degrees).

8. Selection of the size function. The possible selections are: no size effect, spherical size function
(default) or ellipsoidal size function.

9. Specification of the sampling of the theoretical Fourier transforms. The Alim amplitude ratio
limit of the normalized theoretical Fourier transforms can be specified, see section 5.1.6 for
more details. It is noted here that this limit should be specified only in case if no instrumental
profiles are available.

10. Specification of the sampling of the simulated powder pattern data. The value of N1, N2 and
the profile cutting parameter should be specified here. For the details of the meaning of these
parameters see sections 5.1.7 and 5.1.8.

11. Specification of the initial values of the parameters. The initial values of the fitting parameters
have to be specified. They are always saved for subsequent runs. There is an option for fixing
the value of any parameter, which means that the value of that parameter will not be refined
during the fitting procedure. Fixing the value of d · e means that the value of the parameter
M ∗ is fixed. This option can be used if the program enters into an asymptotic minimum (the

64
 values of ρ and R∗e tend to infinity and zero, respectively, and the asymptotic standard errors
of the parameters d and e become extremely large, larger than 100%). The scales of the fitting
parameters can also be specified here.

12. Peak parameter refinement and weighting. Here the refinement of the peak positions and peak
intensities can be chosen. If the peaks have very different intensities, you can choose here to
use weights in the fitting algorithm (see section 5.1.10 for more details about the weighting
algorithm).

13. Fit control. The limit of stopping is the convergence criterium of the Marquardt-Levenberg
algorithm: the fitting is stopped if the specified maximum number of iterations (default: 200)
is reached or if the relative change of the WSSR between two iteration steps is less than the
specified limit (default: 10−9 ).

14. Fitting. In this step the values of the parameters are refined using the gnuplot program exten-
sively modified for the MWP and CMWP procedures. The measured and the fitted theoretical
pattern are plotted and are replotted in each step of iteration, so one can trace continuously how
the theoretical pattern approaches the measured data. The figure is saved after the fit converges.
A typical plot for fitting is shown in Fig. 5.2.

15. Printing of solutions. The resulting physical parameters m, σ, ε, ρ, R∗e , q (or a1 , a2 ), α together
√
with the dislocation arrangement parameter M ∗ = R∗e ρ and the size parameters d and L0 are
printed.

16. Preparation of figures. The program plots the measured and theoretical fitted patterns together
with their differences in linear scale and the data is saved in a file containing 4 columns: 2θ, the
measured intensities, the fitted intensities and their difference. A plot with logarithmic scale is
also created.

17. Saving the results and removal of temporary files.

5.1.5 Determining the background

The background function BG(2θ) is given by an analytical function.

By giving the list of the relevant base points of the physical background, the background func-
tion is determined as the interpolated cubic spline corresponding to the specified base points. Fig.
5.6 shows an example for the spline background. The base points are indicated in Table 5.1. Fig. 5.4
shows the spline background of a real sample.

Another possibility is to use a sum of Legendre polynomials as the background. In this case, the
coefficients must be specified. The Legendre polynomials are orthogonal functions, so there is no
dependence between the coefficients.

65
 x y
30 120
40 420
60 320
65 520
70 580
93 670
Table 5.1: The base points for the example spline background.

800
spline base points
interpolated spline function

700

600

500
Counts

400

300

200

100

0
30 40 50 60 70 80 90
2Theta

Figure 5.6: Example for a cubic spline, which can be used as background. As it can be seen, the
spline goes through the points very smoothly.

66
5.1.6 Sampling of the Fourier transforms
In the CMWP procedure the sampling of the Fourier transforms depends on whether instrumental
broadening is present or not.

• If there is no instrumental effect, the Fourier transforms are evaluated with an arbitrary precision
specified by the user of the program. The sampling of the theoretical Fourier transform is
determined by Alim , an amplitude ratio limit: for each reflection, the normalized theoretical
Fourier transform is calculated with equidistant sampling from L = 0 to Lhkl hkl
max , where Lmax is
A(Lhkl
max ) = A .
defined by the following equation: lim
A(0)
• If instrumental profiles are available, the sampling of the theoretical Fourier transforms depends
on the Fourier transform of the measured instrumental profiles. Since the measured instrumen-
tal profile is represented by a set of numerical data which terminates at a finite 2θ value, its
Fourier transform is periodic. If the sampling of the instrumental intensity profiles is equidis-
tant and the intensity values are given in points with a distance of ∆K instr. , the period of the
1
instrumental Fourier transforms is . This period determines the maximum L value for
∆Kinstr.
calculating the theoretical Fourier transforms: it has no sense to calculate the theoretical Fourier
transform above this value and include the other maxima of the instrumental Fourier transform
in the convolution. Therefore, the Fourier transforms are calculated with equidistant sampling
from L = 0 to Lhkl hkl hkl
max , where Lmax is the L value where Ainstr. , the Fourier transform of the mea-
sured instrumental profile has the first minimum value. This value is determined as follows:
the (complex) instrumental Fourier transform, Ainstr. (L) is calculated from L = 0 by increasing
RL ′ ′
L with small steps and in each point the integral Iint (L) = |Ainstr. (L )|d L is calculated and
′
L =0
MAX , the
the calculation is stopped if Iint (L) exceeds Iinstr. (0) =
Iinstr.maximum intensity value of
the instrumental profile. The so obtained L value is corresponding approximately to the sec-
ond maximum of Ainstr. (L), so the correct value for Lhkl
max is half of this L value (where the
instrumental Fourier transform has its first minimum value).
Fig. 5.7 shows the Fourier transform of a measured instrumental profile, the periodic function
is plotted until its second maximum value. The period of this function is 1526 nm, which is in
good agreement with the sampling of the measured instrumental profile: the instrumental peak
1
is measured in equidistantly sampled points: ∆K = 0.00065 , so the calculated period of the
nm
1
Fourier transform is: LNY QUIST = = 1538nm.
∆K
Depending on the value of Lhkl
max , the number of Fourier samples is:

max ∆K ,
NFT = 2Lhkl P
(5.4)

where:
2 2θB + P 2θB
    
∆K =
P
sin − sin , (5.5)
λ 2 2

67
and the profile cutting parameter is denoted by P and the 2θ value of the peak center is denoted by
2θB . So by using this sampling the Fourier transforms are calculated equidistantly at points with a
Lhkl 1 1
distance of ∆L = max = and the period of the inverse Fourier transform is = 2∆K P . This
NFT 2∆K P ∆L
2 sin θB
means that the peak intensity profile has minimums at the points KB ± ∆K P (where KB = ) and
λ
it is cut at these points. If P is small, this means that it is cut approximately at the points 2θB ± P.

0.008
Instr-FT.dat

0.007

0.006

0.005

0.004
|A(L)|

0.003

0.002

0.001

-0.001
0 200 400 600 800 1000 1200 1400
L [nm]

Figure 5.7: The absolute value of the Fourier transform of an instrumental profile as a function of L.
It is plotted from L = 0 until its second maximum value.

5.1.7 Sampling of the intensity patterns

In the least squares algorithm, the value of the theoretical intensity pattern is required at the 2θ values
of the measured pattern data. However, since the measured pattern can contain many data points,
by calculating the theoretical pattern only at relevant points and by using simple linear interpolation
between these points, considerable speed-up can be achieved in the evaluation procedure. These rel-
evant points are determined by two parameters: N1 is a parameter which represents an equidistant
sampling in the measured 2θ range and the theoretical pattern is calculated in additional positions
whose number is N2 . These points are selected around the measured peaks using the following algo-
rithm:

1. the measured 2θ range is divided into N1 equidistant sub-intervals.

2. in each sub-interval additional points are added equidistantly where the theoretical pattern is
also calculated. The integral of the background stripped measured pattern is calculated for

68
 the n-th sub-interval. Denoting this integral by In and the whole area under the background
stripped measured pattern by I pat , the number of the additional sampling points added in the
I
n-th sub-interval are: N2 n .
I pat
So the total number of sampling points is: N1 + N2 .

5.1.8 The profile cutting parameter

The theoretical intensity profiles are calculated only in the points corresponding to the sampling
determined by N1 and N2 (see the previous section for the details). Since the inverse Fourier transform
is also periodic, the profiles cannot be calculated in the entire 2θ range, and the usage of a cutting
parameter is necessary. It is noted that the inverse Fourier transform is periodical in K only and not
in 2θ, this is the reason why the cutting is performed in K and not in 2θ. The profiles are evaluated
only in the proximity of ∆K P of the corresponding Bragg positions, where P is the “profile cutting
parameter” and ∆K P is defined by eq. (5.5). As we have seen in section 5.1.6, the period of the inverse
Fourier transform is two times ∆K P , so it has a minimum value at the distance of ∆K P , where it is
cut. The theoretical intensity pattern is put together from the background and the theoretical intensity
profiles sampled in the manner described above.

5.1.9 Instrumental profiles

hkl , the measured instrumental profiles corresponding to the hkl indices are
In the CMWP procedure Iinstr.
used directly. The ideal sample for measuring the instrumental profiles has the following properties:

1. it should possibly have a large number of well separated Bragg reflections in the whole investi-
gated 2θ angular range.

2. it should be strain free with a large enough grain size in order to avoid size broadening, this
way it shows only the instrumental properties.

These conditions are more-or-less well satisfied in the case of LaB6 specimens provided by the Na-
tional Institute of Standards and Testing (NIST). The CMWP evaluation program selects automati-
cally the nearest instrumental profile to each of the peak center of the evaluated pattern. Since the
overlapping between the peaks of the instrumental patterns is negligible, it is easy to separate the
spectrum (e.g. with the program MKDAT, see section 4.1.2 for more details). Fig. 5.8 demonstrates the
angular dependence of the instrumental broadening measured on the high resolution powder diffrac-
tion beam line, BM16 of ESRF in Grenoble, France (Martinetto, 2000). Fig. 5.9 shows a part of the
instrumental pattern of a Philips X’Pert powder diffractometer measured on LaB6 sample.

69
 0.01
Si
NaCl
0.009

0.008

0.007

0.006
∆K [1/nm]

0.005

0.004

0.003

0.002

0.001

0
0 5 10 15 20 25 30 35
K [1/nm]

Figure 5.8: The FWHM values of the instrumental profiles measured on the high resolution powder
diffraction beam line, BM16 of ESRF in Grenoble, France, as a function of K for Si and NaCl samples.

20000
LaB6

18000

16000

14000

12000
Intensity

10000

8000

6000

4000

2000

0
25 30 35 40
2Theta

Figure 5.9: Part of the instrumental pattern of a Philips X’Pert powder diffractometer measured on a
LaB6 sample.

70
5.1.10 Weighting algorithms
Without proper weighting the least squares algorithms are strongly influenced by the points with
large y values (these points are overweighted). If the difference between the y values is significant,
the usage of weighting is justified. In the MWP method the profiles are normalized, so there is no
need for weighting. In the CMWP method the intensity of the measured peaks can be different by
orders of magnitude.
The intensity maxima decrease with 2θ since the atomic scattering factors decrease with 2θ and, on
top of that, the structure factor can also cause significant differences in the peak maxima. A further
effect can be caused by texture. The CMWP program offers to use appropriate weighting if needed
in order to take into account the peaks with smaller intensity maxima. In the least squares method
the weights are usually 1/y2 . A special weighting is used for the CMWP procedure in which the
profiles are weighted uniformly (provided that the overlapping between the peaks is weak), however,
the smaller peaks have larger weights. The weights are generated from the maximum intensity values
according to the equation:

wi (2θi ) = max{I MAX

j , where|2θi − 2θ j | < P}. (5.6)

In this equation the I MAX

j intensities are the background-stripped peak intensities, 2θ j are the peak
centers and P is the profile cutting parameter. The actual weights used in the least squares procedure
1
are the 2 values.
wi
Fig. 5.10 shows the intensity pattern and the corresponding wi values for a CdF2 sample ball milled
for 30 minutes as a function function of 2θ. By using these weights, the results provided by the
CMWP procedure are in better correlation with the results of the MWP procedure, where the peaks
are also uniformly weighted. It is noted that due to the different weighting algorithms, a difference
is expected between the results of the MWP method on the intensity profiles, the results of the MWP
method on the Fourier transforms and the results of the CMWP method. The usage of the weights
1/y2 is also possible in CMWP, however this is not optimal if the level of the background is elevated.

5.1.11 Interpretation of the errors of the fitting parameters

In the least squares procedures the errors of the fitting parameters are usually given by their confi-
dence levels. In the fitting algorithm of gnuplot (which is used by the MWP and CMWP procedures
also) the error estimates of the parameters are given rather than confidence intervals. These error es-
timates are calculated from the variance-covariance matrix after the final step of the iterations. These
estimates are the so called “asymptotic standard errors” (which are given by the program gnuplot
at the end of the fitting). They are calculated as the standard deviation of each parameter. It should
be noted, that these asymptotic standard errors are usually over-optimistic and therefore should not
be used instead of the confidence levels. However, these asymptotic standard errors can be used as a
qualitative measure of the fit.

71
The program gnuplot also gives a correlation matrix, which indicates the correlation of the parame-
ters in the region of the solution. The meaning of the elements of the correlation matrix is: if the value
of a parameter is changed, in this way increasing the WSSR, does the changing of the other parameter
compensate? The elements in the main diagonal elements are always 1 (they correspond to autocor-
relation) and if all parameters were independent, all other elements would be nearly 0. Two different
parameters which completely compensate each other would have an element of unit magnitude, their
sign is depending on whether the relation is proportional or inversely proportional. If the magnitudes
of the non-diagonal elements are small, it means that the estimates of the standard deviation of each
parameter are close to the asymptotic standard error given by gnuplot.

5.1.12 Convergence and numerical stability

By using the Marquardt-Levenberg nonlinear least squares algorithm MWP and CMWP tries to find
the minimum value of WSSR in a multi dimensional parameter space (the dimension is denoted by n
and it is the number of the fitting parameters). In the nonlinear Marquardt-Levenberg algorithm the
gradient vector (whose components are the partial derivatives of WSSR as a function of the fitting
parameters) is calculated and the parameter values are modified according to the direction of the
gradient vector and a step parameter called λ. In order not to jump over the minimum - which would
be the solution - the step is modified adaptively, i.e. after each step of the iteration, λ is either
multiplied or divided by a numerical factor and, if the WSSR is better in the new point, the new value
for λ is accepted. Depending on the properties of this n-dimensional surface, finding the physical
minimum is not a trivial problem since local (non-physical) minimums can also be obtained. If this
surface has a sharp minimum there is usually no problem. However if it is flat, there is a chance to find
an inappropriate minimum. By changing the initial values of the parameters and examining the value
of WSSR and the residuals corresponding to the final solution one can try if the solution corresponds
to the absolute minimum or not.
In the following an example will be presented for the examination of the stability of the solution by
running systematically selected fittings starting from different initial values of the parameters. Since
practice has shown that in some cases it is difficult to get a solution for the dislocation parameters,
the stability of the parameters d and e will be examined (d and e are related to the dislocation density
and effective outer cut-off radius of dislocations, respectively).
Table 5.2 shows the examination of the solution of the CMWP fit for SrF2 sample ball milled for
30 minutes: the solution parameters are indicated for different initial values of the parameters d and
e (the initial value of the other parameters was always the same: a = −1.3, b = 1.7 and c = 1.3).
It can be seen that the solution for this sample is well defined and unique: it finds approximately
the same minimum starting from a wide range of the initial parameter values. However, at the same
time, by using inappropriate values for the parameters used for the sampling of the theoretical Fourier
transforms and/or the intensity pattern (e.g. for the profile cutting parameter) it is possible to get an
unstable behaviour for the solution even in the case of the same sample.
Fig. 5.11 shows the WSSR corresponding to the solution of the CMWP fit for the same SrF2 sample

72
 5000
pattern
wi values
4500

4000

3500

3000
Counts

2500

2000

1500

1000

500

0
40 60 80 100 120 140
2θ [o]

Figure 5.10: The measured pattern (solid lines) and the corresponding wi values used in the CMWP
algorithm (dashed line) for CdF2 sample ball milled for 30 minutes, as a function of 2θ. The actual
1
weights are the 2 values.
wi

dini eini asol bsol csol dsol esol WSSR

4 0.5 -0.8062 (13%) 2.054 (1.6%) 1.204 (0.72%) 19.5 (21%) 0.87 (43%) 2.6698
6 0.4 -0.7934 (14%) 2.069 (1.4%) 1.200 (0.63%) 15.8 (17%) -1.33 (34%) 2.6709
8 0.3 -0.8062 (13%) 2.054 (1.6%) 1.204 (0.72%) 19.5 (21%) -0.87 (43%) 2.6698
10 0.25 -0.8062 (13%) 2.054 (1.6%) 1.204 (0.72%) 19.5 (21%) -0.87 (43%) 2.6698
20 0.2 -0.8156 (13%) 2.044 (1.7%) 1.207 (0.74%) 21.5 (18%) 0.71 (37%) 2.6696
30 0.15 -0.8331 (13%) 2.028 (1.4%) 1.211 (0.62%) 24.4 (5.7%) 0.55 (12%) 2.6702
40 0.1 -0.8597 (13%) 2.005 (1.4%) 1.218 (0.61%) 28.1 (4.4%) 0.41 (8.9%) 2.6724
60 0.08 -0.8323 (13%) 2.028 (1.4%) 1.211 (0.61%) 24.4 (5.6%) 0.55 (11%) 2.6702
80 0.05 -0.8331 (13%) 2.028 (1.4%) 1.211 (0.62%) 24.4 (5.7%) 0.55 (12%) 2.6702
100 0.03 -0.8605 (13%) 2.005 (1.4%) 1.218 (0.61%) 28.1 (4.5%) 0.41 (9.1%) 2.6724

Table 5.2: The initial values (dini and eini ), the solution parameters (asol , · · · , esol ) and the WSSR
values obtained as the solution of the CMWP fit for SrF2 sample ball milled for 30 minutes. The
initial value of the other parameters was always the same: a = −1.3, b = 1.7 and c = 1.3. It is noted
that the WSSR values have to be multiplied by a factor of 107 . The asymptotic errors of the fitting
parameters are also indicated in parentheses.

73
as a function of the value of the d parameter (this value was fixed during the fitting and only the
values of the other parameters were refined). It can be seen that it has a flat minimum value around
d = 20. However if the same type of curve for a sample is monotonously increasing (so its derivative
is always positive and does not reach zero at a particular dsol value), when refining all parameters
instead of a well defined solution, the parameter d approaches asymptotically zero (and at the same
time e approaches asymptotically infinity). By fixing the value of de (which means that the value
of M, the dislocation arrangement parameter is fixed) a solution can be obtained for ρ, however it
is better to solve the problem without this (ad-hoc) restraint. These problems have several possible
sources:

• the incorrect choice of the theoretical function is a possible cause

• the incorrect choice of the sampling parameters of the theoretical Fourier transform or theoret-
ical intensity pattern should also be avoided

• the noise and low quality measurements have negative effects on the evaluation

In section 5.5 a solution for a similar stability problem will also be presented: by using a more
appropriate physical model for the size broadening of these samples the unstable behaviour of the
dislocation parameters can be eliminated.

3.1e+07
d-wssr.dat

3.05e+07

3e+07

2.95e+07

2.9e+07
WSSR

2.85e+07

2.8e+07

2.75e+07

2.7e+07

2.65e+07
0 20 40 60 80 100 120
dfix

Figure 5.11: The WSSR corresponding to the solution of the CMWP fit for SrF2 sample ball milled
for 30 minutes, as a function of the value of the d parameter (which was fixed during the fitting and
only the value of the other parameters were refined).

74
5.2 Modeling asymmetric diffraction profiles
In this section a method suitable for modeling asymmetric line profiles is presented. It is based on
Mughrabi’s composite model (Ungár et al., 1984; Mughrabi et al., 1986) and an asymmetric strain
profile is composed by the sum of two shifted strain profiles calculated by using the Wilkens model
(Wilkens, 1970). The ρ parameters of these functions correspond to the dislocation density in the
inner region of the cell and in the cell’s wall, respectively. In this procedure only a single peak is
analysed and the size broadening is neglected. The evaluation procedure consists of the following
steps:

1. Correction of the peak’s center and normalization: The center of gravity of the measured
intensity profile Im (s) is defined as:

R∞
s Im (s) ds
−∞
s= R∞
. (5.7)
Im (s) ds
−∞

Therefore the area normalized and center corrected intensity profile is given by the transform:

Im (s − s)
Ic (s) = R∞ . (5.8)
Im (s) ds
−∞

2. Evaluation using the symmetrized theoretical intensity profile: The size broadening of the
profiles is neglected, so the theoretical Fourier transform becomes:

πb2 2
  
L
A(L, ρ, R∗e ) = exp − 2
(g C)ρL f , (5.9)
2 R∗e

Since A(L) is real and even function of L, its inverse Fourier transform, the theoretical intensity
profile is real and even function of s:

Z∞
I(s, ρ, R∗e ) = 2 A(L, ρ, R∗e ) cos(2πLs) dL. (5.10)
0
R∞
Note that the theoretical intensity function I(s) is area normalized, since: I(s) ds = A(0) = 1.
−∞
In this step Ic (s), the area normalized and center corrected intensity profile is fitted by I(s −
s01 , ρ, R∗e ), so the value of the parameters s01 , ρ and R∗e is refined.

3. Evaluation using the asymmetrical theoretical intensity profile:

75
In this step, the theoretical intensity function Ias (s) is defined as:

Ias (s) = f1 I(s − s1, ρ1 , R∗e,1 ) + f2 I(s − s2, ρ2 , R∗e,2 ), (5.11)

where:
f1 + f2 = 1

s1 f 1 + s2 f 2 = 0
(5.12)
f 1 ρ1 + f 2 ρ2 = ρ

 f 1 ρ1  f 2 ρ2
ρ ρ
 
R∗e,1 · R∗e,2 = R∗e

The parameters ρ and R∗e are known from the previous step, so the parameters f2 , s2 , ρ2 and R∗e,2
are determined unequivocally by the parameters f1 , s1 , ρ1 and R∗e,1 . In this step Ic (s) is fitted by
Ias (s), so the value of the parameters f1 , s1 , ρ1 , R∗e,1 is refined and the value of the parameters
f2 , s2 , ρ2 , R∗e,2 is calculated using equations (5.12).
Although the method presented here was developed for single profile analysis, it was presented
here, because the CMWP program package [S14] includes an evaluation program developed
for this procedure.

76
5.3 CMWP application to Al-Mg alloys [S14, S16]
In [S14, S16] the CMWP fitting procedure is used for the evaluation of the X-ray diffraction patterns
of Al-Mg alloys prepared by mechanical alloying. The microstructural parameters determined by the
CMWP method were compared to the results obtained by the MWP fitting procedure. The aluminium-
magnesium samples were prepared and measured as described in section 4.3 and in [S13]. In [S14]
three samples were investigated, each after 3h ball-milling with nominal Mg concentrations 0, 3 and 6
at. %. The lattice constants of the ball milled samples were determined by the standard Nelson-Riley
method (1945). The Mg concentration in the solid solution was determined by using the slope in
Vegard’s law (Pool & Axon, 1952). The true Mg concentrations were 0, 2.26 and 5.52 at. % for the
samples with nominal Mg concentrations 0, 3 and 6 at. %, respectively.
Figs. 5.12 and 5.13 show the X-ray diffraction pattern obtained on the Al-3 wt.% Mg powder after
3 h ball milling in linear and logarithmic intensity scales, respectively. The open circles represent
the experimental data. As it can be seen from the figure the main phase is the Al-Mg solid solution,
however, traces of Mg around the first reflection of Al-Mg can be observed in the logarithmic plot.
The solid line in Figs. 5.12 and 5.13 represent the model based, theoretical pattern fitted by the
CMWP procedure. The agreement between the the measured and the fitted patterns is satisfactory.
The severe plastic deformation during milling resulted in the formation of ultrafine-grained structure
(< x >vol =39 nm) with very high dislocation density (ρ = 44 · 1014m−2 ).
The microstructural parameters were determined for each sample by applying both the CMWP and
MWP fitting methods. It was found that the microstructural parameters obtained by the MWP fitting
method are in reasonable agreement with those determined by the CMWP fitting procedure, as it is
shown in Table 5.3. The observed differences are due to the different weightings used in the MWP
and CMWP procedures.

Table 5.3: Microstructural parameters obtained by the MWP and CMWP procedures for ball milled
Al-xMg (x=0, 3, 6 wt.%) specimens

x [wt.%] Method m [nm] σ ρ [1014 m−2 ] M q

0 MWP 60 (6) 0.18 (2) 19 (2) 1.0 (2) 0.7 (1)
3 MWP 33 (3) 0.21 (3) 40 (4) 1.2 (2) 0.5 (1)
6 MWP 19 (2) 0.40 (3) 72 (6) 1.0 (2) 0.4 (1)
0 CMWP 73 (7) 0.10 (2) 12 (2) 1.1 (2) 1.3 (1)
3 CMWP 29 (3) 0.29 (3) 44 (4) 1.2 (2) 1.1 (1)
6 CMWP 26 (3) 0.13 (3) 100 (15) 1.3 (2) 0.7 (1)

77
 60000
measured data
theoretical curve
difference

50000

40000
111
Intensity

30000

20000

200

10000
220 311

222 331 420 422

400
0
40 60 80 100 120 140
2θ [ o ]

Figure 5.12: The measured and fitted patterns (fitted by the CMWP method) of the Al-3 wt.% Mg
alloy. In this figure the intensity values are plotted as a function of 2θ. The indices of reflections
are also indicated. The differences between the measured and fitted intensity values are plotted at the
bottom of the figure.

100000
measured data
theoretical curve

111

200

10000
220 311
Intensity

222 331
420
422

400
1000

100
40 60 80 100 120 140
2θ [ o ]

Figure 5.13: The measured and fitted patterns (fitted by the CMWP method) of the Al-3 wt.% Mg
alloy (logarithmic intensity scale). In this figure the logarithm of the intensity values are plotted as a
function of 2θ. The indices of reflections are also indicated.

78
5.4 CMWP application to nanocrystalline Cu containing planar
faults [S18]
The effect of stacking faults and twins on X-ray diffraction patterns has been calculated numerically
by using the DIFFaX software (Treacy et al., 1991) for the first 15 Bragg reflections in fcc crystals
up to 20 % fault densities by Mr. Levente Balogh [S18]. It was found that the Bragg reflections
consist of up to 5 sub-reflections which are broadened and shifted to different extent according to
particular hkl conditions, for more details see section 2.4 and [S18]. Balogh has also shown that
the sub-reflections can be well described by Lorentzian profile functions over an intensity range of
about 4 to 5 orders of magnitude. It was further shown that this feature is in good correlation with
the fact that stacking faults or twin boundaries act as boundaries of coherently scattering domains
and that they cause homogeneous strain. In that sense faulting or twinning causes a kind of size
broadening associated with homogeneous strain. About 15000 sub-reflections were evaluated and
parametrized according to their FWHM and positions relative to the exact Bragg angles. The stacking
faults profile function defined by eq. (2.78) and the calculated parameter files were incorporated
into the CMWP software for evaluating planar faults together with dislocations and crystallite or
subgrain size distributions [S18]. Bulk nanocrystalline copper specimens were prepared by inert
gas condensation and hot compaction at Argonne National Laboratory (Ungár et al., 1998). They
are denoted as O2 − IS, P2 − IS, and N2 − IS in the as-prepared state and P2 − T and N2 − C after
tensile and compression tests, respectively. By applying the CMWP procedure with the inclusion
of the stacking or twin faults effect the microstructure of these nanocrystalline copper samples were
evaluated. The twin density obtained by the CMWP procedure is increasing up to a few percents as
the grain size decreases. Fig. 5.14 shows the twin density as a function of the area averaged mean
crystallite or subgrain size for the different specimens. It can be seen that when the subgrain size is
larger than about 40 nm the twin density is close to zero, within experimental error. At subgrain-size
values smaller than about 40 nm the twin density increases sharply and reaches values up to 6%.
This means that in nanocrystalline Cu twinning becomes a substantial mode of deformation when the
subgrain size becomes smaller than about 40 nm.

5.5 CMWP application to ball milled fluorides [S17]

Nanocrystalline fluorides are important because of their usage as sensor materials. Homogeneous
nanocrystalline materials can be produced by ball milling and at the same time, the milling procedure
also introduces several types of strains. In [S17] 4 different types of fluorides produced by ball milling
for different milling periods were investigated by X-ray diffraction. About 20 diffraction patterns of
CaF2 , SrF2 , BaF2 and CdF2 fluorides produced by ball milling were measured by a D500 Siemens
high resolution powder diffractometer with incident monochromator. By applying the CMWP proce-
dure, the microstructural parameters for size and strain broadening were determined. In about half of
the cases, the measured and fitted diffraction patterns were in perfect agreement throughout the entire

79
Figure 5.14: The twin density β as a function of the area averaged mean crystallite or subgrain size
hxiarea for the different specimens. Inert gas condensed and compacted specimens: O2 -IS, P2 -IS, P2 -
T, N2 -IS, and N2 -C. Submicron grain-size samples deformed by ECAP: ECAP(a) open left triangles
and ECAP(b) open squares, respectively. The dashed line is to guide the eye. The solid vertical line
indicates the experimental uncertainty.

80
2θ range. However, in the rest of the cases the first few reflections of the measured patterns were
significantly narrower than the fitted ones, calculated on theoretical basis. At the same time, also in
these latter cases, the rest of the reflections were in good agreement for the higher angular part of the
patterns consisting of about 15 profiles. It is worth to note, that in the cases when the above mentioned
differences between the measured and calculated patterns were observed, it was almost impossible to
obtain a physically realistic solution for the microstructural parameters, i.e. it was not possible to
determine the parameters of the strain effect: the solution for ρ, the dislocation density and Re , the ef-
fective outer cut-off radius of dislocations was asymptotic as described in section 5.1.12. In the cases
when the first few measured profiles were narrower than the calculated ones, this discrepancy were
interpreted by an X-ray optical interference effect [S17] similar to what was observed by Rafaja et al.
(2004) in nanocrystalline thin layers. The interference effect has been corrected successfully, either
by (i) excluding the affected profiles from the evaluation procedure or by (ii) assuming a diffraction
angle dependent apparent bimodal size distribution of crystallites [S17].

5.5.1 The interference effect

Rafaja et al. (2004) have observed recently an X-ray optical interference effect in nanocrystalline
thin film of Ti0.38Al0.62N. As it can be clearly seen in the Williamson-Hall plot in Figure 6 of
Rafaja et al. (2004), the broadening of the first few reflections (corresponding to the smallest g
values) is significantly smaller than the broadening of the other reflections. This means that the
apparent domain size corresponding to these peaks is increased due to an interference phenomenon,
consequently these peaks are narrowed. The other peaks corresponding to larger g values are not
affected by this interference effect, so their broadening corresponds to the real physical coherently
scattering domain size. As pointed out by the authors, the effect was present in the cases when the
following conditions were simultaneously present: (i) the coherently scattering domain size was less
than about 5 nm, (ii) a texture was present in the sample and (iii) the absolute value of the diffraction
vector was small. In the following a simple interpretation of the interference effect is presented. Fig.
5.15 shows the reciprocal lattice spots of two adjacent crystallites rotated to each other by a relatively
small angle. The difference in their orientation also means that their reciprocal lattices are also rotated
to each other by this small angle. Due to the small size of crystallites, the reciprocal lattice points are
broadened (in the case of isotropic crystallite shape the extent of broadening is the same in each point).
If the crystallite size is small, the broadening becomes large and if the difference in orientation is also
small, the reciprocal lattice spots corresponding to small g vectors can overlap. Due to the increase of
the coherently scattering region in the reciprocal lattice, the peaks corresponding to these points are
narrowed. In Fig. 5.15 the encircled reciprocal lattice-spot pairs #1 and #2 are overlapping, so their
peak profiles are narrowed and the others corresponding to larger g values (like spot pair #3) do not
overlap anymore, so their peak profiles are not influenced by this interference effect. Consequently the
coherently scattering domain size determined from these latter peaks corresponds to the appropriate
crystallite size. This is the simple and direct reason why this effect influences only the first few
reflections in a diffraction pattern, as can be seen in a direct way in Figure 6 of Rafaja et al. (2004).

81
Figure 5.15: Schematic drawing for the interpretation of the interference effect. The dashed and the
dotted spots are the blown up reciprocal lattice nodes corresponding to two adjacent crystallites ro-
tated in orientation with respect to each other by a small angle. The orientations of the two crystallites
are indicated schematically by the dashed and dash-dotted arrows. The pairs of encircled diffraction
spots numbered #1 and #2 overlap, whereas those numbered #3 are well separated.

82
Figure 5.16: Observed and fitted patterns of SrF2 after 12 min of ball milling.

Figure 5.17: Observed and fitted patterns of SrF2 after 40 min of ball milling.

83
 Figure 5.18: Observed and fitted patterns of SrF2 after 60 min of ball milling.

Figure 5.19: Observed and fitted profiles of the 111 reflection of SrF2 after 40 min of ball milling.

84
Figure 5.20: Evolution of the interference effect with milling periods for the 111 diffraction line of
SrF2 samples.

The measured and fitted patterns of SrF2 samples corresponding to 12, 40 and 60 minutes of ball
milling periods are plotted in Figs. 5.16, 5.17 and 5.18, respectively. As it can be seen in Fig. 5.16,
the measured and the model based, theoretical patterns of SrF2 ball milled for 12 minutes are in
perfect agreement throughout the entire 2θ range of the measurement. On the other hand, according
to Fig. 5.17, there is significant difference between the measured and fitted curves of SrF2 ball milled
for 40 minutes at lower 2θ angles, the first few measured profiles are significantly narrower than the
fitted ones. Nevertheless, also in this case the agreement between the measured and fitted profiles
for 2θ angles larger than about 60◦ is satisfactory. After even longer milling periods, e.g. after 60
min ball milling, as can be seen in Fig. 5.18, a similar discrepancy can be observed between the
measured and fitted patterns as in the case of the 40 min ball milled sample, however, the differences
are definitely smaller, meaning that the effect is less pronounced in this case. As can be seen in Fig.
5.19 showing the measured and fitted profiles of the 111 reflection of SrF2 ball milled for 40 minutes,
far enough from the peak center the measured and fitted profiles coincide, so the problem cannot be
related to the incorrect determination of the background level. This suggests that either the shape of
the peak or the extent of broadening is incorrect. The evolution of the interference effect with milling
periods is shown in Fig. 5.20. It can be seen that the differences are increasing up to 40 min and start
decreasing at longer milling periods. The microstructural parameters obtained by the CMWP fitting
procedure also show a discrepancy for the samples for which differences were observed. Instead of a
well defined solution for the parameters of the microstructure, the solution for the dislocation density

85
and the effective outer cut-off radius of dislocations approaches asymptotically infinity and zero,
respectively, so without applying some artificial restraints (e.g. by fixing the value of M) one can get
only a physically unrealistic solution. By omitting the first few reflecitions, the fitting procedure works
well for the rest of the profiles and gives a reasonable solution. The observed differences between the
measured and the model based, fitted patterns suggest the presence of a similar interference effect as
observed by Rafaja et al. (2004) in Ti0.38Al0.62N thin films. However, in the case of these fine loose
powder samples, there can’t be a strong orientational correlation (real texture) between the particles,
but a local orientational correlation (local texture) is possible between adjacent crystallites. If the
ratio of these locally textured particles is large enough, it is producing a similar interference effect
as the one observed by Rafaja et al. (2004). From this we can conclude that two main conditions
make the existence of this kind of interference plausible. (i) A fraction of the fluoride powder consists
of nanocrystalline particles. This condition is produced by ball milling. (ii) In this fraction of the
powder the orientation of adjacent particles is crystallographically not too far from each other, in
other words, there is a local texture in the powder. Since the lattice parameters of these fluorides are
relatively large, the first few diffraction spots are relatively close to the origin of reciprocal space,
meaning that this orientational correlation condition is not too strict. The orientational correlation
between adjacent crystallites can be caused by the Coulomb interaction present in these ionic crystals.
This can produce some kind of epitaxial interaction between adjacent crystallites resulting in a local
orientation correlation between them. In ball milled cobalt, for example, epitaxially touching particles
were observed by HRTEM (Yao & Thölén, 2000). For longer milling periods the milling procedure
enables coagulation and/or coalescence of smaller particles thus increasing the average crystallite size
and diminishing the magnitude of the interference effect.

5.5.2 Correcting for the interference effect

The angular dependence of the apparent size distribution in the case of the SrF2 sample ball milled
for 40 min (where the interference effect is the most pronouncedly present) was examined as follows:
with fixed strain parameter values, the peaks were fitted individually and the lognormal size distribu-
tion parameters were determined for each profile. The m and σ values obtained this way are shown in
Fig. 5.21 as a function of g. It can be seen that m decreases strongly whereas σ is slightly increasing
in good correlation with the model in section 5.5.1.
The interference effect was corrected for by the following two fundamentally different approaches.
(i) The first two reflections were omitted from the fitting, thus the remaining peaks, which were
practically not affected by the interference effect provided the correct microstructural parameters. (ii)
In accordance with the model of the effect of interference on the diffraction patterns, it is assumed
that a fraction of the powder has an apparent crystallite size larger than the true value. Furthermore
it is assumed that the fraction of the crystallites affected by the interference effect and the rest of
the powder, both can be described by separate size distributions. Latter are denoted by f1 (x) and
f2 (x), respectively. A decaying mixing parameter ξ is introduced in order to describe the angular
dependence of the interference effect. The effective size distribution function, fe f f is obtained as the

86
Figure 5.21: The m and σ size parameters, determined individually for the first few profiles in the
sin θ
diffraction pattern of SrF2 ball milled for 40 min, as a function of g (g = 2 ).
λ

87
weighted sum of f1 (x) and f2 (x):

fe f f (x) = ξ f1 (x) + (1 − ξ) f2 (x) (5.13)

The simplest and most plausible selection for the size distribution functions is the lognormal function,
so f1 (x) and f2 (x) are described by lognormal size distribution functions with the median and variance
parameters: m1 , σ1 and m2 , σ2 respectively. Eq. (5.13) shows that the effective size distribution is
a g dependent bimodal size distribution function in which the apparent part disappears with g. The
numerical analysis was shown that the g dependence of the mixing parameter ξ was well described by
g − g0 1 g − g0
a simple Lorentzian function of : ξ= 2 , where s = , g0 = 2 sin(θ111 )/λ and 2θ111
g 1 + As g
is the Bragg angle of the 111 reflection. By using this correction considerable improvement of the
quality of the fitting was achieved. A typical example of the fitting with this correction, corresponding
to SrF2 ball milled for 40 min is shown in Fig. 5.22 which has to be compared to the uncorrected fit
plotted in Fig. 5.17.
It is also noted that by using this correction the fitting was stable and gave reasonable values for the
microstructural parameters without applying any restraints or omitting any part of the patterns. In the
following the microstructural parameters obtained by using this correction are analyzed.

Figure 5.22: Fitting with correction for the interference effect for SrF2 ball milled for 40 min.

88
5.5.3 The microstructure of the fluorides
Figs. 5.23 and 5.24 show typical lognormal distribution functions for the ball milled SrF2 specimens
according to the parameters m and σ given by the CMWP procedure. The figures show that after
shorter ball milling periods, i.e. about 6 and 12 min, the specimens consist of almost monodisperse
and relatively large crystallites. At longer ball milling periods the size distributions indicate that the
size of crystallites collapses dramatically. At the longest ball milling periods a slight re-increase of the
crystallite size can be observed, see Fig. 5.24. This small but definite reincrease has been observed
in three fluorides: SrF2 , BaF2 and CaF2 . It means that during the ball milling of these fluorides the
particles are first fragmented and subsequently become welded again. Similar behaviour during ball
milling has been observed frequently in the literature. Guerrero-Paz & Jaramillo-Vigueras (1999)
ball milled ductile materials such as Cu, Ni and Cu-Ni alloys and observed coalescence of grains
after longer milling periods by TEM investigations. Rác and coworkers (2005) ball milled dodeca-
tungstophosphoric acid (HPW) and Nafion-H for catalytic purposes. Authors claim that because
of coalescence catalysts with high catalytic performance require the use of optimal milling time.
Coalescence during ball milling was also observed by Pál et al. (1986). Janot & Guerard (2005)
also observed coalescence during ball milling of anodic materials for lithium-ion batteries. They state
the importance of surface energy in the procedure of coalescence. The importance of coalescence
is well known in the process of sintering, where surface energy is one of the controlling parameters
(Wakai & Aldinger, 2003). The area weighted mean crystallite size values, < x >area , as a function of
the milling period are shown in Fig. 5.25 for the different fluorides. The figure shows that < x >area
decreases first with the milling period and after a minimum value it re-increases slightly in accordance
with the size distribution functions. The absence of the slight re-increase of the crystallite size in CdF2
might be due to the rather long incubation period observed in the evolution of the dislocation density
discussed in detail below. The evolution of the dislocation density, ρ, with milling period is shown
in Fig. 5.26. In the initial state of the as-grown crystals the dislocation density is practically zero,
i.e. ρ < 1014 m−2 , this value is marked schematically in the lower left corner of the figure. The
figure indicates that the investigated fluorides fall into two different groups. (i) In the case of BaF2
and SrF2 the dislocation densities increase almost monotonously up to about 7 · 1015 m−2 during the
first 30 min milling period and saturate for the rest of the milling procedure. (ii) In the other group,
CaF2 and CdF2 , during the first 6 to 12 min ball milling ρ increases to about 2.5 · 1015 m−2 where it
saturates for about 20 and 30 min in the case of CaF2 and CdF2 , respectively. After this intermediate
plateau the dislocation density increases again during further ball milling and reaches the same large
values of ρ as in the case of the other two fluorides. The figure suggests that after even longer milling
periods also in these two fluorides, i.e. in CaF2 and CdF2 , the dislocation densities would saturate
at around values of 7 · 1015 m−2 . The presence of an incubation period in the milling procedure of
CaF2 and CdF2 is unequivocally present. A much weaker, however, existing short incubation-like
stage in the case of BaF2 and SrF2 can also be observed, as shown in Fig. 5.26. The latter lasts
up to about 12 minutes. The incubation periods in the evolution of the dislocation densities can be
T
correlated to the homologous temperatures of the fluorides, Thom = RT (where TRT and Tm are room
Tm

89
temperature, TRT = 300K, and melting temperature, respectively). In Fig. 5.27 the incubation periods
are plotted versus the homologous temperatures. A good correlation between the two parameters can
be seen. The rapid increase of ρ, without longer incubation periods can be observed in the case of
BaF2 and SrF2 with the lowest Thom values. On the other hand, the lengths of the incubation periods
increase monotonously with Thom . Based on these observations the development of the dislocation
density can be discussed as follows. At the very beginning of ball milling a rapid production of
dislocations is necessary in order to enable plastic deformation. In the case of CaF2 and CdF2 , once
a relatively large number of dislocations (about 2.5 · 1015 m−2 ) are present, due to the higher Thom
and lower Tm , it can be assumed that the existing dislocations are mobile and plastic deformation can
proceed relatively easily without further increase of ρ. However, at the end of the incubation period,
further plastic deformation needs even more dislocations, therefore ρ increases rapidly again up to
values of about 7 · 1015 m−2 . In the case of BaF2 and SrF2 , where Thom is lower and Tm higher, the
mobility of dislocations is considerably restricted as compared to the other two fluorides. Therefore,
plastic deformation can only be enabled by intensive dislocation production resulting in a rather short
incubation-like stage followed immediately by further rapid increase of ρ. Finally, this rapid increase
is followed by the saturation of ρ at about 7 · 1015 m−2 which is close to the theoretical maximum
value of average dislocation densities (Essmann et al., 1981).

Figure 5.23: Size distribution functions of SrF2 samples ball milled for 6, 12 and 60 min.

90
 Figure 5.24: Size distribution functions of SrF2 samples ball milled for 30, 40 and 60 min.

Figure 5.25: The area averaged mean crystallite size as a function of milling period for the different
specimens.

91
Figure 5.26: The dislocation density as a function of milling period for the different specimens.

Figure 5.27: The incubation period as a function of homologous temperature.

92
Chapter 6

Summary and Conclusions

Most of the physical properties of crystalline materials are fundamentally determined by the mi-
crostructure. The most important fundamental microstructural properties are the type, density and
distribution of crystal defects and the grain or crystallite structure. X-ray line profile analysis (XLPA)
is one of the most important and powerful alternative methods besides electron microscopy for the
characterization of microstructures.

This dissertation presents the theoretical models of the microstructure, the classical methods of
XLPA and my newly developed methods, the MWP and CMWP procedures, which are based on the
modeling of the physical properties of the different crystal defects. Several applications are presented
in order to show the efficiency of the MWP and CMWP methods.

• The theoretical models of the microstructure are presented in Chapter 2: the size broadening
of the profiles is calculated for the most important crystallite shapes and lognormal crystallite
size distribution. The Fourier transform of the size profile is also given. The Wilkens model of
dislocations is a sophisticated model for the description of the strain profile corresponding to a
physically realistic dislocation configuration. By linking it with the theory of strain anisotropy,
the model can be applied for a general description of strain broadening of diffraction peaks.
The broadening caused by planar faults is also discussed.

• Chapter 3 gives a review of the classical methods of XLPA. The basic breadth and Fourier
methods: the classical and modified Williamson-Hall methods and the classical and modified
Warren-Averbach procedures are presented. The recently developed full profile and whole pat-
tern fitting methods and the moment methods are also presented.

• Chapter 4 describes the MWP method, which is a Fourier method, based on the modeling of
the physical properties of the X-ray diffraction profiles for the different crystal defects. By
comparing the so obtained model based line profiles with the measured ones the method gives
the microstructural parameters of the investigated materials. The steps of the data preparation
and the steps of the MWP evaluation procedure are presented. Several applications of the

93
 method have been selected and the results provided by MWP have been compared to the results
of other microstructural methods, e.g. electron microscopy.

• In Chapter 5, the CMWP method is presented, which is also based on the modeling of the phys-
ical properties of the different crystal defects, but instead of fitting the separated peak profiles, it
works directly on the measured intensity pattern. By comparing the model based pattern to the
measured one the microstructural parameters are determined. The principles of the method, the
steps of the evaluation procedure and the several available frontends of the CMWP program are
presented. The details of the numerical procedure, e.g. the determination of the background,
the usage of the instrumental profiles, the weighting algorithms and the different aspects of the
sampling of the intensity pattern and the Fourier transforms are also given. The convergence and
numerical stability of the procedure are also examined. By applying both the MWP and CMWP
procedures for the determination of the microstructural parameters of ball milled Al-Mg alloys,
the results obtained by the two different procedures were compared. The method was applied
to nanocrystalline Cu and the planar fault density was determined. A special application of the
CMWP procedure for systematically prepared ball milled fluorides is presented: for some of
the samples an unusual X-ray optical interference effect was observed and the microstructural
parameters have been determined by correcting for this effect.

By applying the MWP and CMWP methods, the microstructural properties of several different mate-
rials or groups of materials are determined and these properties are examined as a function of several
mechanical or thermal treatments.

The most important results of my work are:

• The size profile [S1] and the size Fourier transform of coherently scattering domains with
spherical or ellipsoidal crystallite shape and lognormal size distribution have been determined
[S4, S5, S6, S8]. The expression of the Fourier transform supports the fast numerical evaluation.

• The method for applying the strain profile for XLPA based on the Wilkens model of dislocations
and the model of average contrast factors have been worked out [S4, S6, S14].

• By using the theoretical size and strain profiles new methods have been developed for the ex-
traction of microstructural parameters from X-ray measurements:

– by fitting simultaneously the whole Fourier transforms or intensity profiles obtained by

separation and instrumental deconvolution, (the method Multiple Whole Profile Fitting:
MWP) [S4, S6], and
– by the convolutional fitting (with the inclusion of the instrumental effect) of the whole
diffraction spectrum, (the method of Convolutional Multiple Whole Profile Fitting:
CMWP) [S14].

94
 • A software package has been developed implementing the methods described in sections 4 and
5, which is available to the public [S4, S6, S14] through the web: http://www.renyi.hu/mwp
and http://www.renyi.hu/cmwp.

• By using the MWP method I have shown that:

– the dislocation density has the average value of 1016 m−2 in severely deformed Ti, which
is in accordance with electron microscopy investigations; and in this deformation stage
mainly the slip systems <a> and <c+a> are activated [S11, S12, S22],
– by applying ball milling and heat treatment of PbS (galena) samples, a systematic set of
samples was produced and by evaluating the X-ray line profiles of these samples using the
MWP method a map of microstructure was obtained. By comparing the microstructural
parameters of samples from ancient cosmetics fabricated in the Egyptian Kingdom 3500
b.c. with the parameters of the systematically prepared samples I have shown that only
gentle crushing and no heat treatment or heat treatment at a temperature of less than 300
◦ C was applied [S9] for the fabrication of the archaeological cosmetics.

• By using the CMWP method I have shown that:

– in ball milled Al-Mg alloys the dislocation density as well as the average crystallite size
shows a saturation after 2h of ball milling and the results obtained by the MWP and
CMWP methods have been compared [S13, S16, S14],
– in ball milled fluorides (MF2 , M=Ca,Ba,Cd,Sr) the presence of an X-ray optical interfer-
ence effect can be observed which is present primarily if the average crystallite size of
a larger portion of crystallites is about 5-10 nm or smaller [S17]. The X-ray optical in-
terference effect means that the first few peak profiles become narrower than what would
correspond to the domain size.

• I have shown that if the method of XLPA is carried out correctly, which means that the ex-
periments are done with an angular resolution high enough and that the evaluation is based
on correct microstructural models, then good correlation is observed between TEM and XLPA
parameters. For example, in the case of nanocrystalline Si3 N4 particles the size distribution
determined either by TEM or XLPA are in excellent agreement, cf. [S1]. In a large number of
cases very good correlation was found between the microstructural parameters, especially the
subgrain size, and size distribution, the dislocation density and dislocation types determined
either by TEM or XLPA [S1–S22].

As the results of this work, a coherent set of methods has been developed which is suitable for the
characterization of the microstructure of most different crystalline materials, e.g. metals, alloys,
ceramic materials, minerals or polymers, in terms of the size and size distribution of crystallites or
grains, the density, the distribution and character of dislocations and the density and type of planar
defects.

95
Own Publications

[S1] Gubicza, J., Szépvölgyi, J., Mohai, I., Ribárik, G. and Ungár, T., The effect of heat-treatment
on the grain-size of nanodisperse plasmathermal silicon nitride powder, J. of Materials Science,
35, 3711 (2000)

[S2] Gubicza, J., Ribárik, G., Bakonyi, I. and Ungár, T., Crystallite-size distribution and dislo-
cation stucture in nanocrystalline HfNi5 determined by X-ray diffraction profile analysis, J.
Nanoscience & Nanotechnology, 1, 343–348, (2001)

[S3] Gubicza, J., Ribárik, G., Goren-Muginstein, G. R., Rosen, A. R. and Ungár, T., The density and
the character of dislocations in cubic and hexagonal polycrystals determined by X-ray diffrac-
tion, Mat. Sci. and Eng. A, 309–310, 60–63 (2001)

[S4] Ribárik, G., Ungár, T. and Gubicza, J., MWP-fit: a program for multiple whole-profile fitting of
diffraction peak profiles by ab initio theoretical functions, J. Appl. Cryst., 34, 669–676 (2001)

[S5] Ungár, T., Gubicza, J., Ribárik, G., Microstructure of carbon blacks determined by X-ray diffrac-
tion profile analysis, Carbon, 40, 929–937, (2001)

[S6] Ungár, T., Gubicza, J., Ribárik, G. and Borbély, A., Crystallite size-distribution and dislocation
structure determined by diffraction profile analysis: Principles and practical application to cubic
and hexagonal crystals, J. Appl. Cryst., 34, 298–310 (2001)

[S7] Ungár, T., Gubicza, J., Ribárik, G. and Hanák, P., Dislocation structure and crystallite size
distribution in plastically deformed metals determined by diffraction peak profile analysis, J.
Eng. Mat. and Technology (ASME), 124, 2–6 (2002)

[S8] Ungár, T., Gubicza, J., Ribárik., G., Pantea, C. and Zerda, T. W., Microstructure of carbon blacks
determined by X-ray diffraction profile analysis, Carbon, 40, 929–937 (2002)

[S9] Ungár, T., Martinetto, P., Ribárik, G., Dooryhee, E., Walter, Ph. and Anne, M., Revealing the
powdering methods of black makeup in Ancient Egypt by fitting microstructure based Fourier
coefficients to the whole X-ray diffraction profiles of galena, J. Appl. Phys., 91, 2455–2465
(2002)

96
[S10] Borbély, A., Dragomir, I. C., Ribárik, G, Ungár, T., Computer program ANIZC for the calcula-
tion of diffraction contrast factors of dislocations in elastically anisotropic cubic, hexagonal and
trigonal crystals, J. Appl. Cryst., 36, 160–162, (2003)

[S11] Gubicza, J., Dragomir, I. C., Ribárik, G., Baik, S. C., Zhu, Y. T., Valiev, R. Z. and Ungár, T.,
Dislocation structure and crystallite size-distribution in plastically deformed Ti determined by
X-ray peak profile analysis, Z. Metallkunde, 94, 1185–1188 (2003)

[S12] Gubicza, J., Dragomir, I. C., Ribárik, G., Zhu, Y. T., Valiev, R. Z. and Ungár, T., Microstructure
of severely deformed titanium from X-ray diffraction profile analysis, Mat. Sci. Forum, 414–415,
229–234 (2003)

[S13] Gubicza, J., Kassem, M., Ribárik, G. and Ungár, T., The evolution of the microstructure in
mechanically alloyed Al-Mg studied by X-ray diffraction, Mat. Sci. Eng. A, A372, 115–122
(2004)

[S14] Ribárik, G., Gubicza, J. and Ungár, T., Correlation between strength and microstructure of ball
milled Al-Mg alloys determined by X-ray diffraction, Mat. Sci. Eng. A, A387–389, 343–347
(2004)

[S15] Dragomir, I. C., Li, D. S., Castello-Branco, G. A., Garmestani, H., Snyder, R. L., Ribárik, G.,
Ungár, T., Evolution of dislocation density and character in hot rolled titanium determined by
X-ray diffraction, Materials Characterization, 55, 66–74 (2005)

[S16] Révész, Á., Nagy, L., Ribárik, G., Kovács, Zs., Ungár, T., Lendvai, J., Microstructural
evolution in mechanically alloyed nanocrystalline Al-20at.% Mg alloy, J. of Metastable and
Nanocrystalline Materials, 24-25, 149–152 (2005)

[S17] Ribárik, G., Audebrand, N., Palancher, H., Ungár, T., Louër, D., Dislocations and crystallite
size distributions in ball-milled nanocrystalline fluorides MF2 (M = Ca, Sr, Ba, Cd) determined
by X-ray diffraction-line-profile analysis, J. Appl. Cryst., 38, 912–926, (2005)

[S18] Balogh, L., Ribárik, G. and Ungár, T., Stacking Faults and Twin Boundaries in fcc Crystals
Determined by X-ray Diffraction Profile Analysis, J. Appl. Phys., 100, 023512 (2006)

[S19] Dragomir, I. C., Li, D. S., Castello-Branco, G. A., Garmestani, H., Snyder, R. L., Ribárik, G.
and Ungár, T., Microstructure evolution in hot rolled titanium determined by X-ray diffraction
peak profile analysis, Z. Kristallogr. Suppl., 23, 99–104 (2006)

[S20] Fátay, D., Spassov, T., Delchev, P., Ribárik, G., Révész, Á., Microstructural development in
nanocrystalline MgH2 during H-absorption/desorption cycling, International Journal of Hydro-
gen Energy, in press (2007)

[S21] Ungár, T., Castelnau, O., Ribárik, G., Drakopoulos, M., Béchade, J. L., Chauveau, T., Snigirev,
A., Snigireva, I., Schroer, C. and Bacroix, B., Grain to grain slip activity in plastically deformed

97
 Zr determined by X-ray microdiffraction line profile analysis, Acta Materialia, 55, 1117–1127
(2007)

[S22] Ungár, T., Glavicic, M. G., Balogh, L., Nyilas, K., Salem, A. A., Ribárik, G., Semiatin, S. L.,
The Use of X-Ray Diffraction to Determine Slip and Twinning Activity in Commercial-Purity
(CP) Titanium, Mat. Sci. Eng. A, in press (2007)

[S23] Ribárik, G, Aifantis, E. C., Konstantinidis, A. and Ungár, T., in preparation (2007)

98
Bibliography

[1] Aitchinson, J. and Brown, J., The Lognormal Distribution, Cambridge University Press, Cam-
bridge (1957)

[2] Arley, N. and Buch, K. R., Introduction to the Theory of Probability and Statistics, New York -
Wiley, (1950)

[3] Balogh, L., Ribárik, G. and Ungár, T., J. Appl. Phys., 100, 023512 (2006)

[4] Balogh et al., in preparation (2008)

[5] Balzar, D., J. Res. Natl. Inst. Stand. Technol., 98, 321 (1993)

[6] Berkum, J. G. M. van, PhD thesis, Delft University of Technology, The Netherlands (1994)

[7] Bertaut, F. C. R., Acad. Sci., Paris, 228, 187 (1949)

[8] Borbély, A., Dragomir, I. C., Ribárik, G. and Ungár, T., J. Appl. Cryst., 36, 160–162 (2003)

[9] Borbély, A., Driver, J. H. and Ungár, T., Acta Mat., 48, 2005–2016 (2000)

[10] Borbély, A. and Groma, I., Appl. Phys. Lett., 79, 1772–1774 (2001)

[11] Caglioti, G., Paoletti, A. and Ricci, F. P., Nucl. Instrum., 3, 223–234 (1958)

[12] Cordier, P., Ungár, T., Zsoldos, L., Tichy, G., Nature, 428, 837–840 (2004)

[13] Csikor, F. F. and Groma, I., Phys. Rev. B, 70, 064106 (2004)

[14] Dong, Y. H. and Scardi, P., J. Appl. Crystallogr., 33 (1999)

[15] Dragomir, I. C., Li, D. S., Castello-Branco, G. A., Garmestani, H., Snyder, R. L., Ribárik, G.,
Ungár, T., Materials Characterization, 55, 66–74 (2005)

[16] Dragomir, I. C., Li, D. S., Castello-Branco, G. A., Garmestani, H., Snyder, R. L., Ribárik, G.
and Ungár, T., Z. Kristallogr. Suppl., 23, 99–104 (2006)

[17] Dragomir, I. C. and Ungár, T., Powder Diffr. Journ., 17, 104–111 (2002)

[18] Essmann, U., Gösele, U. and Mughrabi, H., Philos. Mag. A, 44, 405–426 (1981)

99
[19] Fátay, D., Spassov, T., Delchev, P., Ribárik, G., Révész, Á., International Journal of Hydrogen
Energy, in press (2007)

[20] Gaál, I., Acta Phys. Acad. Sci. Hung. 33, 411–418 (1973)

[21] Gaál, I., Dissertation to the Hungarian Academy of Sciences, Budapest, Hungary (1976)

[22] Gaál, I., Proc. 5th Riso Int. Syrup. Metall. Mater. Sci., edited by N. H. ANDERSEN, M. EL-
DRUP, N. HANSEN, D. JUUL JENSEN, T. LEFFERS, H. LIFHOLT, O. B. PEDERSEN B. N.
SINGH, pp. 249-254. Riso Natl Lab., Roskilde, Denmark (1984)

[23] Groma, I., Phys. Rev. B, 57, 7535–7542 (1998)

[24] Groma, I. and Monnet, G., Journal of Applied Cryst., 35, 589-593 (2002)

[25] Groma, I., DSc thesis, Eötvos Loránd University, Budapest, Hungary (2003)

[26] Gubicza, J., Dragomir, I. C., Ribárik, G., Baik, S. C., Zhu, Y. T., Valiev, R. Z. and Ungár, T., Z.
Metallkunde, 94, 1185–1188 (2003)

[27] Gubicza, J., Dragomir, I. C., Ribárik, G., Zhu, Y. T., Valiev, R. Z. and Ungár, T., Mat. Sci.
Forum, 414–415, 229–234 (2003)

[28] Gubicza, J., Juhász, A., Lendvai, J., J. Mater. Res., 11, 2964 (1996)

[29] Gubicza, J., Kassem, M., Ribárik, G. and Ungár, T., Mat. Sci. Eng. A, A372, 115–122 (2004)

[30] Gubicza, J., Ribárik, G., Bakonyi, I. and Ungár, T., J. Nanoscience & Nanotechnology, 1, 343–
348, (2001)

[31] Gubicza, J., Ribárik, G., Goren-Muginstein, G. R., Rosen, A. R. and Ungár, T., Mat. Sci. and
Eng. A, 309–310, 60–63 (2001)

[32] Gubicza, J., Rozlosnik, N., Juhász, A., J. Mater. Sci. Lett., 16, 1904 (1997)

[33] Gubicza, J., Szépvölgyi, J., Mohai, I., Ribárik, G. and Ungár, T., J. of Materials Science, 35,
3711 (2000)

[34] Guerrero-Paz, J. and Jaramillo-Vigueras, D., Nanostruct. Mater., 11, 1123–1132 (1999)

[35] Guinier, A., X-ray diffraction, Freeman and Co., San Francisco, CA (1963)

[36] Hall, E. O., Proc. Phys. Soc., Ser. B, Vol. 64, 747–753, (1951)

[37] Hecker, M., Thiele, E. and Holste, C., Mat. Sci. Eng. A, 234–236, 806–809 (1997)

[38] Hinds, W. C., Aerosol Technology: Properties, Behavior and Measurement of Airbone Particles,
Wiley, New York (1982)

100
[39] Janot, R. and Guerard, D., Prog. Mater. Sci., 50, 1–92 (2005)

[40] Kamminga, J. D. and Delhez, R., J. Appl. Cryst., 33, 1122–1127 (2000)

[41] Krill, C. E. and Birringer, R., Philos. Mag. A, 77, 621–640 (1998)

[42] Krivoglaz, M. A., Theory of X-ray and Thermal Neutron Scattering by Real Crystals, Plenum
Press New York (1969)

[43] Krivoglaz, M. A. and Ryaboshapka, K. P., Fizika Metall., 15, No. 1, 18 (1963)

[44] Langford, J. I. and Wilson, A. J. C., J. Appl. Cryst., 11, 102–113 (1978)

[45] Langford, J. I., Louër, D. and Scardi, P., J. Appl. Cryst., 33, 964–974 (2000)

[46] Leoni, M. and Scardi, P., J. Appl. Cryst., 37, 629–634 (2004)

[47] Levenberg, K., Quart. Appl. Math., 2, 164–168 (1944)

[48] Louër, D. and Audebrand, N., Adv. X-ray Anal., 41, 556–565 (1999)

[49] Marquardt, D. W., SIAM J. Appl. Math., 11, 431–441 (1963)

[50] Martinetto, P., Anne, M., Dooryhée, E., Tsoucaris, G. and Walter, Ph., Radiation in Art and
Archeometry, edited by D. Creagh and D. Bradley, Elsevier Science BV, Prague (2000)

[51] Martinetto, P., Castaing, J., Walter, P., Penhoud, P. and Veyssiere, P., J. Mater. Res., 17, 1601–
1611 (2002)

[52] Mughrabi H., Ungar T., Kienle W., Wilkens M., Phil. Mag., 53, 793 (1986)

[53] Nelson, J. B., Riley, D. P., Proc. Phys. Soc. London, 57, 160 (1945)

[54] Nusinovici, J. and Rehfeldt-Osierski, A., Collected Abstracts, IUCr Powder Diffraction Satellite
Meeting, Toulouse, France, 315–316 (1990)

[55] Pál, F., Tóth, K., Pungor, E., Farkas-Jahnke, M., Ebel, H. and Ebel, M. F., Anal. Chim Acta, 180,
313–321 (1986)

[56] Paton, N. E. and Backofen, W. A., Metallurgical Trans., 1, 2839–2847 (1970)

[57] Pederson, D. O. and Brewer, J. A., Phys. Rev. B, 16, 4546–4550 (1977)

[58] Petch, N.J., J. Iron and Steel Institute, May, 25–28 (1953)

[59] Pool, D. M. and Axon, H. J., J. Inst. Met., 80, 599 (1952)

[60] Rác, B., Mulas, G., Csongrádi, A., Loki, K. and Molnár, A., Appl. Catal. A, 282, 255–265 (2005)

101
[61] Rafaja, D., Klemm, V., Schreiber, G., Knapp, M. and Kužel, R., J. Appl. Cryst., 37, 613–620
(2004)

[62] Révész, Á., Nagy, L., Ribárik, G., Kovács, Zs., Ungár, T., Lendvai, J., J. of Metastable and
Nanocrystalline Materials, 24-25, 149–152 (2005)

[63] Ribárik, G, Aifantis, E. C., Konstantinidis, A. and Ungár, T., in preparation (2007)

[64] Ribárik, G., Audebrand, N., Palancher, H., Ungár, T., Louër, D., J. Appl. Cryst., 38, 912–926,
(2005)

[65] Ribárik, G., Gubicza, J. and Ungár, T., Mat. Sci. Eng. A, A387–389, 343–347 (2004)

[66] Ribárik, G., Ungár, T. and Gubicza, J., J. Appl. Cryst., 34, 669–676 (2001)

[67] Scardi, P. and Leoni, M., J. Appl. Cryst., 32, 671–682 (1999)

[68] Scardi, P. and Leoni, M., Acta Cryst., A58, 190–200 (2002)

[69] Scherrer, P., Göttinger Nachrichten; R. Zsigmondy, Kolloidchemie (3rd Ed. 1920), pp. 394
(1918)

[70] Somogyvári Z., Sváb, E., Mészáros, Gy., Krezhov, K., Konstantinov, P., Ungár, T. and Gubicza,
J., Mat. Sci. Forum, 378-381, 771-776 (2001)

[71] Sprackling, M. T., The Plastic Deformation of Simple Ionic Crystals, Academic Press, London
(1976)

[72] Stokes, A. R., Proc. Phys. Soc., 61, 382–393 (1948)

[73] Stokes, A. R. and Wilson, A. J. C., Proc. Phys. Soc., 56, 174–180 (1944)

[74] Terwilliger, Ch. D. and Chiang, Y. M., Acta Metall. Mater., 43, 319–328 (1995)

[75] Treacy, M. M. J., Newsam J. M., and Deem, M. W., Proc. R. Soc. London, Ser. A 433, 499
(1991)

[76] Ungár, T., Alexandrov, I. and Hanák, P., Proc. Int. Conf. Severe Plastic Deformation, Moscow,
1999, KLUWER A. P. (2000)

[77] Ungár, T. and Borbély, A., Appl. Phys. Lett., 69, 3173–3175 (1996)

[78] Ungár, T., Castelnau, O., Ribárik, G., Drakopoulos, M., Béchade, J. L., Chauveau, T., Snigirev,
A., Snigireva, I., Schroer, C. and Bacroix, B., Acta Materialia, 55, 1117–1127 (2007)

[79] Ungár, T., Dragomir, I. C., Louër, D. and Audebrand, N., J. Phys. Chem. Solids, 62, 1935–1941
(2001)

102
[80] Ungár, T., Dragomir, I. C., Révész, Á. and Borbély, A., J. Appl. Cryst., 32, 992–1002 (1999)

[81] Ungár, T., Glavicic, M. G., Balogh, L., Nyilas, K., Salem, A. A., Ribárik, G., Semiatin, S. L.,
Mat. Sci. Eng. A, in press (2007)

[82] Ungár, T., Gubicza, J., Ribárik, G., Carbon, 40, 929–937, (2001)

[83] Ungár, T., Gubicza, J., Ribárik, G. and Borbély, A., J. Appl. Cryst., 34, 298–310 (2001)

[84] Ungár, T., Gubicza, J., Ribárik, G. and Hanák, P., J. Eng. Mat. and Technology (ASME), 124,
2–6 (2002)

[85] Ungár, T., Gubicza, J., Ribárik., G., Pantea, C. and Zerda, T. W., Carbon, 40, 929–937 (2002)

[86] Ungár, T., Martinetto, P., Ribárik, G., Dooryhee, E., Walter, Ph. and Anne, M., J. Appl. Phys.,
91, 2455–2465 (2002)

[87] Ungár, T., Mughrabi, H., Rönnpagel, D. and Wilkens, M., Acta Met., 32, 333–341 (1984)

[88] Ungár, T., Ott, S., Sanders, P. G., Borbély, A. and Weertman, J. R., Acta Mater., 10, 3693 (1998)

[89] Ungár, T. and Tichy, G., Phys. Status Solidi. A, 171, 425–434 (1999)

[90] Ungár, T., Tichy, G., Gubicza, J., Hellmig, R. J., J. Powder Diffr., 20, 366 (2005)

[91] Valiev, R. Z., Kozlov, E. V., Ivanov, Yu., F., Lian, J., Nazarov, A. A. and Baudelet, B., Acta
Metall. Mater., 42, 2467–2476 (1994)

[92] Wakai, F. and Aldinger, F., Acta Mater., 51, 641–652 (2003)

[93] Walter, Ph., Martinetto, P., Tsoucaris, G., Bréniaux, R., Lefebvre, M-A., Richard, G., Talabot, J.
and Dooryhée, E., Nature, 397, 483–484 (1999)

[94] Warren, B. E., Progr. Metal Phys., 8, 147–202 (1959)

[95] Warren, B. E., X-ray Diffraction, Reading, Addison-Wesley, MA (1969)

[96] Warren, B. E. and Averbach, B. L., J. Appl. Phys., 23, 497 (1952)

[97] Warren, B. E and Bodenstein, P., Acta Cryst., 18, 282 (1965)

[98] Wilkens, M., Theoretical aspects of kinematical X-ray diffraction profiles from crystals con-
taining dislocation distributions, Fundamental Aspects of Dislocation Theory, edited by J. A.
Simmons, R. de Wit and R. Bullough., Vol. II, pp. 1195-1221, Nat. Bur. Stand. (US) Spec. Publ.
No. 317., Washington, DC, USA (1970)

[99] Wilkens, M., Proc. Int. Conf. Strenght Met. Alloys (ICSMA 8), Tampere, Finland, 1988, eds.
Kettunen, P. O., Lepistö, T. K., Lehtonen, M. E., Pergamon Press, 47–152 (1988)

103
[100] Williamson, G. K. and Hall, W. H., Acta metall., 1, 22–31 (1953)

[101] Wilson, A. J. C., X-ray Optics, Methuen, London (1962)

[102] Winther, G., Huang, X., Godfrey A. and Hansen N., Acta Materialia, 52, 4437–4446 (2004)

[103] Yao, Y. and Thölén, A., Mater. Charact., 44, 441–452 (2000)

[104] York, B. R., Adv. X-ray Anal., 41, 544–555 (1999)

[105] Zhu, Y. T., Huang, J. Y., Gubicza, J., Ungár, T., Wang, Y. M., Ma, E., Valiev, R. Z., J. Mat.
Res., 18, 1908–1917 (2003)

104
Acknowledgements

I express my deep gratitude to my supervisor, Prof. Tamás Ungár for introducing, encouraging and
guiding me in the field of materials science. I am in his debt for his care and boundless patience in
revising this manuscript.

I’m grateful to my family, first of all to my wife, Orsolya Ribárik for her help in typography and
to my son, Balázs Ribárik for his patience and letting me work.

I would like to thank for the fruitful cooperation to the co-authors of our common papers, espe-
cially to Prof. Tamás Ungár, Dr. Jenő Gubicza, Dr. Ádám Révész, Dr. Iulianna Dragomir and Mr.
Levente Balogh.

Thanks are due to my colleagues, especially to Prof. István Groma, Dr. András Borbély, Dr. Erhard
Schafler, Mr. Michael Kerber, Dr. Avraam Konstantinidis and Prof. E.C. Aifantis for the inspiring
discussions.

I would like to thank for the careful reading of the thesis to Dr. Jenő Gubicza and Mr. Michael Kerber.

I am especially grateful to Prof. János Lendvai for the opportunity to have been able to work at
the Department of Materials Physics of the Eötvös Loránd University.
 Short summary

Most of the physical properties of crystalline materials are fundamentally determined by the mi-
crostructure. Electron microscopy is an important method for the visualization of the microstructure.
The most important fundamental microstructural properties are the type, density and distribution of
crystal defects and the grain or crystallite structure. X-ray line profile analysis (XLPA) is one of the
most important and powerful alternative methods besides electron microscopy for the characterization
of microstructures. This method gives information about the following basic microstructural proper-
ties: (i) the size and distribution of crystallites, (ii) the shape anisotropy of crystallites, (iii) the den-
sity, character and distribution of dislocations and (iv) the density and type of stacking faults and twin
boundaries. This dissertation presents the most important theoretical models of the microstructure
according to the literature, my results related to the development of these models, the most important
methods of XLPA: the classical ones and some of the most recent methods of literature. The most
important part of my work is the development and implementation of new XLPA methods, which are
based on the modeling of the physical properties of the different crystal defects. By comparing the
so obtained model based line profiles with the measured ones the methods give the microstructural
parameters of the investigated materials. By applying these methods the microstructural properties of
several different materials or groups of materials are determined and these properties are examined as
a function of several mechanical or thermal treatments.
The most important new scientific results of my work are: (i) I have determined the size profile [S1]
and the size Fourier transform of coherently scattering domains with spherical or ellipsoidal crystal-
lite shape and lognormal size distribution [S4, S5, S6, S8]. (ii) I elaborated the method for applying
the strain profile for XLPA based on the Wilkens model of dislocations and the model of average
contrast factors [S4, S6, S14]. (iii) I developed two new methods by using the theoretical size and
strain profiles for the extraction of microstructural parameters from X-ray measurements: (a) by fit-
ting simultaneously the whole Fourier transforms or intensity profiles, (the method Multiple Whole
Profile Fitting: MWP) [S4, S6], and (b) by the convolutional fitting of the whole diffraction spec-
trum, (the method of Convolutional Multiple Whole Profile Fitting: CMWP) [S14]. (iv) I worked
out a software package for these methods described which is publicly available [S4, S6, S14]. (v)
By using the MWP method I have shown that: (a) the dislocation density has the average value of
1016 m−2 in severely deformed Ti, which is in accordance with electron microscopy investigations;
and in this deformation stage mainly the slip systems <a> and <c+a> are activated [S11, S12, S22],
(b) by applying ball milling and heat treatment of PbS (galena) samples, a systematic set of samples
has been produced and by evaluating the X-ray line profiles of these samples using the MWP method
a map of microstructure has been obtained. By comparing the microstructural parameters of samples
from ancient cosmetics fabricated in the Egyptian Kingdom 3500 b.c. with the parameters of the
systematically prepared samples I have shown that only gentle crushing and no heat treatment or heat
treatment at a temperature of less than 300 ◦ C was applied [S9] for the fabrication of the archaeolog-
ical cosmetics. (vi) By using the CMWP method I have shown that: (a) in ball milled Al-Mg alloys
the dislocation density as well as the average crystallite size shows a saturation after 2h of ball milling
and the results obtained by the MWP and CMWP methods have been compared [S13, S16, S14], (b)
in ball milled fluorides (MF2 , M=Ca,Ba,Cd,Sr) the presence of an X-ray optical interference effect
can be observed which is present primarily if the average crystallite size of a larger portion of crystal-
lites is about 5-10 nm or smaller [S17]. (vii) I have shown that in nanocrystalline Si3 N4 particles the
size distribution determined either by TEM or XLPA are in excellent agreement, cf. [S1]. In a large
number of cases very good correlation was found between the microstructural parameters, especially
the subgrain size, and size distribution, the dislocation density and dislocation types determined either
by TEM or XLPA [S1–S22].
As the results of this work, a coherent set of methods has been developed which is suitable for the
characterization of the microstructure of most different crystalline materials.
 Rövid összefoglalás

Kristályos anyagok számos fizikai tulajdonságát alapvetően meghatározza a mikroszerkezet. A

mikroszerkezet szemléltetésének fontos módszere az elektronmikroszkópia. A legalapvetőbb mik-
roszerkezeti tulajdonságok a kristályhibák típusa, sűrűsége és eloszlása valamint a szemcse-, il-
letve krisztallit szerkezet. Az elektronmikroszkópia mellett az egyik legfontosabb alternatív mód-
szer a röntgen vonalprofil analízis (RVPA). Ez a módszer alapvetően a következő mikroszerkezeti
tulajdonságokról ad felvilágosítást: (i) krisztallitok méretéről és méreteloszlásról, (ii) krisztalli-
tok alak anizotrópiájáról, (iii) diszlokációk sűrűségéről, típusáról és eloszlásáról, valamint (iv)
rétegződési hibák illetve iker határok sűrűségéről és típusáról. A disszertációban bemutatom a
mikroszerkezet legfontosabb irodalomban fellelhető modelljeit, illetve eredményeimet melyek ezek
továbbfejlesztésével kapcsolatosak, ismertetem a RVPA fontosabb módszereit a klasszikusaktól
kezdve a legújabb irodalomban fellelhető módszerekig. Munkám legjelentősebb része a RVPA olyan
módszereinek a kifejlesztése, amelyek a különböző kristályhibák fizikai tulajdonságai alapján mo-
dellezik a röntgendiffrakciós vonalprofilokat. Ezeknek a modellezett vonalprofiloknak a mérésekkel
való összevetéséből, a kifejlesztett módszerek segítségével megkaphatjuk a vizsgált anyag mikro-
szerkezeti paramétereit. A kidolgozott módszerek alkalmazásával meghatározom számos különböző
anyag illetve anyagcsalád mikroszerkezeti tulajdonságait, valamint azt, hogy ezek hogyan változnak
meg különböző mechanikai vagy termikus kezelések hatására.
Ennek a munkának a során kapott főbb új tudományos eredményeim a következők: (i) Meghatároz-
tam gömb- [S1], illetve forgási ellipszoid alakú, lognormális méreteloszlású koherens domének méret
profil függvényeit és azok Fourier transzformáltját [S4, S5, S6, S8]. (ii) Kidolgoztam a deformá-
ciós profil felhasználásának módszerét a diszlokációk Wilkens-féle modellje és az átlagos kontraszt
faktorok modellje alapján [S4, S6, S14]. (iii) Az elméleti méret- és deformációs profilok alapján
módszereket dolgoztam ki a mikroszerkezet paramétereinek meghatározására röntgendiffrakciós
mérésekből: (a) a teljes Fourier transzformáltak, illetve teljes intenzitásprofilok együttes illesztésével
(Multiple Whole Profile Fitting: MWP) [S4, S6], illetve (b) a teljes intenzitás spektrum konvolú-
ciós illesztésével (Convolutional Multiple Whole Profile Fitting: CMWP) [S14]. (iv) Bárki számára
elérhető programcsomagot fejlesztettem ki ezekhez a módszerekhez [S4, S6, S14]. (v) Az MWP
eljárás alapján megmutattam, hogy: (a) erőteljesen alakított Ti-ban a diszlokációsűrűség 1016 m−2
átlagos értéket ér el, összhangban az elektronmikroszkópos vizsgálatokkal; továbbá, hogy ilyen de-
formált állapotban döntően a <a> és <c+a> típusú csúszási rendszerek dominálnak [S11, S12, S22],
(b) a PbS (galenit) őrlésével és hőkezelésével készült mikroszerkezeti állapottérkép alapján megál-
lapítottam, hogy az ie. 3500 Egyiptomi Királyságokban használt kozmetikumok készítésekor csupán
rövid idejű őrlést és 300 ◦ C-nál nem magasabb hőmérsékletű hőkezeléseket alkalmaztak [S9]. (vi)
A CMWP eljárás alapján megmutattam, hogy: (a) golyós malomban őrölt Al-Mg ötvözetben mind
a diszlokáció sűrűség, mind pedig a szemcseméret 2 óra őrlés után telítődést mutat, továbbá az
MWP és CMWP módszerekkel kapott eredményeket összehasonlítottam [S13, S16, S14], (b) golyós
malomban őrölt alkáliföldfém fluoridokban kimutattam egy újszerű röntgen optikai interferencia
effektus fellépését, amely elsősorban akkor jelenik meg, ha a méreteloszlás jelentős hányadában a
szemcseméret 5-10 nm, vagy annál kisebb [S17]. (vii) Nanokristályos Si3 N4 részecskéinek TEM
és röntgen méreteloszlására igen jó egyezést kaptam [S1]. Számos esetben igazolódott, hogy az
elektronmikroszkópos szubszemcse méret, diszlokációsűrűség, valamint diszlokáció típus kitünően
egyezik a vonalprofil analízis módszerével nyert értékekkel [S1–S22].
Munkám eredményeképpen előállt egy olyan koherens módszer együttes, amely alkalmas a legkülön-
bözőbb kristályos anyagok, nevezetesen, fémek, ötvözetek, kerámiák, kőzetek, illetve polimerek
mikroszerkezetének jellemezésére a krisztallit és szemcseméret, a diszlokáció sűrűség, eloszlás és
típus, illetve a rétegződési hibák alapján.