X-Ray Analysis in Materials Chemistry

KEMI6510 (5 cu)

Mika Lastusaari
University of Turku, Department of Chemistry
Laboratory of Materials Chemistry and Chemical Analysis
Inorganic Materials Chemistry Research Group
Autumn 2016
GENERAL THINGS

Lecturer
Mika Lastusaari
Room K348, tel 02-333-6736
Email: [email protected]

Lectures (Wed 2.11. – Fri 9.12.2016)

Wed 14.15-16 Sem1
Fri 8.30-10 Sem1

Course material in Moodle

Key: XRA16

”Exercise”
Visit to the X-ray diffraction lab

Examination
On official examination days (December 12 and 20, 2016; beginning of 2017)
Based on lectures 2
INTENDED LEARNING OUTCOMES

The student is familiarized with different experimental X-ray techniques for

materials chemistry.
The student acquires the ability to select and use these techniques to different
research and industry based problems.

3
CONTENTS

1. Introduction
2. Generation and detection of X-rays
3. Safety issues
4. Elementary crystallography
5. X-ray diffraction
6. Applications of X-ray powder diffraction
7. X-ray absorption spectroscopy
8. X-ray fluorescence spectroscopy

4
1. INTRODUCTION
1.1. History and definition
• X-rays were discovered by German Physicist
Wilhelm Röntgen (Fig. 1) in 1895.
◘ He noticed a barium platinocyanide screen fluorescing in
his laboratory as he generated cathode rays (electron beam)
in a Crookes tube (Fig. 2) some distance away.

Figure 1. Wilhelm Röntgen.

Figure 2. Demonstration by Crookes that cathode rays travel in straight lines:

a) cathode; b) aluminum cross and anode; d) dark shadow; c) fluorescent image.
Wikipedia: Crookes tube in operation: several thousand volts of DC electricity is applied between the two electrodes in the
5
tube. The electrode to the right is the cathode or negative electrode. Electrons released by it are attracted toward the positive
space charge in center of the tube, created by the positive anode at bottom, and they accelerate down the tube to the left.
 ◘ 6 weeks later he brought his wife into his laboratory, and they emerged with a
photograph of the bones in her hand and of the ring on her finger (Fig. 3).
• The New-York Times (Jan. 16, 1896) predicted: “transformation of modern
surgery by enabling the surgeon to detect the presence of foreign bodies.”
• Within a month of Röntgen’s announcement doctors were using the X rays to
locate bullets in human flesh and photograph broken bones: “The surgical
imagination can pleasurably lose itself in
devising endless applications of this wonderful
process.” (The New-York Times, Feb. 15, 1896).

Figure 3. Hand of Anna Bertha, wife of Wilhelm

Röntgen. The black glob on the fourth finger is a ring
made of gold, which absorbs x-rays. (Image by
Wilhelm Röntgen, via SSPL/Science Museum/Getty
Images).

6
• In the first six months after their discovery Viennese mummies were undressed,
doctors claimed to have photographed their own brains, and the human heart
was uncovered.
• By 1897 the rays’ dangerous side began to be reported: examples included loss
of hair and skin burns of varying severity.
• After the discovery of X-rays, everything about the new rays was dazzling and
fascinating to the public. This resulted in the “X-ray mania”: X-rays appeared in
advertising, songs and cartoons (Figs. 4-7).

Figure 4. X-ray mania: Advertisement for ”X-ray” headache pills (Cosmopolitan, 1896) 7
(left) and can of ”X-ray” stove polish (right).
Figure 5. X-ray mania: Advertisement from the February 1917 issue of National 8
Geographic (left) extolling the virtue of so-called X-ray Reflectors (right).
Figure 6. X-ray mania: Shoe-fitting fluoroscope (1930-1940), the
scientific way of approaching the problem of poorly fitted shoes. 9
Figure 7. X-ray mania: Modern
ads (2014).

10
• Electricians and physicists speculated on the nature of these X-rays, but it wasn’t
until 1912 that the exact nature was established by the discovery of diffraction.
• DEFINITION: X-rays = radiation in the wavelength range 0.01 – 10 nm
(energy: 120–120000 eV).
◘ In the range of the sizes of molecules and atoms (Fig. 8).

Figure 8. The
electromagnetic
spectrum.

11
1.2. Interaction of X-rays with matter
• The interaction with matter (Fig. 9) can change the properties of the
original X-ray beam.

Figure 9. Interactions of X-rays with matter.

12
• The changes in the beam properties are due to the following processes:
◘ Coherent scattering (Rayleigh/Thomson scattering): the incident photon undergoes a
change in direction without a change in wavelength (Fig. 10).
◘ Photodisintegration: collision of a high energy photon with an atomic nucleus. The
photon is completely absorbed in the process, and a neutron, proton, or alpha particle
is ejected from the excited nucleus (Fig. 11).

Figure 10. Coherent scattering. Figure 11. Photodisintegration.

13
◘ Pair production: A high-energy photon is completely transformed into an electron and a
positron. Thus, this is a process whereby energy is transformed into matter.
◘ Compton effect: A photon impinges on an electron in matter, and in this process
transfers part of its energy to it. The excited electron is ejected or moved into an excited
atomic state, while due to the law of conservation of energy the photon energy is
reduced (Fig. 12).

Figure 12. Compton effect.

14
◘ Photoelectric effect: A photon transfers its entire energy to an electron in the material
on which it impinges. The electron thereby acquires enough energy either to free itself
from the material to which it is bound or to be elevated into the conduction band of a
semiconductor or insulator (Fig. 13).
 The created core hole is then filled with an electron, which will lead to the emission
of X-rays or Auger electrons.
 Element specific (characteristic) absorption and X-ray emission energies (Fig. 14).

Figure 13. Photoelectric effect.

15
Figure 14. X-ray absorption coefficient of Pb.
16
1.3. Material characterization techniques using X-rays
• The different interactions of X-ray with matter are employed in many techniques
for materials characterization, e.g.
Scattering
◘ X-ray Scattering
- X-ray Diffraction (XRD, WAXS)
 Crystal structure and related properties, phase purity and composition.
- Small Angle X-ray Scattering (SAXS)
 Particle/pore sizes, shapes, distribution, and surface-to-volume ratio.
- Grazing-Incidence X-ray Diffraction (GID)
 Surface XRD.
- Grazing-Incidence Small-Angle X-ray Scattering (GISAXS)
 Combines SAXS and GID.
- X-ray Raman Scattering (XRS)
 Coordination number, electronic structure, bonding.
 Resembles XAS.
- Resonant Inelastic X-ray Scattering (RIXS)
 Electronic structure.
- Compton Scattering / Inelastic X-ray Scattering (IXS)
 Electronic structure.
17
Absorption
◘ X-ray Absorption Spectroscopy (XAS)
- X-ray Absorption Near Edge Structure (XANES)
- Extended X-ray Absorption Fine Structure (EXAFS)
- Near Edge X-ray Absorption Fine Structure (NEXAFS)
 Coordination number, electronic structure, bonding.
 Local structure.

Photoionization & emission

◘ X-ray Fluorescence (XRF)
 Elemental analysis.
- Energy Dispersive X-ray Spectroscopy (EDS, EDX)
- Wavelength Dispersive X-ray Spectroscopy (WDS, WDX)
◘ X-ray Photoelectron Spectroscopy (XPS, ESCA)
 Elemental analysis, empirical formula, chemical state, electronic state.
 Quantitative.
 Surface.
◘ X-Ray-Excited Auger Electron Spectroscopy (XAES, XEAES)
 Elemental analysis.
◘ X-ray Emission Spectroscopy (XES)
 Electronic structure.
- Resonant Inelastic X-ray Emission Spectroscopy (RIXS)
18
- Non-resonant Inelastic X-ray Emission Spectroscopy (NXES)
 Mixed & miscellaneous
◘ X-ray Porosimetry (XRP)
 Thin film porosity.
◘ X-ray Reflectometry (XRR)
 Thickness, roughness and density of thin films.
◘ X-ray Microscopy (TXM, STXM)
 3D structure of samples.
 XANES.
◘ X-ray Free Electron Laser (XFEL) Spectroscopy
 Atomic resolution snapshots on the ultrafast timescale associated with the intrinsic
atomic motions of atoms in matter.
◘ X-ray magnetic circular dichroism
 Magnetic properties of solids, thin films and surfaces.
 Orbital and spin magnetic moments.

• In this course:
◘ X-ray powder diffraction.
◘ X-ray absorption spectroscopy.
◘ X-ray fluorescence spectroscopy.

19
2. GENERATION AND DETECTION OF X-RAYS
2.1. The X-ray spectrum
• X-rays may be produced by two means:
◘ When an electrically charged particle of sufficient kinetic energy rapidly decelerates.
 Conventionally produced by electrons hitting a metal target.
 Not all of the particles decelerate in the same way: Some lose their energy at once,
some lose it gradually.
 Continuous spectrum
(i.e. Bremsstrahlung) (Fig. 15).
◘ Photoelectric effect (see 1.2.).
 Emission of X-ray photons with energies
characteristic to the element of the
metal target.
 Characteristic spectrum (Fig. 15).

Figure 15. Schematic X-ray spectrum of Mo

as a function of applied voltage.
20
• In the continuous spectrum, the short wavelength limit (SWL) (i.e. the highest
energy) is obtained with electrons that are stopped with one impact.
◘ These electrons transfer all their energy into photon energy:
SWL = hc/eV = 12.40·103ÅV/V (2.1)

(h: Planck’s constant, c: velocity of light, e: electron mass, V: applied voltage)

◘ No dependence on the target element.
• With high enough kinetic energy for the decelerating electrons, i.e. when the
element specific (characteristic) absorption edge energies (Fig. 14, Table 1) are
reached, characteristic emission (Fig. 16, Table 2) is produced due to the core-
hole filling in the photoelectric effect:

◘ The voltage needed to generate characteristic lines can be calculated by using the
wavelength of the characteristic emission line, K (in Å). For example, for the K edge:
VK = hc/eK = 12.40·103ÅV/K (2.2)
Table 1. Electron binding energies (in eV) for elements 1 - 36.

22
Figure 16. Transitions giving rise to characteristic emission lines (left) and
emission of Mo at 35 kV (right).
23
Table 2. Photon energies (in eV) for principal emission lines for elements 3 - 36.

24
2.2. X-ray tubes
• X-ray tubes contain the following parts:
◘ Source of electrons (cathode).
◘ High accelerating voltage.
◘ Metal target (anode).
◘ Cooling device: only 1 % of the energy of the electrons is converted to X-rays, 99 % to
heat.
• Two basic types.
◘ Gas tubes: electrons are produced by the ionization of a small quantity of gas (residual
air in a partly evacuated tube), e.g. the Crookes tube (Fig. 2).
 Resemble those used by Röntgen, not used nowadays.
◘ Filament tubes: Electrons are produced by a hot filament.
 Invented by William D. Coolidge (Fig. 17) in 1913.

1940’s X-Ray Physics Documentary By

William D. Coolidge
http://www.youtube.com/watch?v=I3s5HFQ2YME

Figure 17. William D. Coolidge. 25

• Filament tube (Fig. 18) operation:
◘ When the heating circuit is activated, the current flowing through the filament heats
the wire to about 2200°C, giving some of the electrons in the metal enough energy to
move a little way from the surface of the metal.
◘ When the high voltage circuit is activated, the tungsten filament becomes extremely
negative, forcing the electrons to escape into the vacuum of the tube, where they are
attracted towards the positive anode.

Figure 18. The filament X-ray tube.

◘ Some tungsten may occasionally evaporate from the filament and deposit on the target.
 W(L) emission may be observed as an impurity.
26
• Variations of the filament tube:
◘ Rotating anode tube (Fig. 19).
 Provides continously fresh metal target area.
◘ Microfocus tube (Fig. 19).
 Focal spot diameters smaller than 0.01 mm.
◘ Pulsed/flash tube.
 Removes the limitations for maximum operating power set by cooling requirements.
◘ High voltage tube.

27
Figure 19. Rotating anode (left) and transmission microfocus X-ray tube (right).
• Typical anode materials:

28
2.3. Synchrotron radiation
• If electrons or positrons moving at relativistic velocities are deflected along a
curved trajectory by a magnetic field, synchrotron radiation is produced.
◘ Continuous spectrum from microwaves to hard X-rays (ca. 10-4 – 105 eV).
• In synchrotrons, electrons (or positrons) are injected periodically (e.g. every 8 h)
to a storage ring (Fig. 20), where they circulate.
• The deflection may be produced by using bending magnets or insertion devices
(Figs. 20 - 23).

Bending magnets
Ring radius tens
or hundreds of
meters.
Straight sections
containing insertion
devices.

Figure 20. Synchrotron storage ring. 29

• Bending magnet (Fig. 21):
◘ Needed to guide electron beam around the ring.
◘ Gives a wide horizontal fan of radiation.
◘ Smooth intensity distribution versus energy.

Figure 21. Bending magnet (left) and its spectrum (right).

30
• Insertion devices (Fig. 22):
◘ Built from multi-pole magnet linear arrays.
◘ Wigglers use high magnetic fields.
◘ Undulators use low magnetic fields and utilize interference between the photon beam
and the periodic magnetic pole structure.

Wiggler

Undulator

Figure 22. Schematic of an insertion device (left)

and typical spectra of wigglers (top right) and
undulators (bottom right). 31
Table 3. Comparison of bending magnet and insertion device properties.
Bend Magnet Wiggler Undulator
Flux Low High Very high
Beam size
Beam divergence
Tuneable output Yes Yes Tricky
Energy range < 20 keV ~100 keV < 20 keV

Figure 23. Comparison of spectral

brightness from bending magnets
and insertion devices.
32
• From the bending magnets and insertion devices the radiation is transported to
experimental stations (and finally to the sample to be studied) via beamlines
(Figs. 24 & 25).

Figure 24. Schematic of a synchrotron radiation beamline.

33
Figure 25. A typical synchrotron radiation
beamline (top) and inside of the
experiment hutch at HASYLAB (Hamburg,
Germany) X-ray absorption spectroscopy
beamline A1 (bottom).

34
• Synchrotrons give much more beam intensity than laboratory X-ray sources
(Fig. 26).

Year

Figure 26. Comparison of spectral

brightness from bending magnets,
insertion devices and X-ray tubes.
35
• Synchrotron based free-electron
lasers (FELs) give even higher beam
intensity (Fig. 27).

Figure 27. Comparison of spectral

brightness from undulators and FELs.

36
• Advantages of synchrotron radiation:
◘ Tunability.
◘ High intensity.
◘ High resolution.
• Disadvantages:
◘ Intensity may be too high for some samples.
◘ Very expensive to produce.
◘ Limited geographical (Fig. 28) and temporal accessibility.

Figure 28. Map of synchrotron sources (top)

and aerial view of the PETRA storage ring at
DESY in Hamburg, Germany (right).
37
38
https://www.youtube.com/watch?v=-X8vjtnB0pg

39
2.4. X-ray detectors
• A wide variety of X-ray detectors is available (Table 4), some counting single
photons, some providing only measurements of count rate or total flux, others
measuring the energy, position, and/or incidence time of each X-ray.

Table 4. Properties of common

X-ray detectors.

Figure 29. Comparison of the

energy resolution of
proportional, scintillator and
Si(Li) detectors for Mn(K)
radiation. 40
3. SAFETY ISSUES
• The user of a laboratory X-ray apparatus is exposed to two dangers: radiation
injury and electric shock (high-voltage).
• Both can be reduced to neglible proportions by proper design of equipment and
reasonable care on the part of the user.
Electric shock
• The anode of an X-ray apparatus is usually grounded and thus safe. However, the
cathode must be made inaccessible.
Radiation hazard
• X-rays can kill human tissue, and the biological effects include burns, radiation
sickness and genetic mutations.
• The most intense and therefore dangerous part of each instrument is the path of
the incident X-ray beam. Thus care should always be exercised to know the
expected path of the beam.
• Scattered radiation is typically of such reduced intensity that it poses a much
smaller health risk to the researcher.
• However, potientially dangerous beams may come from an instrument in very
unexpected directions. The X-ray beams may bounce off surfaces of the
41
instrument and bend around corners.
4. ELEMENTARY CRYSTALLOGRAPHY
4.1. Unit Cells
• Crystals consist of a regular three-dimensional arrangement of atoms, ions or
molecules, i.e. the structure has long-range order.
• In amorphous state, the structure is short-range only.
• The whole structure of a crystal can be described with a basic repeating unit:
the unit cell.
• The shape and size of a unit cell are defined by unit cell axes (a, b and c) and the
angles between them (α, β and γ) (Fig. 30).
• Together, these are called the unit cell parameters.

Figure 30. Unit cell parameters in

the xyz-coordinate system.
42
 Copper

Figure 31. Examples of unit cells built up of atoms (top left), ions (bottom left)
and molecules (right).

43
• There are seven possibilities for the shape of the unit cell. These are known as the
crystal systems:

44
• The unit cells are further classified as follows:
◘ Primitive (P): same atoms in the corners, but not in the unit cell center or centers of the
faces.
◘ Body-centered (I): same atoms in the corners and cell center.
◘ Face-centered (F): same atoms in corners and face centers.
◘ End-centered (C): same atoms in the corners and centers of two opposite faces.
• Combining these with the
seven crystal systems gives
the 14 Bravais lattices (Fig. 32).

Figure 32. The 14 Bravais

lattices.
45
• A further classification of the contents of unit cells can be made by the symmetry
operations that they obey. This gives the 32 crystal classes (i.e. point symmetry
groups):

46
• Finally, by combining the Bravais lattices and the crystal classes gives 230 space
groups.
◘ Every crystal has a structure that can be described by one of the space groups.
• In the unit cells, the positions of the atoms are defined by their xyz-coordinates
corresponding to their relative position in the unit cell abc axes directions.
◘ Example: if an atom has the coordinates (¼, 0.327, 0), it is located (Fig. 33):
 ¼ * a-axis length from (0, 0, 0).
 0.327 * b axis length from (0, 0, 0).
 0 * c axis length from (0, 0, 0) i.e. on the ab plane.

b (0, 0, 0) c

a b
(0, 0, 0)
a
Figure 33. Two views of a unit cell with one atom at (1/4, 0.327, 0).
• All possible atomic positions in all of the 230 space groups are collected to the
International Tables for X-ray Crystallography (Fig. 34).
47
Figure 34. An example of
the International Tables
for X-ray Crystallography:
space group F222.

48
4.2. Atomic Planes
• A unit cell can be considered to have atomic planes, lattice planes (Fig. 35).
• The planes are identified by the Miller indices hkl.
◘ The indices describe the plane’s intercepts with the unit cell axes: a/h, b/k and c/l.
◘ The numbers are reciprocals of these intercepts, i.e. h = 4 means that the plane goes
through x = ¼ and h = 0 that the plane is on the a-axis.
• Each Miller index describes a
family of lattice planes with a
constant distance between them.
• These interplanar distances are
labelled as dhkl.

Figure 35. Examples of lattice

planes and their Miller indices.
49
5. X-RAY DIFFRACTION
• Wikipedia:
“In classical physics, the diffraction phenomenon is described as the apparent bending
of waves around small obstacles and the spreading out of waves past small
openings.”
◘ X-rays diffract from electrons, i.e. the electron shells of atoms.
Diffraction occurs with all waves, including sound waves, water waves, and
electromagnetic waves such as visible light, X-rays and radio waves.
While diffraction occurs whenever propagating waves encounter obstacles, its effects
are generally most pronounced for waves where the wavelength is roughly similar
to the dimensions of the diffracting objects.”
◘ X-ray wavelengths are of the order of atomic sizes (1 Å).
“If the obstructing object provides multiple, closely-spaced openings, a complex
pattern of varying intensity can result. This is due to the superposition, or
interference, of different parts of a wave that traveled to the observer by different
paths.”
◘ This gives the X-ray diffraction pattern.
• Note: Powder diffraction works with all polycrystalline materials, not only
powders.
50
• X-ray diffraction is among the most common analysis methods:

ISI Web of Knowledge / December 2013:

Method Number of published articles in 2013
Luminescence Spectroscopy 45656
X-ray Diffraction (XRD) 39730
Mass Spectrometry (MS) 35192
Liquid Chromatography (LC) 18235
Infrared Spectroscopy 13528
Differential Scanning Calorimetry (DSC) 7992
Gas Chromatography (GC) 7873
NMR Spectroscopy 7632
Raman Spectroscopy 7473
Cyclic Voltammetry (CV) 4965
UV-vis Absorption Spectroscopy 2359
Atomic Absorption Spectrometry (AAS) 1540
Magnetometry 1475
Thin Layer Chromatography (TLC) 819
51
Circular Dichroism Spectroscopy (CD) 326
• In 1914, Max von Laue won the Nobel Prize in physics for discovering how crystals
can diffract X-rays. Thus, 2014 was named the International Year of Crystallography.

52
53
5.1. Bragg’s law
• A diffraction pattern is observed, when waves diffracting from adjacent atomic
planes are in the same phase (Fig. 36). This is because only then, constructive
interference takes place.

Figure 36. Constructive interference of waves diffracting from atomic planes.

• This constructive interference is obtained, when when Bragg’s law (Eq. 5.1) is
fullfilled:
n = 2dsin (5.1) 54
• If only two planes of atoms were diffracting, the transition from constructive to
destructive interference would be gradual as the angle is varied. However, since
many atomic planes are interfering in real materials, very sharp peaks surrounded
by mostly destructive interference result.
• For a single crystal, diffraction gives spots around the crystal (Figure 37).
• A powder (polycrystalline sample) has crystallites with all possible orientations.
Thus diffraction results in cones (Figure 37).

Single crystal

Powder / Polycrystalline

Figure 37. Diffraction from a single crystal (top) and a polycrystalline sample
55
(bottom). Both cases depict transmission only.
5.2. X–ray diffraction intensity
• The intensity of diffraction from a specific set of atomic planes is determined by
the following factors:
◘ The structure factor, which is determined by
o The symmetry of the crystal.
o The positions of the atoms in the unit cell (Fig. 38).
o The number of electrons in the atoms (the amplitude of diffraction is quadratically
dependent on this number) (Fig. 39).
◘ The amount of material.
◘ The absorption in the material.

Figure 38. Wave fronts in CsCl.

56
 • In addition to the atomic number,
diffraction intensity is also
dependent on the diffraction angle
(Fig. 39).
Atomic Number

Figure 39. X-ray scattering

factor (f) values for neutral
atoms as a function of sin/ .
Diffraction Angle 57
• Note: in addition to X-rays, also neutrons and electrons can be used to obtain
diffraction patterns.
• In neutron diffraction, there is no clear systematic dependence between atomic
number and diffraction intensity (Fig. 40).

Figure 40. Nuclear scattering lengths (b) as a function of atomic weight. 58

5.3. Powder diffraction pattern
• The powder diffraction pattern is collected along a single line of the diffraction
circles (Figure 41).
• It is plotted as diffraction intensity vs. diffraction angle 2 (Figure 41).
• The peaks are called reflections.

Figure 41. Diffraction rings converted to a powder diffraction pattern.

• As a consequence of Bragg’s law, the reflection positions give information on the

distances (d) between atomic layers.
◘ The d values can then be used to obtain the unit cell paramaters (a, b, c, α, β and γ).
59
 ◘ The general format for these calculations is as follows (V = unit cell volume):

2 b c sin  2 a c sin  2 a b sin 

2 2 2 2 2 2 2 2 2
1
2
h k l
d hkl V2 V2 V2
abc 2 (cos  cos   cos  ) a 2bc(cos  cos   cos  )
 2hk  2kl
V2 V2
ab 2 c(cos  cos   cos  ) (5.2)
 2lh
V2

◘ For the higher symmetries, this is reduced to:

Cubic: 1/d2= (h2+k2+l2)/a2 (5.3)

Tetragonal: 1/d2= [(h2+k2)/a2] + (l2/c2) (5.4)
Orthorhombic: 1/d2= (h2/a2) + (k2/b2) + (l2/c2) (5.5)
Hexagonal: 1/d2= 4/3[(h2+hk+k2)/a2] + (l2/c2) (5.6)

• The reflection intensities give information about the atoms on the diffracting
planes (see Chapter 5.2). A more detailed descrition is given in equation 6.1 of
chapter 6.2.
60
• The powder diffraction pattern also contains a lot of more information:

Powder diffraction pattern

Reflections Background

Sample Scattering from

Position Intensity Profile sample holder,
(FWHM, peakshape) air etc.

Amorphous fraction
Lattice parameters Instrument Sample
Space group (local order/ disorder)
broadening broadening
Crystal structure:
Atomic positions
Qualitative Temperature factor
phase analysis Occupancy Real structure:
(Disorder)
Stress,
Strain
Crystallite size
Domain size,...
Quantitative
phase analysis

• Using this information will be discussed later on during this course. 61

Versatility of X-ray powder diffraction: CSI Miami

62
5.4. Laboratory X-ray powder diffractometers

Figure 42. A typical powder

diffractometer setup. 63
5.4.1. DIFFRACTION GEOMETRIES
Bragg-Brentano
• Brentano (1917) and Bragg (1921).
• The divergent and diffracted beams are focused at a fixed radius (the goniometer
circle) from the sample position (Figure 43).
• The anode can be fixed and the sample and detector can be rotated by θ and 2θ,
respectively.
• The alternative is to fix the sample
(usually in the horizontal position) and to
move both the source and the detector
by -θ and θ, respectively.
• True focusing only occurs when the
sample plate has a curved surface, but
the flat-plate approximation works well
since the footprint of the beam on the
sample plate is considerably smaller than
the radius of the focusing circle.
Goniometer circle
64
Figure 43. Bragg-Brentano geometry.
• Bragg-Brentano parafocusing:
◘ The anode (S), sample (O) and detector (D) lie on the focusing circle (Figure 44).
◘ The goniometer circle has a fixed radius, R, but the radius of the focusing circle, r
varies from ∞ to R/2 as the 2θ angle varies from 0° to 180°.
◘ Due to the geometric property of circles, X-rays produced from S and diffracted from
any point on the focusing circle by angle 2θ
(e.g. from O) all arrive at exactly the same point D, Focusing circle
i.e. the relationship φ1 = φ2 = 180°- 2θ is exact.
• Only reflection (no transmission)
measured.
• Flat samples.
• Pros:
◘ Very good resolution possible.
◘ Easy to align.
• Cons:
◘ Conventionally slow data collection
(one point at a time).

Figure 44. Bragg-Brentano parafocusing. 65

Debye-Scherrer
• Debye & Scherrer (1916) and Hull (1916).
• Used to be one of the most common instruments (Figure 45) for obtaining
powder diffraction data from a sample in transmission geometry :
◘ Powder diffraction patterns were collected
on film wrapped around the inside of the camera.
◘ The sample was positioned at the center of
the camera.
• Pros:
◘ Very high angles reached.
◘ Can reveal texture effects.
• Cons:
◘ Non-focusing (Figure 46) = low resolution.
• The resolution may be considerably
enhanced by the introduction of a
focusing narrow-band pass
monochromator (Figure 47).
• Transmission and reflection.
• Capillary or thin plate samples. 66
Figure 45. Debye-Scherrer camera.
 Figure 46. The detector is never on the
focusing circle in the Debye-Scherrer
geometry.

Figure 47. Better resolution: The

divergent beam from the X-ray anode
is focussed using a curved silicon
monochromator, not onto the
cylindrical sample, but beyond onto
the 2θ measuring circle of the
67
detector.
Seemann-Bohlin
• Seemann (1919) and Bohlin (1920).
• The anode (S), sample (AB) and detector (MN) lie on the focusing circle
(Figure 48).
• Pros:
◘ High resolution.
◘ Fast measurement (the whole pattern simultaneously).
• Cons:
◘ Very difficult to align.
◘ Low diffraction angles unreachable.

Figure 48. Seemann-Bohlin geometry. 68

Guinier
• Guinier (1939) and de Wolf (1948).
• Modification of the Seemann-Bohlin: primary beam monochromator added
(Figure 49).
• Transmission (for low angles) and reflection (high angles) modes possible.
• Pros:
◘ High resolution.
◘ Fast measurement (the whole pattern simultaneously).
• Cons:
◘ Very difficult to align.
◘ Low diffraction angles unreachable in reflection mode.
◘ High diffraction angles unreachable in
transmission mode.
◘ Low-angle reflections may present
severe asymmetry.

Figure 49. Guinier geometry

69
in transmission.
Parallel beam
• Primary beam optics: channel-cut Si monochromator or parabolic graded
multilayers (X-ray mirror) (Figure 50).
• Secondary beam optics: crystal analyzers or long Soller slits (Figure 50).
• The geometry produces many focus points on the sample, which enables the use
of non-flat samples.

Figure 50. Parallel beam geometry with an X-ray mirror and a soller slit.
70
5.4.2. X-RAY TUBE
• Usually, only the K-lines are useful in X-ray diffraction (Table 7).
◘ Highest intensity.
◘ Wavelength in the Å range.

Table 7. Some commonly used X-ray K wavelengths (Å).

Element Kα(av.) Kα1 Kα2 Kβ1
Cr 2.29100 2.28970 2.29361 2.08487
Fe 1.93736 1.93604 1.93998 1.75661
Co 1.79026 1.78897 1.79285 1.62079
Cu 1.54184 1.54056 1.54439 1.39222
Mo 0.71073 0.70930 0.71359 0.63229

• Choice of radiation (Table 8). Figure 51. Cu X-ray tube.

◘ Short wavelength: more reflections – less resolution.
◘ Long wavelength: less reflections – more resolution.
◘ The X-ray energy should not be just above the absorption edge energies of the studied
elements (Figure 52): fluorescence will result.
◘ The mass attenuation coefficient (Table 9) is also important.
71
Table 8. Specific applications for commonly used X-ray K wavelengths.

Anode Material Application

Cu Suitable for most diffraction examinations. Most widely used.
Mo Preferably used for examinations on steels and metal alloys with
elements in the range Ti to approx. Zn. Used to maximize the number
of reflections.
Co Often used with ferrous samples, the Fe fluorescence radiation would
cause interference and cannot be eliminated by other measures.
Fe Examination of ferrous samples. Also for use with minerals where Co
and Cr tubes cannot be used.
Cr Used for complex organic substances and also stress measurements
on steels. Used to minimize reflection overlap and maximize resolution.
W Used where an intensive white spectrum is of more interest than the
characteristic.

72
 Metallic copper
Cu K radiation
From Tables 1 & 2:

Cu K: 8040 eV

Cu K edge: 8979 eV
Metallic iron
Cu K radiation Fe K edge: 7112 eV

Figure 52. Diffraction patterns of Cu (top) and Fe (bottom) measured using

CuK radiation. The diffraction signal for Fe is buried under its X-ray emission.

73
Table 9. Mass attenuation coefficients (in cm2g-1) for CuK for selected elements.

3rd period
Na Mg Al Si P S Cl Ar
30 39 49 61 74 89 106 123
Group IIA
Be Mg Ca Sr Ba Ra
2 39 162 125 330 226
Lanthanides
Ce Pr Nd Pm Sm Eu Gd Tb Dy
352 363 374 386 397 425 439 273 286
Ho Er Tm Yb Lu
128 134 140 146 153
5.4.3. FILTERS AND MONOCHROMATORS
• Each of the K-lines (1, 2, ) will produce a diffraction pattern complicating the
interpretation.
• The undesirable lines can be removed by using filters or monochromators for the
primary beam.
• A secondary monochromator can be used to remove anything but coherent
scattering from the radiation entering the detector.

Filters No Filter Ni Filter

•  lines can be suppressed by
passing the beam through an
absorber (filter) whose
absorption edge lies just above
the parasitic wavelength
(Figure 53).

Figure 53. Suppressing the Cu K

line by using a Ni filter. 75
• The filtration is never perfect (Table 9): a thicker the filter gives a better
suppression of the Kβ component but also results in weaker Kα.

Table 9. Suppression characteristics of commonly used K filters.

Anode Beta filter Thickness Kβ suppression Kα suppression
material material / m /% /%
Mo Zr 75 97 54
Cu Ni 20 99 58
Co Fe 16 99 51
Cr V 13 98 45

Figure 54. 20 m V, Fe, and Ni filters. 76

Monochromators
• X-ray beam monochromatization can be achieved by diffraction from flat single
crystals: Si, Ge, LiCl, and graphite (Figure 55).
◘ Kβ component easy to remove.
◘ Kα2 cannot be removed.
• The wavelength separation is based on different diffraction angles for different
wavelengths.

Figure 55. Flat single crystal monochromator. 77

• The focusing geometry with a curved single crystal (Figure 56) can be used to
obtain truly monochromatic radiation (also Kα2 removed ).
• Also known as Johansson-Guinier type.

Figure 56. Curved single crystal monochromator: schematic (left) and Huber 611.

78
• Although the Kα2 can be
removed, the
monochromatization also
leads to the loss of intensity
(Figure 57).

Figure 57. Y2O3 measured with Cu K

radiation: no monochromator (top)
and with monochromator (left).

79
5.4.4. SLITS
Divergence slits
• Divergence slits are placed in the incident beam path to control the equatorial
divergence of the incident beam, and thus, the amount (length) of the sample
that is irradiated by the incident X-ray beam (Figure 58).
• Slit size must be chosen so as not to irradiate anything else but the sample.

Figure 58. Effect of divergence slit

size on irradiated sample area. 80
Receiving slits
• Receiving slits are placed at the detector
entrance.
• Size should be as small as possible to
improve the resolution, but very small slit
sizes reduce diffracted beam intensity.

Figure 59. Slits with different opening sizes.

Fixed or variable slits

• In Bragg-Brentano geometry, the slit sizes are usually kept fixed: irradiated area
decreases with increasing angle (Figure 60).
• Using variable slits the irradiated area can be maintained constant (Figure 60).

Figure 60. Fixed (left) vs. variable (right) slits. 81

• Fixed slits will give different intensities than variable slits (Figure 61), which
must be taken into account if reflection intensities are used for analysis.

2.0 Y2O3
CuK
Normalized Intensity

Fixed Slits
1.5

1.0
Variable Slits

0.5

0.0

20 30 40 50 60
o
2 /

Figure 61. Calculated diffraction pattern for Y2O3 with fixed (red curve) and variable
(blue) slits.
82
Soller slits
• Soller slits (Figure 62) are used to limit the axial (vertical / out-of-plane)
divergence of the incident & diffracted X-ray beams.
• Using soller slits improves reflection shape and resolution, especially at low
scattering angles.

Figure 62. Soller slits.

83
5.4.5. SAMPLE PREPARATION
• Accurate reflection positions require that the sample surface is flat and on the
focusing circle.
• Accurate intensities require that the crystallites are randomly oriented and that
there are enough crystallites (grains) for good statistics (Table 10, Figure 63).

Table 10. The effect of crystallite size on crystallite (particle)

statistics and reproducibility of intensity.

84
Figure 63. 2D diffraction images of a sample with large (left) and smaller (right)
crystallites.

85
Reflection geometry
• The sample holders for the Bragg-Brentano geometry are usually metal (Al, Cu) or
plastic plates with a circular or rectangular cavity for the sample (Figure 64).
◘ The flat surface is obtained with sufficient (but not too strong to destroy the sample)
grinding of the sample.
◘ Evening the powder surface to be on the same level as the sample holder surface will
place the sample on the focusing circle.
◘ The surface should not be pressed too heavily, since that may cause the crystallites to
orientate preferentially, i.e. to lose random orientation.

Figure 64. A typical Bragg-Brentano

sample holder (top) and sample
preparation (bottom).

86
• If the sample surface is not on the focusing circle, reflections will be displaced
(Figure 65).
• Surface roughness (too large crystallites/particles) and sample transparency (too
loose packing) will also displace reflections.

Figure 65. Effect of wrong sample surface level on reflection angle. 87

Transmission geometry
• The sample may be inserted to a glass capillary or fixed on a thin polymer film
placed on a frame (Figure 66).
• Surface flatness is equally important as for the reflection geometry.
• Sample thickness effects reflection widths (Figure 67).

Figure 66. Sample holders for Figure 67. Effect of sample thickness in
transmission geometry. transmission geometry. 88
5.4.6. DETECTORS
• The regular detector for a Bragg-Brentano diffractometer is the scintillator
counter (Figure 68).
◘ Collects point-like data (one point at a time).
◘ Rather slow measurements.
◘ Proportional counters and solid state detectors also available.

Figure 68. Bruker Bragg-Brentano diffractometer with a scintillator counter (left)

and Bruker Sol-X energy dispersive Si(Li) solid state detector (right). 89
• The data collection can be made faster with a 1D detector that can collect data
within a range of 2 – 15 degrees (Figure 69).
• Different types:
◘ Gasflow proportional.
◘ Silicon strip.
◘ Microgap technology.

Figure 69. Operation of a 1D detector (left) and Bruker Våntec-1 1D detector

based on microgap technology (right).
• In Debye-Scherrer, Seemann-Bohlin and Guinier geometries, position sensitive
detectors (Figure 70) are used.
◘ Storage phosphor (optically stimulated luminescence material).
◘ Gas proportional counter.
◘ Whole diffraction pattern collected simultaneously  fast measurements.

Figure 70. INEL CPS120 gas proportional counter position sensitive detector in
Debye-Scherrer geometry. 91
• Area detectors (2D detectors) offer the possibility for even faster measurements.
• Different possibilities for type:
◘ Gas filled multiwire.
◘ CCD.
◘ Microgap.
• Diffraction rings are detected (Figure 71):
◘ Gives the possibility to average out texture effects (ring spottiness).
◘ Limited angle range (typically < 35o).

Figure 71. Bruker Våntec2000 2D detector and a diffraction pattern measured with it.
92
5.5. Sample-based errors in powder diffraction
• In ideal conditions, the powder diffraction pattern would only show features
originating from the composition and structure of the material studied, as
demonstrated e.g. in
http://www.doitpoms.ac.uk/tlplib/xray-diffraction/flash/Graph%20Sim.swf
• The diffraction pattern, however, also contains effects due to the diffractometer
(Figure 73) as well as the sample conditions.
• The sample-based errors may significantly affect
the intensities, shapes and angles of reflections and
subsequently the analysis results derived from these.

Intensity (a.u.)
Figure 73. Diffractometer effects:
asymmetry due to axial divergence
(left) and reflection shape and width
variation between different
diffractometers (right).
47.0 47.2 47.4 47.6 47.8
93
2 (deg.)
Flat sample error
• For diffraction to be geometrically correct, the sample should fulfill the impossible
condition of being variably curved to always lie on focusing circle.
• The use of flat samples induces the flat sample error (Figure 74).
◘ The sample is tangent to the focusing circle (rf) .
◘ The extreme edges of the sample lie on another focusing circle (rf ‘) which results in
the overall diffracted intensity being skewed to a lower value of 2.
◘ The decrease in 2 takes the form of an asymmetric peak broadening towards lower
angles.

Figure 74. The flat specimen error. 94

• The effect is most pronounced at lower angles where the beam hits more of the
sample.
• Reducing the width of sample that is exposed to the incident beam can control
the magnitude of the flat sample error. This is usually accomplished by
narrowing the beam by the use of smaller divergence slits.
• The flat sample error is expressed as follows (equation 5.7):
2 = - (2cot) / 343.8 (5.7)
: angular aperture of the divergence slit (degrees), : diffraction angle (radians)
• In general, for most diffraction work in which low-angle (under ca. 8o in 2) data
is not required, a divergence slit angle of 0.50 provides a reasonable compromise
between intensity and minimizing flat sample error.

95
Changes in sample composition
• Excessive grinding, principally when highly percussive in nature, can induce
changes in the sample, e.g.
◘ Amorphism.
◘ Strain.
◘ Decomposition due to local heating.
◘ Loss of volatile components.
• Non-automatic and gentle grinding should thus always be employed when
preparing the sample for powder diffraction.
• Some materials react and change composition in the X-ray beam. These
effects can be significant in the analysis of organic compounds, but are not
generally an issue for inorganic materials.
• Some materials may react in ambient conditions and decompose even without
the help of the X-ray beam.
• The effect of the instability/reactivity of the sample can be taken into account and
observed by making many short measurements instead of one long.

96
Sample displacement
• Sample displacement can be a significant source of errors in diffraction angles
measured with a diffractometer. The geometry of diffraction requires that the
sample lies on the focusing circle and be at the center of the diffractometer
circle (see Figure 65 and chapter 5.4.5). Anything that causes the sample to
deviate from this geometry will cause angular errors in the resulting diffraction
data.
• It is difficult to prepare a specimen which is precisely flat and uniform. For this
reason specimen displacement is a significant cause of angular diffraction errors.
• This error is most pronounced at lower 2 angles, since it follows a cosine
function (see Figure 65).
• It will cause asymmetric broadening of the profile toward low 2 values, and can
produce about 0.01 of angular error for each 15 m of displacement.

97
Sample transparency
• Sample transparency errors are related to the effective depth of penetration of
the X-ray beam (Figure 75). The errors result primarily in
changes in the observed
diffraction angles.

Figure 75. Effect of sample

transparency on the observed
reflection angle.

• The transparency error for a thick specimen (thicker than penetration depth of
the beam) may be defined mathematically as follows (equation 5.8):
2 = sin2 / 2R (5.8)
: diffraction angle (radians), R: radius of diffractometer circle, : linear absorption
coefficient of the sample for the radiation used
• The “working thickness”, i.e. the average depth from which diffraction is
generated, of the sample (t1/2) is defined by beam penetration depth
(equation 5.9):
t1/2 = 1 /  (5.9)
• The mass absorption coefficient * (i.e. /) is tabulated for different elements
and is dependent on the X-ray wavelength.
• For SiO2 and CuK, = 97.6/cm. t1/2 thus is ca. 100 m.
• For high-density, high * materials (e.g. metals, alloys), t1/2 will be on the order of
10 m.
• For low-density organics, t1/2 will be on the order of 1000 m, and a thick sample
will induce very significant displacement errors. For this reason, organics are
usually mounted as thin films on a zero-background plate.
• For powders which are not tightly packed, because  will be a function of both
the sample density and the air in the pore spaces, the actual displacement error
will be greater than that estimated for a powder having the density of the solid
material.
Sample thickness
• As evident from the sample transparency (and displacement) effects, thin samples
will yield the most accurate reflection positions.
• Thick samples will yield the most accurate reflection intensities as the statistics
are superior to those in thin samples.

99
Sample inhomogeneity
• Many multi-phase samples are inhomogeneous in character (Figure 76).
• If the phases in the sample have different * values, their diffraction intensities
will not accurately reflect their proportions in the sample.
• If the phases are also inhomogeneously distributed in the sample, each phase will
show different transparency effects.

Figure 76. Effect inhomogeneity in the penetration depth of the X-rays in a sample.
100
Preferred orientation
• If the crystallites in the sample are not randomly orientated, the intensities of
some reflections will be exaggerated while others lose intensity.
• The effects of preferred orientation can be made less severe e.g. by spinning or
shaking the sample.
• Preferred orientation will be discussed in more detail later during the course.
Crystallite statistics
• Since the sample must contain randomly orientated crystallites representing all
possible orientations, there must be enough crystallites in the sample to fulfill
this condition (i.e. good statistics are required).
• This was addressed already earlier (see Table 10 and Figure 63).
• Factors affecting statistics:
◘ Concentration: multi-phase samples have less crystallites of a given phase in a unit area
thus increasing error.
◘ Reflection multiplicity: Multiplicity in higher symmetry crystal structures results in more
diffraction per unit cell, improving statistics.
◘ Specimen thickness: improves diffraction volume, limited by penetration depth.
◘ Crystallite size: uniform crystallites of smaller size than particle size can greatly improve
statistics as long as the crystallite orientation is random. Extremely small size, however,
will result in reflection broadening (this will be discussed later during the course).
6. APPLICATIONS OF X-RAY POWDER DIFFRACTION
6.1. Qualitative phase analysis (Phase identification)
• The powder diffraction pattern is characteristic for each crystalline substance: the
crystal structure is represented by the position and intensity of the reflections.
• Substances with similar chemical formula but different structure differ in their
diffraction patterns, as well (Figure 77).

3000 YOCl
 = 1.5406 Å
Intensity / Arb. Units

2000
Hexagonal

1000
Tetragonal

30 40 50 60
o
2 /

Figure 77. X-ray powder diffraction patterns of hexagonal and tetragonal YOCl (yttrium
oxychloride).
102
• The phase identification is carried out by comparing the measured pattern with
those in a database (own or commercial).
• The most comprehensive (commercial) database is the ”Powder Diffraction File”
from International Centre for Diffraction Data (ICDD):

103
• If the identification is based on using a database, a program that can do a
”Search-Match” from e.g. the ICDD Powder Diffraction File must be used.
• Such programs are for example:
◘ Match! (Crystal Impact).
◘ Highscore (PANalytical).
◘ SIeve (ICDD).
• Each program has its own search algorithms and thus they may give somewhat
different results.
• The general way of how the programs are operated is as follows:
1) Import the experimental diffraction pattern.
2) Find where the reflections are either manually or with automation.
3) Do a search-match of the database either with or without filters:
o Inorganic / Organic / Mineral
o Possible elemental contents
o Number of elements
o Chemical formula, etc.
4) Accept and/or decline suggested phases.
5) Repeat until all reflections have been identified.
104
Example
• Product of a reaction between NaF(aq), YCl3(aq), EuCl3(aq), MgCl2(aq) and
TiCl3(aq).
• Aim: NaYF4:Eu,Mg,Ti

3+ 2+
9000 NaYF4:Eu ,Mg ,Ti
xEu:0.014 , xMg:0.043 , xTi:0.015
8000
annealed at 500 C for 5 h
7000
Intensity / Counts

white
6000

5000

4000

3000

2000

1000

20 30 40 50 60 70 80 90

2/ Degrees
105
6.2. Quantitative phase analysis
• After qualitative analysis has provided information on which phases are present
in the sample, quantitative phase analysis (QPA) will give information on the
relative amounts of the phases in the mixture.
• These are the most important industrial applications of X-ray powder diffraction,
often taking place in quality control.
• Quantitative analysis is based on the determination and comparison of diffraction
intensities due to the different phases in the mixture.
• Due to the many possible factors resulting in erroneous diffraction intensities,
good accuracy is not always to be expected from quantitative phase analysis.
• The analysis methods may use single peak or whole pattern methods.
◘ Single peak methods rely on the measurement of a peak or some peaks for each phase
of interest and assumes that the intensity of the peaks are representative of their
amount in the mixture.
◘ Whole pattern methods rely on the comparison of wide range of diffraction data with a
calculated pattern formed from the summation of individual phases that have been
either (i) measured from pure phase samples or (ii) calculated from crystal structure.

106
• The intensity, I, of a reflection (hkl) for a phase  in a mixture can be calculated
from (equation 6.1):

(6.1)

I0: incident beam intensity Mhkl: multiplicity of reflection hkl of phase 

: X-ray wavelength V: volume of the unit cell of phase 
e: electron charge F(hkl): structure factor for reflection hkl of
r: distance from specimen to detector phase  (i.e. the vector sum of scattering
me: electron mass intensities of all atoms contributing to that
C: velocity of light reflection).
Constant for the setup used 2: diffraction angle for reflection hkl
2m: diffraction angle of the monochromator
s = linear absorption coefficient of the
X: Weight of phase  in the mixture specimen
: Density of phase 
m*: Mass absorption coefficient of the Constant for reflection hkl for phase α
mixture

107
• Due to the constancies, Eq. 6.1. can be expressed in a simpler form (Equation
6.2):
I = C/(Bm*) (6.2)
B: Calibration constant, C: Concentration (weight fraction) of phase 
• The concentration is thus (Equation 6.3):
C = IBm* (6.3)
• Assuming that intensities can be measured accurately, the big problem relates to
the mass absorption coefficient for the sample, since in most experiments it is a
function of the amounts of the constituent phases (Equation 6.4):
m* =  i*Ci (6.4)
• All peak intensity-related methods for quantitative analysis involve circumventing
this problem to make this equation solvable.
• The mass absorption coefficient of a sample can be determined experimentally
by e.g. transmission measurements in a diffraction setup or X-ray fluorescence
spectrometer.

108
6.2.1. ABSOLUTE CALIBRATION-BASED METHODS
General model
• If the m* for the sample is known, Equation 6.5 can be used.
C = IBm* (6.5)
• Calibration by a set of standard samples is needed to determine the value of B.
• The standards can be pure materials or mixtures.
• Calibration is carried out by fitting a straight line in a C vs. Im* plot (Figure 78),
which yields B as the slope of the line.

Figure 78. General model

calibration curve.
109
• Because the general model is absolute, the calculated concentration may be
erroneus due to intensity variations (due to ambient conditions and the X-ray
tube).
• These variations must be accounted for by using a drift factor Fd. This is obtained
by comparing the intensity of a selected reflection measured for a standard
material when generating the calibration curve (Im0) and when doing the analysis
(Im) (Equation 6.6):
Fd = Im0/Im (6.6)
Im0: initial reference intensity Im: current reference intensity
• The corrected intensity is then obtained as (Equation 6.7):
Icorrected = FdIobserved (6.7)

110
Straight line model
• If the m* can be assumed to be the same for each sample, it can be included in
the calibration constant (Equation 6.8):
C = IB (6.8)
• This model can thus be applied, if the phases in a mixture all have the same i or
if the mixtures studied differ very little in composition (i.e. the m* can be
assumed to remain unchanged from one sample to the next).
• As for the general model, calibration is carried out by fitting a straight line in a C
vs. I plot (Figure 79), which yields B as the slope of the line.
• As for the general model, drift correction should be used.

Figure 79. Straight line

model calibration curve.
111
6.2.2. RELATIVE CALIBRATION-BASED METHODS
Internal standard model
• If both the m* and the sum of the concentrations of the analyzed phases are
unknown, a known amount of an other phase (the internal standard, S) can be
added. Then (equations 6.9 & 6.10):
Ci / CS = (IiBim*) / (ISBSm*) (6.9)
Ci = CS [(IiBi) / (ISBS)] (6.10)
• Then:
Ctotal = 1 + CS = 100 % + CS (6.11)

• The calibration is carried out by fitting a

straight line in a Ci / CS vs. Ii / IS plot,
which yields Bi as the slope of the line
(Figure 80).
• BS is defined to be equal to 1.

Figure 80. Internal standard model calibration curve.

• Intensity drift correction needs not to be used as only relative intensities are
needed.
• The calibration curve must go through origin since relative values are used.
• Requirements for the internal standard material:
◘ Chemically stable.
◘ Sufficiently small crystallite size.
◘ No chemical reaction with the sample’s components.
◘ Its reflections must not overlap with those of the sample.
◘ Its diffraction pattern must have a high-intensity reflection to obtain good counting
statistics in a reasonable time. The precision of the internal standard intensity has a
direct impact on the accuracy of the concentrations.
◘ Only a few reflections (= no overlap and good intensity).
◘ Mass absorption coefficient comparable to that of the sample.
◘ No fluorescence.
• The concentration(s) of the internal standard must be chosen so that the
expected analyte concentrations are covered in the calibration curve.

113
6.2.3. SEMI-QUANTITATIVE METHODS
Reference intensity ratio (RIR) method
• If calibration standard samples and pure phases are not available, the reference
intensity ratio (RIR) may be used to establish the calibration constants.
• The RIR is the ratio of the net reflection intensities of the strongest reflections of
the phase to be analyzed and the reference corundum (-Al2O3, marked as C) in a
1-to-1 mixture, then (equations 6.12 & 6.13):
Ci / CC = (IiBi) / (ICBC) = 1 (6.12)
Bi = IC / Ii = 1 / RIRi (6.13)
• BC is defined to be equal to 1.
• The RIRs can also be used as an alternative to relative calibration constants if
several samples need to be analyzed.
• The Powder Diffraction File (PDF) of the International Centre for Diffraction Data
(ICDD) contains RIRs (I/ICor values) for thousands of materials (Figure 81).
• These values are usually based on peak heights, not integrated intensities.
• The tabulated RIR value gives one point on the calibration curve, while the other
point is the origin (similarly to Figure 80).
• The measured RIR value from a 1-to-1 mixture of corundum and the sample 114
will give the concentration of the analyst in the sample.
Figure 81. The Powder Diffraction File entry for Bi2O3 with the RIR value highlighted.

115
• The RIR values are usually given for fixed slits. If variable slits are used, the
following correction must be applied (equation 6.14):
Bi,variable slits = (di / dC) * (1 / RIRi) (6.14)
di and dC: d values for the strongest reflections of of phase i and corundum, respectively.
• If the strongest reflection of phase i (with relative intensity of 100 %) cannot be
measured due to e.g. overlap, another reflection with a smaller intensity (I’i) must
be used instead.
• The RIR can then be adapted as follows (equations 6.15 & 6.16):
B’i = (100 / I’i) * (1 / RIRi) (6.15)
B’i,variable slits = (100 / I’i) * (d’i / dC) * (1 / RIRi) (6.16)
• The RIR method should not be used if the reference pattern has been measured
using a different wavelength than the sample pattern.

116
Doping with a foreign substance
• If the RIR of the analytical phase i is known, its concentration can be determined
by doping the original sample.
• This is done by adding a known amount of the foreign substance x, whose RIR is
known.
• After measuring the doped sample, the concentration Ci in the original sample is
calculated as follows (equation 6.17):
Ci = Ax * (Ii / Ix) * (RIRx / RIRi) (6.17)
Ax: known amount of dopant x added to the sample relative to the original weight
Ii and Ix: intensities of phases i and x in the doped sample
RIRi and RIRx: reference intensity values of phases i and x in the doped sample
• If variable slits are used, correction must be applied (as in equation 6.14).
• Intensity drift correction needs not to be used as only relative intensities are
needed.

117
6.2.4. STANDARD ADDITION METHODS
• Sometimes only one sample of a type needs to be analyzed for a single
compound. Then, it is not practical to perform a complete calibration procedure.
• With standard addition analysis, the compound to be analyzed must be available
also in pure form.

Reference to a second crystalline compound

• If the sample contains at least one other crystalline compound (r = reference
compound) as well as the compound to be analyzed (i), the intensity of both
phases can be measured and their ratio calculated (to eliminate the effect of
m*).
• First, the original sample (o) is measured. Next, the sample is enriched with an
added weight fraction A of phase i and measured again (e).
• Then, by using equation 6.5, for the original sample o (equation 6.18):
Ci / Cr = (BiIi,o) / (BrIr,o) = (Bi / Br) * Ro (6.18)

118
• For the enriched (original + Ai) sample e (equations 6.19 – 6.21):
(Ci + Ai) / (1 + Ai) = Ii,eBim,e* (6.19)
(Cr + Ai) / (1 + Ai) = Ir,eBrm,e* (6.20)
(Ci + Ai) / Cr = (BiIi,e) / (BrIr,e) = (Bi / Br) * Re (6.21)
• Combining 4.21 and 4.24 gives (equations 6.22 – 6.24):
Ci / (Ci + Ai) = Ro / Re (6.22)
Ci * (1 – [Ro / Re]) = Ai * (Ro / Re) (6.23)
Ci = (Ai * Ro) / (Re - Ro) (6.24)
• Next Ra is defined as (Re-Ro)/Ro giving (equation 6.25):
Ci = Ai / Ra (6.25)
• If more enriched samples are prepared, Ci can be obtained from the slope of a
Ai vs. Ra plot (Figure 82).
• Intensity drift correction needs not to be used as only relative intensities are
needed.

119
 Figure 82. Multiple
addition calibration curve.

Measurement of analytical compound only

• If there is no second compound to be used as a reference, e.g. the other phases
are amorphous, quantitative information can be obtained by measuring three
samples: pure compound (p), original sample (o) and enriched sample (e).
• The concentration of i in the original sample can be obtained from (equation
6.26):
Ci = Ai * (Io / Ip) * [(Ip - Ie) / (Ie - Io)] (6.26)
• Intensity drift correction should be used, since absolute intensities are required.
6.2.5. RIETVELD METHOD
• In the Rietveld method, the whole diffraction pattern is calculated based on a
structural model taking into account instrumental effects. The calculated pattern
is then fit to the experimental one by adjusting the values of both structural and
instrumental parameters.
◘ In a Rietveld refinement, the calculated intensities are compared to the measured ones
and parameter values are changed until reasonable fit is obtained by least squares
minimization of function.
• This method will be discussed in more detail later during this course.
• Whole pattern methods are potentially more accurate than conventional single
line methods because
◘ All of the reflections in the pattern contribute to the analysis, regardless of the degree
of overlap.
◘ Some sample effects, such as preferred orientation (to be discussed later), are
minimized (or at least can be taken into account) when using all reflections.

121
• The weight of a phase in a mixture is proportional to the product of the scale
factor, as derived in a multiphase Rietveld analysis of the powder diffraction
pattern, with the mass and volume of the unit cell (equation 6.27):
C = S(ZMV) /  Si(ZMV)i (6.27)
S: Rietveld scale factor, Z: number of formula units per unit cell, M: mass of the formula
unit, V: unit cell volume.

6.2.6. DIRECT COMPARISON OF INTENSITIES

• Sometimes the concentrations of the phases in a mixture can be approximated by
comparing the intensities of their most intense reflections.
• As shown above, the intensity is affected by other factors than the concentration,
as well. Thus this method is highly likely to give incorrect results.

122
6.2.7. EXAMPLE: NaCl-KBr MIXTURE
• A mixture that has been identified to contain NaCl and KBr gives the following
diffraction pattern (Figure 83):
• Measurements for samples with known NaCl/KBr ratios were carried out for
calibration (Figure 84).

1200 NaCl
NaCl/KBr Mixture
KBr NaCl
Intensity / Arb. Units

1000
Height: 752
Area: 114

800 KBr
Height: 450
Area: 81
600

20 30 40 50
o
2 /

Figure 83. X-ray powder diffraction pattern of a NaCl/KBr mixture. 123

 2500

NaCl
Intensity / Arb. Units

2000
m(NaCl):m(KBr)
2:1

1500 1:1

1:2
1000
KBr

500

20 30 40 50
o
2 /
Figure 84. X-ray powder diffraction patterns of pure NaCl and KBr as well as their mixtures
with known compositions.
124
• Straight line methods were used to analyze the NaCl/KBr ratios (Figure 85).

700
Straight line model with heights Straight line model with areas
600
NaCl NaCl
KBr KBr
500 100
Height / Arb. Units

Area / Arb. Units

400

300
50
200

100

0 0

0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
x(NaCl) x(NaCl)

Straight line model with areas

30000
Mass absorption coefficient
I(area) x m* / Arb. Units

accounted for
20000

NaCl
KBr
10000

0
0.0 0.2 0.4 0.6 0.8 1.0
x(NaCl)

125
Figure 85. Calibration curves for NaCl/KBr mixtures.
• Results (Table 11):
◘ None of the calibration curves were truly linear.
◘ Accounting for the mass absorption coefficient of the mixture improved linearity, but this
method can not be used unless the coefficient is known also for the analyzed mixture.
◘ Absolute intensities (straight line methods) could not be used for the analysis. This is
because the measurements were made in transmission geometry, where the sample
thickness can not be kept constant from one measurement to another.
◘ The relative method (simple direct comparison) yielded very good
results.

Table 11. Quantitative phase analysis results for the

”unknown” NaCl/KBr mixture using different methods.
Method x(NaCl)
Direct comparison of peak heights 0.63
Direct comparison of peak areas 0.58
Straight line model with heights 0.35
Straight line model with areas 0.52
Correct answer 0.63

126
 6.3. Determination of crystallite size (domain size)
• X-ray powder diffraction can be used for the approximation of the size of the
diffracting domains present in the sample.
• In powder materials, the diffracting domain is the crystallite. The crystallite can
be a single crystal or a part of a particle or cluster of particles (Figure 86).
• In polycrystalline materials that are not powder-like (e.g. metal or ceramic
pieces), the diffracting domain is usually called a grain.
◘ The grain size may be of great importance in these materials, since strength and
hardness increase with decreasing grain size.
• The crystallites and grains may aggregate to form
larger particles.
(cluster) • A direct observation and quantification of the
crystallite/grain sizes in a sample can be
obtained only by electron microscopies
(domain/ (SEM/TEM) or AFM.
crystallite)
• Moreover, microscopies give information on
(particle) the degree of aggregation.

Figure 86. The definition and hiearchy of crystallites,

Crystallite ≤ Particle ≤ Cluster 127
particles and particle clusters.
• In general, the diffraction linewidth is inversely proportional to the diffracting
domain size (Figure 87).
• The (integrated) intensity of a reflection is not influenced by the domain size.
• As discussed earlier, the linewidths are influenced also by sample conditions as
well as true sample properties such as structural strains (to be discussed later).
• The analysis of sample properties from the linewidths and/or shapes is called line
profile analysis (LPA).
400 5000

350 N=20 4500 N=99

4000 N=20
300 N=10 3500
250 N=10
N=5 3000
N=5
200 N=2 2500
2000
N=2
150
1500
100
1000
50 500
0 0
2.4 2.9 3.4 2.4 2.9 3.4

Figure 87. The effect of the number of unit cells (N) taking part in the diffraction
process on the observed width of a reflection.
• The sizes can be determined either with single peak (Scherrer) or multi-peak
128
methods. The latter methods will be discussed later.
6.3.1. REASON FOR INVERSE PROPORTIONALITY
• Consider a set of p crystallographic planes with a distance d between two
adjacent planes (Figure 88).

+ +
1
2
d
3
...
p ...
planes
p-2
Figure 88. A set of crystallographic
p-1
atomic planes taking part in diffraction.
p

• According to Bragg’s law (2dsin = n) , the amplitude of a diffracted ray will be a
maximum when the path difference between rays reflected from successive
planes is equal to a multiple of the wavelength.
• When the angle differs from  by a small amount , the path difference is
(equations 6.28):
(6.28)
129
• Since the diffracted amplitude is appreciable only for small values of , it can be
approximated that (equation 6.29):
(6.29)
• The corresponding phase difference of two successive planes is (equation 6.30):
(6.30)

• Thus, the phase difference of the first and pth planes is (equation 6.31):

(6.31)

• From optics, it is known that if n equal vectors of amplitude differ in phase by

successive uniform increments, the resultant amplitude A is (equation 6.32):

(6.32)
where  =  / 2

130
• Thus, with n = p, the resultant amplitude of the reflected wave is (equation 6.33):

(6.33)

• The amplitude of the reflected wave is a maximum (A0), when  = 0 and the
reflected rays from all planes are in phase (equation 6.34):
(6.34)

• The intensity of an electromagnetic wave varies as the square of the amplitude.

Thus, at half-maximum intensity (equations 6.35 & 6.36):

(6.35)

sin2(2pdcos/) / (2pdcos/)2 = 1/2 (6.36)

131
• Then (equation 6.37):
2 = 2 2 sin2(pdcos/) / (42 p2 d2 cos2) (6.37)
Distance between
two successive
Reflection width Wavelength Diffracting domain size atomic planes

• This shows that reflection width  increases with:

◘ Decreasing coherently scattering domain (crystallite/grain) size p.
◘ Decreasing d spacing (increasing diffraction angle).
◘ Increasing wavelength λ.

• Equation 6.35 is satisfied when /2 = 1.40. Then the FWHM (Full Width at Half
Maximum, hkl) becomes (equation 6.38):
(6.38)

• This is the Scherrer formula for the determination of diffracting domain size. It
assumes:
◘ Spherical domains.
◘ No distribution of domain sizes.
132
◘ Reflection broadening is due to domain size only.
6.3.2. USE OF THE SCHERRER FORMULA
• The general form for the Scherrer formula (equation 6.38) can be written as
(equation 4.44):
D = K / cos (6.39)
D: mean crystallite size (m), K: Scherrer constant, λ: X-ray wavelength (m), β: full width at
half maximum (rad), θ: half of the Bragg´s angle (2θ) (o)
• The constant K may assume values between 0.62 and 2.08. The value is chosen
based on how the width is determined, the shape of the crystal, and the size
distribution.
• The common (universal) value used for K is 0.89.
• The Scherrer formula was originally developed for colloidal crystallites in the size
range from 1 to ca. 30 nm. With those sizes, the reflection widths are large and
the instrumental contribution to the width is negligible (Figure 89).
• For larger crystallites, the FWHM value (β) must be corrected for the instrumental
contribution (Figure 89) by using the FWHM value of a microcrystalline standard
material (equation 6.40):
 2 =  2sample -  2reference (6.40)
◘ Without using the instrumental correction, the FWHMs of ordinary laboratory
diffractometers (ca. 0.1o) for microcrystalline materials will give false domain 133
sizes of < 80 nm.
 Scherrer formula
Calculated Crystallite Size / nm
400 for a reflection with 2: 30
o

No reference
o
LaOCl Reference (FWHM: 0.149 )
300 o
Si Reference (FWHM: 0.140 )

Figure 89. Effect of instrumental

200 broadening and the choice of the width
reference material to the domain size
100
obtained from the Scherrer formula. The
FWHMs given for the references are
0
0.0 0.1 0.2 0.3 0.4 0.5
from the Huber G670 Guinier camera.
o
FWHM /

• From the shape of the crystallite size vs. FWHM curve it can be concluded that
the applicability range for the Scherrer formula is for sizes smaller than 100 nm.
• Depending on the size range of the crystallites, the results obtained may be
strongly dependent on the accuracy of the determination of the FWHM
(Table 12).
Table 12. Domain sizes calculated for Huber G670 data from different
FWHMs with Si reference correction.

FWHM / o 1 0.5 0.4 0.3 0.2 0.19 0.18 0.17 0.16 0.15
134
D / nm 8 17 22 31 58 64 73 85 106 153
6.3.3. SCHERRER FORMULA AND NON-SPHERICAL CRYSTALLITES
• The Scherrer formula can also be used for non-spherical crystallites, by calculating
the diffracting domain size in different directions.
Example:
• For hexagonal NaYF4 nanocrystallites the crystallite shape can be
calculated with the Scherrer formula by using the following reflections
(Figure 90):
◘ 00l: distance between the hexagonal faces.
◘ h00: distance between the rectangular faces.
Figure 90. A hexagonal crystal showing the h00 and 00l
directions.

1200
3+
NaYF4:Yb ,Er
3+ (121)
Sample 1
• The powder diffraction patterns of Intensity / Arb. Units Sample 2
(300)
two NaYF4 samples were measured 900
600
(200)

(Figure 91). 500

34 35
Figure 91. Parts of the powder diffraction 600
(002)
(102)
patterns of two NaYF4 samples showing
the h00 and 00l reflections used for the 50 52 54
135
analyses. 2 /
o
• For sample 1, the FWHM values suggest (on average) 26 nm between the
hexagonal faces and 43 between the rectangular ones (Table 13), i.e. the shape
is plate-like (Figure 92).
Table 13. Domain sizes calculated for NaYF4 sample 1.

hkl 2θ / o FWHM / o Domain size / nm

200 34.56 0.2635 41
002 52.11 0.4060 26
300 52.97 0.2571 46

Figure 92. Particle shape/size

results for sample 1.

• The results are in very good agreement with

transmission electron microscopy (TEM)
(Figure 93).

Figure 93. TEM image for sample 1. 136

• For sample 2, the FWHM values suggest (on average) 27 nm between the
hexagonal faces and 21 between the rectangular ones (Table 14), i.e. the shape
is short rod-like (Figure 94).
Table 14. Domain sizes calculated for NaYF4 sample 2.
hkl 2θ / o FWHM / o Domain size / nm
200 34.62 0.4601 21
002 52.09 0.3860 27
300 53.05 0.5020 20

Figure 94. Particle shape/size

results for sample 2.

• The results are in very good agreement with

transmission electron microscopy (TEM)
(Figure 95).

Figure 95. TEM image for sample 2. 137

6.3.4. EXTENDED SCHERRER FORMULA: SIZE DISTRIBUTION
• The Scherrer method allows for the determination of the average grain size (D) of
a polycrystalline material from the of FWHM of the diffraction peak profile.
• The measurement of two widths of the same peak, allows for two parameters to
be defined: the average grain size (D) and dispersion of sizes (σ). These
parameters can be used to draw a grain size distribution (GSD) curve.
• For this, full widths at 1/5 (FW1/5M) and 4/5 (FW4/5M) maximum are used
(equations 6.41 – 6.45).
D = 2BC / (FW4/5M) (6.41)
 = 2B(C)1/2 / (FW4/5M) (6.42)

(6.43)
(6.44)
(6.45)

• Example: nanocrystalline BN (Figure 96).

138
Figure 96. (220) reflection of BN
(top) and grain size distributions
(right) calculated from the (111),
(220) and (311) reflections for
nanocrystalline BN.
139
6.4. Structure refinement (Rietveld analysis)
• Structure refinement refers to finding out the fine details of a known structure.
• The refinement is carried out by using the Rietveld method, which may also be
called whole pattern fitting.
• As stated before (in chapter 6.2.5.): In the Rietveld method, the whole diffraction
pattern is calculated based on a structural model taking into account
instrumental effects. The calculated pattern is then fit to the experimental one
by adjusting the values of both structural and instrumental parameters.
• The calculated intensities are compared to the measured ones and parameter
values are changed until a reasonable fit is obtained by a least squares
minimization of the residual S (equation 6.46):
S = Σwi(yi-yci)2 (6.46)
wi: 1/yi, yi: observed intensity at step i, yci: calculated intensity at step i.
The sum is over all datapoints i.
• The paramaters to be refined can be classified to global and phase-specific.

140
6.4.1. GLOBAL PARAMETERS
• The global parameters describe instrumental effects. The main parameters are
as follows:
a) 2θ zeropoint (diffraction angle calibration).
b) Instrumental profile (peak shape): Gaussian, Lorentzian or pseudo-Voigt
(mixture of Gaussian and Lorentzian).
c) Background: polynomial function (a + b2 + c3 + d4 +…) or experimental curve.
d) Sample displacement (deviation from ideal position in diffraction geometry).
e) Sample transparency (absorption and attenuation effects).

Figure 97. Comparison of Gaussian and

Lorentzian peak shapes.
141
6.4.2. PHASE-SPECIFIC PARAMETERS
• The phase-specific parameters describe the crystal structure details of the phase
as well as its macrostructure.
• The basic parameters are:
a) Atomic coordinates (x, y, z)i of each atom i of phase j.
b) Isotropic ”temperature factor” Bi of each atomic position i.
 Unit: Å2.
 Describes the spread of electron density in the specific atomic site.
 0 < B with B  1 Å2 considered as an acceptable higher limit.
 Indicates the correctness of the structure model used.
c) Occupancy factor Ni of each atomic position i.
 Describes the percentage by which the atomic position is filled in the total structure.
 Value between 0 and 100 %.
d) Lattice parameters a, b, c, ,  and .
e) Scale factor of phase j.
f) Profile breadth parameters U, V and W (Caglioti parameters):
 (FWHM)2 = U tan2 + V tan + W
• Other parameters are also available, e.g. to account for reflection asymmetry or
preferred orientation, etc. 142
6.4.3. CRITERIA OF FIT
• The Rietveld refinement process will adjust the refinable parameters until the
residual (Eq. 6.46) is minimized in some sense, i.e. until the ”best fit” of the
entire calculated pattern to the entire observed pattern is obtained.
• The ”best fit” obtained will depend on the adequacy of the structure model and
whether the global minimum (true minimum) rather than a local one (false
minimum) has been reached (Figure 98).
Residual

Figure 98. Example of a possible

residual surface obtainable from two
refined parameters.
143
• Various numerical indicators for the goodness of fit are used in Rietveld analysis.
The most commonly used are:
a) Rwp (R: residual; wp: weighted pattern)

 Pro: uses the complete pattern.

 Cons: uses also ranges without reflections; influenced by background height.
b) RB (B: Bragg)

 Pros: considers only the reflections, not affected by background height.

 Con: biased on favor of the model used.
c) 2

 1 < 2 < 2 is OK. Other values indicate that there is something wrong with the model.
 Con: calculated from Rwp and Rexp(ected).
144
• Commonly, for the different R values, those below 10 % are considered
acceptable.

145
• These goodness of fit values should all be considered together to judge the
goodness of the fit.
• These numerical values must always be used with the following criteria:
a) Visual inspection of the fit from the Rietveld plot (Fig. 99).
b) The physical soundness of the structural data obtained as the result:
interatomic distances , bond angles, site occupancies, temperature factors.

20000 Na(Y0.80Yb0.17Er0.03)F4 Observed

As-prepared Calculated
Difference
Bragg reflections
Intensity / Counts

15000
4800

4400

10000
4000

40 50 60

5000

30 40 50 60 70 80 90
2 / Degrees
Figure 99. Example of a Rietveld plot. 146
6.4.4. SOURCES OF ERROR
Sources originating from the Rietveld method
• The structure model from where the refinement is started must be close enough
to allow reaching the global minimum of the residual.
• To reach the global minimum, the parameters to be refined must be turned on in
the correct order and considering the correlation between the parameters.
• Commonly, the turn-on sequence is as follows:
1) Background.
2) 2 offset.
3) Scale.
4) Unit cell parameters.
5) Parameters describing the peak shapes.
6) Atomic coordinates.
7) Isotropic temperature factors.
8) Site occupancies. However, these are usually not refined.

147
Sources originating from the sample
• The following are common sources of systematic errors in Rietveld refinements
due to sample properties:
a) Preferred orientation (may be accounted for by specific parameters).
b) Definition of the background.
c) Determination of the profile shape.
d) Anisotropic reflection broadening (may be accounted for by specific
parameters).
e) Absorption of the sample (may be accounted for by specific parameters).
f) Sample displacement (may be accounted for by specific parameters).
g) Sample transparency (may be accounted for by specific parameters).
h) Graininess of the sample (too few crystallites diffracting).
• Although some of these problems can be overcome by additional parameters,
the parameter values do not have any real physical meaning. This makes it
difficult to judge whether the parameter values obtained are acceptable.

148
6.4.5. RIETVELD SOFTWARE
ANSTO GUI LHPM-Rietica for Windows
ARITVE - Glasses/Amorphous
BGMN - Fundamental Parameters Approach
CSIRO RIET7/SR5/CSRIET/LHPM
DBWS
DEBVIN Rietveld - DOS Rietveld refinement with generalized coordinates subjected to geometrical
restraints
Jana (with Powder opiton for Incommensurate structures)
Fullprof
GSAS
Koalariet/XFIT for Win95 - Fundamental Parameters Approach
MAUD for Java
Philips PC-Rietveld
PDFFIT - Refines the atomic pair distribution function (PDF) extracted from powder diffraction data based
on a structural model.
Philips Quasar - Quantitative Analysis
PREMOS91/REMOS - Programs for Modulated Structures
Rietan
Rietquan 2.2 - Quantitative Analysis
RIQAS - Quantitative Analysis
SIMREF and SIMPRO also handles incommensurate structures.
SiroQuant - Quantitative Analysis
Topas from Bruker AXS
WinMprof
XND with on-line manual also handles incommensurate structures. 149
XRS-82
Example:
• Rietveld refinement with FullProf.

150
6.5. Structure solution/determination
• The diffraction pattern from polycrystalline materials can be described as a one-
dimensional projection of three-dimensional diffraction data that results partial
and/or complete overlapping of some diffraction peaks.
• The overlapping defines the principal difference between the powder and single
crystal diffraction patterns and makes the former more challenging.
• In some cases, it is not possible to grow good or big enough single crystals. Then,
powder diffraction is needed.
• Also, in situ studies are not possible for single crystals.
• For single crystals and polycrystalline materials, crystal structure solution involves
the following steps:
a) Peak search.
b) Peak indexing.
c) Determination of possible unit cell parameters.
d) Determination of possible space groups.
e) Determination of atomic positions.
f) Structure refinement (with the Rietveld method, see chapter 6.4.).
g) Validation of the structure (as for the Rietveld method). 151
6.5.1. PEAK SEARCH, INDEXING AND UNIT CELL DETERMINATION
• Indexing is the process of assigning Miller indices to each diffraction peak and the
subsequent determination of the lattice parameters.
• No mistakes are tolerated at this stage, because those make the structure
solution impossible. Therefore, indexing is the most crucial step in structure
determination.
• Accurate peak search is the fundamental stage for correct powder pattern
indexing.
• The indexing can be performed using variety of programs such as ITO, TREOR or
DICVOL , etc.
• Each program searches for the unit cell based on about 10 to 30 reflection
positions from the low angles of the diffraction pattern.
• All programs generate several best possible solutions and the role of the
investigator is to decide which solution is the true one or at least which one to
start tests with (Fig. 100).
• The final proof of the unit cell correctness is the determination of the crystal
structure.

152
• The main challenges in pattern indexing are as follows:
a) Systematic errors in the reflection positions (e.g. sample transparency, sample
displacement, incorrect 2 calibration).
b) Presence of reflections from impurity phases.
c) Accidental non-crystallographic relationship between the cell dimensions.
d) Factors decreasing the accuracy of the (determination) of the reflection
positions (e.g. overlapping, broadening, asymmetry).

Figure 100. Example of indexing results with the program N-TREOR. 153
6.5.2. SPACE GROUP DETERMINATION
• The space group is determined based on the systematic absences obtained from
Miller indices vs. diffraction intensity data.
• The reflections belonging to each space group are listed in the International
Tables for X-ray Crystallography (Figure 101).

Figure 101. An example of the International Tables for X-ray Crystallography:

space group F222. The reflection conditions are highlighted. See figure 34 for a
more complete view of the data for this space group.

154
• Since only Miller indices are assigned to the reflections during indexing, the
intensity of overlapping reflections need to be decomposed and distributed to
each reflection.
• In the case of totally overlapping reflections, there is no way of knowing the
weight with which each diffracting plane contributes to the observed diffraction
peak.
• As with indexing, space groups typically cannot be determined uniquely, but there
are several possibilities (Fig. 102).

Figure 102. Example of space group determination with the program EXPO. 155
6.5.3. DETERMINATION OF ATOMIC POSITIONS
• This procedure aims in the creation of a structural model that is as close as
possible to the true structure.
• There are two different approaches for this: reciprocal space and real space.
• Both methods require the determination of as precise diffraction intensity as
possible for each set of Miller indices.
• Thus, the extraction of these intensities (the pattern decomposition) has to be
performed after the space group determination but before the determination of
the atomic positions in the unit cell.

Figure 103. Example of pattern decomposition with the program EXPO. 156
6.5.3.1. Reciprocal space methods
• The methods used in reciprocal space are called direct methods.
• They involve the direct use of the measured reciprocal space data, i.e. the
structure factor |F|2 values.
• The structure factor is obtained fro the intensity, I, of a reflection (hkl) (eq. 6.1):

(6.1)

• The other parameters in this equation have been explained in chapter 6.2.
• The structure factor is the vector sum of scattering intensities of all atoms
contributing to a particular reflection.
• Structure factors are vectors, since they have an amplitude (F) (i.e. magnitude)
and a phase (ƒ) (i.e. direction). In the reciprocal space methods, the vectors are
used to calculate a map of real space showing electron density.

157
Use of reciprocal space methods
• In the reciprocal space methods, the F vectors are used to calculate maps
showing electron density in real space (Figure 104).
• A simple grid search can then be used to locate the peaks in the map, which
correspond to atomic positions. Assignment of the peaks to particular atoms
provides a convenient description of the crystal structure.

Figure 104. A Fourier map of electron

densities of NaI at z = 0.5 of the xy plane.
158
• Typically, the reciprocal space programs generate a list peak heights (electron
densities) and their corresponding atomic coordinates (Figure 105).
• It is common that the programs are given the expected types of atoms and
their number in the unit cell.
• The program will place the
atom types based on the
heights.
• The Fourier maps often also
contain so called ghost atoms
(marked e.g. with Q), which
could be real atoms, or not.
• The user needs to check each
assignment and redo those that
are not acceptable.
• This will be repeated until the
whole structure is completed.

Figure 105. A list of electron densities (heights) and their corresponding atomic
coordinates for famotidine in the EXPO program. 159
Difficulties
• The diffraction experiment measures only the magnitude of F and not its phase,
and without phases a Fourier map cannot be constructed.
• It can be shown that the phases actually have a higher "information"
content than amplitudes, so that over 50 % of the information is lost in the
experiment itself.
• Fortunately, the phases are not random, but have well-defined relationships that
relate to the magnitude of the structure factors.
• Direct method programs exploit these relationships and attempt to determine
the phases directly from the observed amplitudes.
• The main assumptions are:
a) Atoms are found as peaks in the density map.
b) Due to chemical bonding, the peaks are separated, typically about 1-2 Å apart.
• Thus one requirement for success is that the data is measured to atomic
resolution, i.e. the lowest d value of a resolvable reflection is typically in the
range 0.9-1.0 Å (2 > 100o for CuKα radiation).
• This condition is easily achieved in most single-crystal experiments, but hard to
achieve in powder diffraction due to the extensive reflection overlap.
160
6.5.3.2. Real space methods
• The real space methods are based on global optimization.
• These methods are effective, if all or some parts of the structure are known, e.g.
from spectroscopic data (NMR, IR, etc.). Thus, this approach is ideal for organic
and organometallic compounds.
• In real space calculations, the known parts of the structure are kept fixed in
shape, but their positions and orientations are optimized.
• As an example, the ibuprofen molecule (Figure 106) has well defined bond
lengths as well as bond angles (determined by spectroscopic methods). However,
the torsion angles (χ) can not be determined by spectroscopic methods.

Figure 106. Ibuprofen molecule.

161
• The crystal structure solution of Ibuprofen simply involves the determination of
the four unknown torsion angles, χ, plus the position (e.g. center of gravity X, Y, Z)
and the orientation (e.g. Euler angles Θ, Φ, Ψ,) of the molecule within the unit.
From the unit cell volume, it has been concluded that there is only one molecule
per unit cell.
• This may seem a simple problem with only 10 parameters to find, but a
systematic grid search of parameter space would involve billions of permutations.
• There are three basic direct space methods that can be used: Monte Carlo,
Simulated Annealing and Genetic Algorithm.

Monte Carlo
• In the simplest form of Monte Carlo search, the parameters (e.g. χ1, χ2, χ3, χ4, X, Y,
Z, Θ, Φ, and Ψ) are assigned randomly to give a possible model for the crystal
structure.
• From this model, structure factors are calculated and compared to the observed
ones obtained from pattern decomposition.
• An agreement index, similar to the R-factors used in Rietveld refinement, is
calculated. If the agreement is good, the parameters are kept.
• The program is run for many thousands of iterations and a best set of parameters
162
is obtained. It is hoped that one or two sets will be close to reality.
Simulated annealing
• This is probably the most effective of the algorithms used, and has led to the
development of many of the "black-box" methods for structure determination
from powder diffraction data.
• The method is similar to Monte Carlo, but at a fixed elevated “temperature”.
• In practice, the “temperature” is simply a tolerance value for the agreement
index, and is simply a parameter that decreases slowly with program cycle
number. Thus it corresponds to a cooling of the system, hence the expression
simulated annealing.
• As with the Monte Carlo method, first a set of model parameters, are generated
at random and an agreement index is calculated.
• The model parameters are then allowed to vary slightly so as to provide a new set
of model parameters. In addition, some of the parameters are allowed to vary
randomly so as to include the possibility of uphill moves in the residual space (see
Fig. 98 in chapter 6.4.3.) for moving out of false minima solutions.
• All of this is repeated until the “temperature” is low and an acceptable structure
has been obtained.
• Many small organic structures can now be routinely solved with this technique.
163
Genetic algorithm
• In the genetic algorithm method, the parameters are aligned in a chain, a bit like
genes along a chromosome (e.g. χ1- χ2-χ3-χ4-X-Y-Z-Θ-Φ-Ψ).
• A population of such chains is generated with random parameters, and associated
with each one a fitness parameter is calculated.
• By applying the genetic operators of crossover (parameters swapped from chain
to another) and mutation (parameters simply altered on a given chain), a child
generation of parameters is generated.
• By using the principle of survival of the fittest, the population should evolve to
the point where the parameters of one or more individuals in the population
should match those of the correct crystal structure, i.e. the parameter set should
correspond to a global minimum.
• For this method to work efficiently, careful control of the genetic parameters (e.g.
population size, crossover rate, mutation rate) is required.

164
6.5.3.3. Software
• As for indexing and Rietveld refinement, there are several programs that can be
used determine atomic positions from powder data. These include:
a) EXPO (reciprocal and real space).
b) SHELX (reciprocal, developed for single-crystal data).
c) FOX – Free Objects for Crystallography (real space).
d) ESPOIR (real space).
e) FOCUS (real space, designed for zeolites).
f) PSSP - Powder Structure Solution Program (real space).
g) ENDEAVOUR (commercial, real space).
h) TOPAS (commercial, real space).
i) DASH (commercial, real space, designed for organic materials).
j) POWDERSOLVE (commercial, real space).

165
6.6. Analysis of residual stresses
Definitions
• Stress that remains in the material after the removal of a load.
• Stress that is not needed in order to keep the equilibrium between the material
and its surroundings.
• Given as force per area with Pascal as unit (typically some 100 MPa).
Effects
• The co-operative effect of a work load and residual stress can cause material
rupture.
• Intentionally induced residual stresses can increase
the durability of brittle materials.
Hardened glass (Figure 107)
• Endures much more hits and twisting than regular glass.
• Shatters to small crums when breaking. No potentially harmful sharp
fragments are formed.
• The hardening includes heating to 600 oC followed by a fast cooling.
This results in 5-6 times the surface tension of regular glass.
Figure 107. A table with hardened glass top. 166
Causes
• Mechanical load/working (e.g. in the preparation of metallic components).
◘ Tensile stresses are usually undesired, since they decrease durability: caused by e.g.
strong grinding or bar drawing (Figure 108).
◘ Compressive stresses may improve durability: e.g. shot peening or gentle grinding
(Figure 108).

Grinding
Tensile

Abrasive
Bar drawing
Residual Sress (MPa)

Regular

Gentle
Compressive

Figure 108. Effect of grinding on residual stress (left). Bar drawing (top right) and 167
shot peening (bottom right).
• Temperature change.
◘ Macroscopic: non-uniform heating or cooling (Figure 109).
◘ Microscopic: different thermal expansion of multicomponent material components.
• Chemical change.
◘ Stresses are caused by volume changes due to chemical changes (reactions,
phase changes and transitions, etc.).
◘ E.g. chemical surface treatments
and coating usually cause notable
stresses on material surfaces.

Figure 109. Cooling curves for the interior and

surface of an aluminum piece. At point A, these
layers partially mix due to local yielding. 168
6.6.1. CLASSIFICATION OF RESIDUAL STRESSES
Type I
• Macro residual stress.
• Long range variation, extends over multiple grains (Figure 110).
• Caused by e.g. welding or shot peening.

Type II
• Micro residual stress.
• Intra-grain variation, extends over a single grain (Figure 110).
• Common in pure materials due to the differences in the elastic and thermal
properties of differently oriented adjacent grains.
• In multi-phase materials, due to the differences in the physical (and chemical)
properties of the different components.

Type III
• Micro residual stress.
• Atomic scale intra-grain variation (Figure 110).
• Caused by e.g. dislocations and other lattice defects. 169
• Micro- and macrostresses may be present simultaneously.
◘ Common for e.g. cold worked metal.

Type I
Stress

Type III
Type II

Distance
Type II
Type I

Type III

Figure 110. Spatial scales of the different 170

types of residual stresses.
6.6.2. CHARACTERIZATION TECHNIQUES
• Different techniques give information on different types of stress.
◘ Different methods may give different values for stress.
◘ Wrong choice of technique may leave stresses undetected.

Hole Drilling
• A hole is drilled (Figure 111) to the investigated sample area and the relaxation of
stress around the hole is monitored.
◘ Detection e.g. by strain gauges, laser-interferometry or holography.
◘ The observations enable the calculation of the original residual stress.
• Penetration: ca. 1,2 x hole diameter.
• Resolution: 50 m (depth).
• Accuracy:  50 MPa (decreases with increasing depth).
• Measures type I stresses.
• Semidestructive.

Figure 111. Hole drilling apparatus for residual stress

171
determination.
M. Lastusaari 2011.
Curvature
• The formation or relaxation of stress is monitored after deposition (Figure 112a)
or removal (Figure 112b) of a layer from the material.

Figure 112. Residual stress determination by the curvature technique.

◘ Detection e.g. by strain gauges or laser interferometry.

◘ Curvature change allows the calculation of residual stresses.
◘ Common in the determination of residual stresses in layered materials.
• Penetration: 10 – 50 % of thickness.
• Resolution: 5 % of thickness.
• Accuracy limited by the accuracy of curvature determination.
• Measures type I stresses.
• Destructive.

172
Diffraction methods
• Effect of stress on crystal structure.
• X-ray diffraction:
◘ Penetration: m range.
◘ Resolution: 1 mm in plane (depends on beam size), m range in depth.
◘ Accuracy: 20 MPa.
• Synchrotron X-ray diffraction:
◘ Penetration: mm range.
◘ Resolution depends on the beam size.
◘ Accuracy: 10-5 x strain.
• Neutron diffraction (Figure 113).
◘ Penetration: cm range.
◘ Resolution: 500 m.
◘ Accuracy: 5x10-4 x strain.
• Type I, II and III stresses.
• Non-destructive.

Figure 113. Mapping of residual stresses by neutron diffraction.

173
Other methods
• Ultrasonics:
◘ Effect of stress on the velocity of sonic waves can be measured.
◘ Penetration: > 10 cm.
◘ Resolution: 5 mm.
◘ Accuracy: 10 %.
◘ Type I stress, but sensitive to microstructure.
• Magnetic measurements:
◘ Stresses cause measurable magnetic anisotropy in magnetic materials.
◘ Penetration: 10 mm.
◘ Resolution: 1 mm.
◘ Accuracy: 10 %.
◘ Type I, II and III stresses.
• Raman spectroscopy:
◘ Stresses affect bond lengths and/or symmetry.
◘ Penetration and resolution: 1 m.
◘ Accuracy 50 MPa.
◘ Type I and II (and III) stresses.

174
6.6.3. CHARACTERIZATION WITH X-RAY DIFFRACTION

1) How do different types

of residual stress affect
the diffraction pattern?
2) Which other factors give
similar effects?

Stress

?
175
Elastic deformation
• If a polycrystalline material is elastically deformed in such a way that the strain is
uniform over a relatively long distance, the distances between the atomic layers
change uniformly according to the strain.
 Macrostress (long range stress): type I.
• The changes in the distances between the atomic layers is constant in all grains
oriented similarly with respect to the stress.
 Reflection positions move.
Plastic deformation
• If the deformation is plastic (or local), the strain varies from one grain to the next
or inside a grain.
 Microstress (short range stress): types II and III.
 Reflections broaden.

• The interplanar distance change obtained from diffraction data can be used to
calculate the associated macrostrain, which subsequently can be used for the
calculation of macrostress. This can be done by using the values of elastic
constants (if known) or a calibration curve of strains produced by known stresses.

176
Effects of residual stresses on the diffraction pattern

No residual stresses.

Long range stress.

 Type I.
• Uniform strain.
 Reflections move (Figure 114).

Short range stress.

 Types II and III
• Non-uniform strain.
 Reflections broaden (Figure 114).
Figure 114. Effect of different types of
stresses on a single reflection. 177
6.6.4. LONG RANGE RESIDUAL STRESS
6.6.4.1. Principles
Uniaxial stress
• The stress affects in one direction only.
• Assume that a bar with cross section area A is being drawn by a tensile force F along
the y axis (Figure 115).
◘ The stress along the y axis (y) is then
y = F/A (6.47)
◘ No stresses along the x and z axes.
◘ The tensile stress causes a strain (y) along the y axis
y = L/L (6.48) (L: bar length)
◘ This causes a compressive strain along x and z
x = z = D/D (6.49) (D: bar diameter)
◘ Assuming that the bar material is isotropic
x = z = y (6.50) (: Poisson’s ratio).
◘ Stress and strain are related as follows:
y = Ey (6.51) (E: Young’s modulus). Figure 115. Isotropic bar under
tensile load. 178
• The strain along y (y) can be determined by measuring the change in the interplanar
spacings perpendicular to the y axis.
◘ E.g. along z: z = d/d (6.52)
◘ Giving y = (-E/)(d/d) (6.53)
• For a given (hkl) plane, the magnitude with which the strain is experienced depends
on the orientation of the crystallite with respect to the orientation of the load
( angle) (Figure 116).
Highest expansion: (hkl)  y

Figure 116. The effect of tilt angle

Highest dilatation: (hkl)  y  on observable interplanar
spacing.

 It is possible to probe whether long range stresses are present in a material by

changing the tilt angle . 179
Bi- and triaxial stress
• Usually, there are stress components in two or three orientations that are in right
angles with respect to one another.
 Bi- and triaxial stress.
• On the surface are the main stress components 1 and 2 (Figure 117).
• Normal to the surface there is the main stress component 3 (Figure 117), which
assumes a non-zero value due to the Poisson’s ratio and because X-rays will always
penetrate below the sample surface.

Figure 117. Main stress components of a planar surface. 180

• To connect bi- and triaxial stresses with diffraction data, two coordinate systems
need to be used simultaneously (Figure 118):
◘ Laboratory coordinates Li, describing the orientation with respect to the studied (hkl) plane.
 L1 are L2 on the (hkl) plane.
 L3 is perpendicular to the (hkl) plane.
◘ Sample coordinates Si, corresponding to the main stress components.
 S1 and S2 are on the sample surface.
 S3 is perpendicular to the surface.
◘ Tilt angle  is the angle between L3 and S3.
◘ Furthermore,  is the angle between L2 and S2 (corresponding to the angle between S1 and
the projection of L3 on the S1,S2 plane).

Figure 118. Definition of the Li and Si

coordinate systems and the angles  and .
181
• By using the coordinate systems Li and Si as well as the angles  and  one gets the
fundamental equation of X-ray diffraction stress analysis binding together the strain
along L3 (’33) and the strain components on the Si coordinates ():

(6.54)
where d0: stress-free lattice spacing.
• Assuming that the material is isotropic, the following applies:
ij = [(1+)/E]ij - ij(/E)kk (6.55)
where ij = 1 (when i = j) or 0 (when i  j).
• Equation (6.54) then becomes

(6.56)

182
6.6.4.2. Biaxial stress determination
The sin2 method
• Assuming only biaxial stress (when 13 = 23 = 33 = 0) and taking into account the
relationship between strain and stress, equation (6.56) can be expressed in a more
simple form:

(6.57)

• This is the equation for the sin2 method predicting a linear dependence between 
and sin2 in case of biaxial stress (Figure 119).

Figure 119.  vs. sin2 plot for biaxial stress.

183
• Stress in the direction  () is then obtained from the slope, if E and  are known.
• Because the determination of d0 may be difficult, the d value at  = 0o is often used
instead.
◘ For most materials, elastic strains can cause at most a 0.1 % change between the real d0
and d.
◘ The error in stress then never exceeds 0.1 %.
• If the stress is not truly biaxial, a non-linear curve is obtained.
◘ If 13 and/or 23  0, negative  angles give a curve that is a mirror image (with respect to
the expected linear behavior) of that from positive  angles, the so called -splitting
(Figure 120), due to the sin2 term in equations (6.54) and (6.56).

Figure 120. -splitting.

184
◘ If the curve is equally non-linear (congruent) with both negative and positive  angles
(Figure 121), the sample possesses a stress gradient.

Figure 121. Stress gradient.

◘ Other curve shapes (e.g. Figure 122) suggest a more complicated stress distribution that
can not be characterized with the sin2 method.

Figure 122. Textured sample. 185

Two tilt method
• Used for fast stress determination.
• Based on the assumption that the d value behaves linearly as a function of sin2.
• Measurements made only at two tilt angles: 0o and .
• If the change in interplanar spacing is d, Bragg’s law (n = 2dsin) states that

(6.58)
• Which gives
(6.59)
where 2: diffration angle change and K: stress constant

(6.60)

186
One tilt method
• Measurement made only at one tilt angle  =   0 (Figure 123).
• Based on the fact that the diffraction ring of a stressed structure is deformed from
circular (Figures 116 & 124). Then, the reflection is observed at different angles on
different sides of the primary beam.
• The reflection shift due to stresses can be examined
with a single measurement by using two position
sensitive detectors.

Figure 124. Effect of

stress on the diffraction
ring shape.

Figure 123. One tilt method. 187

• The stress can be calculated from

(6.61)

where n = 90-.
• Usually equation (6.61) is expressed as

(6.62)
where K: stress constant

(6.63)
• The space required by the detectors limits the usable 2 and  ranges thus limiting
the precision of the method.
• This method is usually used for ultra fast stress determination by specially designed
diffractiometers with position sensitive detectors.

188
6.6.4.3. Triaxial stress determination
• Terms a1 and a2 are defined as

(6.64)

(6.65)
• Then 11, 12, 22 and 33 are obtained from the slope and intercept of the a1 vs.
sin2 plot at  angles 0, 45 and 90o.
• 13 and 23 are obtained from the slope and intercept of the a2 vs. sin2 plot at 
angles 0 and 90o. 189
6.6.4.4. Experimental considerations
Sample and its position
• The sample surface must be planar (flat) in the studied region.
◘ The  angle is then constant throughout the studied area.
• Too large crystallite size or strong texture makes measurements difficult or
impossible.
◘ Usable  angle range is decreased.
• The sample surface must lie in the center of rotation of the ,  and  axes
(Figure 125).
◘ The surface must remain in the center while the angles are changed.
◘ The required precision is about 0.025 mm.

Figure 125. Diffractometer axes.

190
• Errors in alignment result in a 2 shift for the reflections of an unstressed sample
with changing tilt angle .
◘ This can be accounted for by measuring the 2 shift for a stress free fine grained reference,
whose composition matches that of the studied material as closely as possible.
◘ The true 2 shift is then
2 = (2)sample - (2)reference (6.66)
• The detector axis must be symmetric with respect to the  axis.
• The effect of  axis error can be minimized by using parallel beam geometry
(Figure 126).
◘ Even a 2 mm error due to  axis displacement is then negligible.

Figure 126. Parallel beam geometry.

191
Reflection selection
• Choose a sufficiently intense single reflection at as high as possible angle.
◘ The sensitivity to detect a reflection shift increases with 2 (Figure 127).
◘ Asymmetry is lower at high angles.
◘ For metal stress analysis typically at 120 – 165o in 2.
• Limiting factors at high angles:
◘ Decreasing intensity.
◘ Probability for reflection overlap
increases.
 Only materials with high-symmetry
structures may be studied.
• The diffraction angle can be influenced
by the choice of radiation.
◘ E.g. Cu K (1.54 Å) instead
of Cr K (2.29 Å).

Figure 127. Effect of 2 angle on the detection of a 0.01 Å change in d.

192
Determination of reflection angle
• The 2 step must be short enough to allow an unambiguous determination of the
shape of the reflection.
• Enough points must be collected around the reflection to allow the determination of
the background.
• If the studied reflection is broad (as usually for e.g. cold-worked metal) the combined
effect of the Lorenz polarization factor and absorption (LPA) must be accounted for.

(6.67)
◘ The experimental curve must be divided with the LPA before the determination of the
reflection angle.

193
• Many methods may be used for the determination of the angle, e.g.
◘ Half-value breadth method (Figure 128): the reflection angle is obtained as the mean value
of the 2 angles on the left and right sides of the reflection at half the maximum intensity
(background subtracted).
◘ Parabola fitting (Fig. 129), where a parabola is fit to the top 15 % of the reflection.
◘ Fitting with a Lorentz/Gauss function.
 In X-ray diffraction the shape is usually a mixture of the two and thus the pseudo-Voigt
function is used in fits.
 Many programs (e.g. the free XFit) can be used for the fitting.

Figure 128. Half-value breadth 194

method. Figure 129. Parabola fitting.
6.6.4.5. Applications
• The typical applications (Figure 130) are:
◘ Metallic industrial components.
◘ Coatings.
◘ Composite materials.
◘ Ceramic materials.

Figure 130. Papers presented at the MECA SENS IV meeting (Vienna, Austria, 2007)
classified based on material type (upper figure) and use (lower). 195
6.6.5. SHORT RANGE RESIDUAL STRESS
• Stresses of types II and III cause reflection broadening.
• Broadening is also caused by instrumental contributions (Figure 131).
• Instrument-based reflection broadening is due to, e.g.
◘ Non-ideal optics.
◘ Axial divergence.
◘ Sample transparency.
◘ Flat sample effect.
◘ Detector resolution.

Figure 131. Instrument-based reflection broadening for a Bruker D8 196

diffractometer (with a Ge monochromator and position sensitive detector).
• Sample-based reflection broadening is due to
◘ Finite crystallite/domain size.
◘ Microstrains (due to microstresses).
◘ Structural defects.
 Cause both microstrain and domain size effects (e.g. separation of defects).
 May also give rise to asymmetry and/or peak shift.
• For a reflection, integral breadth  is defined as
 = A / I0 (6.68)
where A: integrated intensity (area) and I0: maximum intensity (counts) with the background
removed.
• Correction for instrumental effects to the line broadening can be carried out as
follows:
◘ Lorentzian shape:  sample=  total-instrumental (6.69)
◘ Gaussian shape:  2sample=  2total-2instrumental (6.70)
• The sample boadening contains contributions from the size and strain
effects.
 sample=  size+strain (6.71)
197
• Commonly it is assumed that
◘ Size broadening is Lorentzian.
◘ Strain broadening is Gaussian.
• The shape factor f can be calculated from the integral breadth and the full width at
half maximum (FWHM):
f = FWHM/ (6.72)
◘ Lorentzian: f  0.636.
◘ Gaussian: f  0.939.
• The size and strain effects can be separated more precisely due to their different
dependencies on .
◘ Size (D): D = /(sizecos) (: wavelength) (6.73)
◘ Strain ():  =  straincot (6.74)
• Combining (6.71), (6.73) and (6.74) gives
 sample= /cosD +  /cot (6.75)

198
• The reciprocal d and  are defined as
d* = 2sin/ (6.76)
* = cos/ (6.77)
• Equation (6.75) can be written with the reciprocal terms as
* = 1/D + 2d* (6.78)
◘ Size broadening: no d* dependency (size is absolute).
◘ Strain broadening: d* dependency (strain is relative).
◘ Size and strain can be obtained from the intercept and slope of a * vs. d* plot.
 Williamson-Hall method.
• The value thus obtainable is the r.m.s. microstrain containing both compressive and
tensile strain.
◘ Assuming similar contribution from both, the maximum tensile or compressive strain is
obtained by dividing the r.m.s. strain by two.
◘ Subsequently, the microstress  is obtained by using Young’s modulus.
 = ½E (6.79)

199
Williamson-Hall method
• Integral breadth method of line profile analysis.
• Williamson-Hall plots can be used to separate different sample based factors giving
reflection broadening.
• The Williamson-Hall equation can be presented in FWHM terms as
K = (1/D) + 2d* (6.80)
where K = 2cos()/ (where : ½*half width at half maximum (rad))
• Crystallite size and strain are obtained from the intercept and slope of a
K vs. d* plot.
• Easy to use: only reflection breadth is required.
• Works well with pure Gaussian and Lorentzian reflections.
• May not be used quantitatively, if the breadth has both d*-dependent and
d*-independent contributions.
 Different methods should be used in combination to ensure reliable
results.

200
• Basic types of Williamson-Hall plots:
◘ Horizontal: no strain, spherical crystallites.
 Isotropic size effect.
◘ Horizontal for the different orders of the same reflection, but different reflection families
form their own horizontal parts (Figure 132): no strain, non-spherical crystallites.
 Anisotropic size effect.

c c

h00 10l

b
a
a
00l
In this case, the domains are cylindrical
(spherical in the ab plane). The 10l
plane approaches the 00l with
increasing l resulting in a systematic
dependence of broadening on
Figure 132. Williamson-Hall plot for a direction.
stress-free sample. Prismatic crystallites.
201
◘ Non-horizontal line with all reflections on the same line (Figure 133).
 Isotropic microstrain.
 Through origin: no size effect.
 Not through origin: size effect.
◘ Non-horizontal line for different orders of the same reflection, but different reflection
families have their own slopes and intercepts.
 Anisotropic microstrain.
 Through origin: no size effect.
 Not through origin: size effect.

Figure 133. Williamson-Hall plot: isotropic

strain and size effect.

202
◘ Broadening depends on hkl, but not systematically with respect to d* or direction
(Figure 134): structural defects (e.g. stacking faults).
 Shape factor f vs. d* plots (Figure 135) may be constructed to judge whether
size/strain components are present.

Figure 135. f vs. d* plots for the same

Figure 134. Williamson-Hall plot: reflections.
structural defects and size effect. 203
Other methods
• Rietveld analysis.
◘ Whole pattern fitting to a structural model.
◘ Strong correlation with different peak shape parameters may cause
unreliable results.
• Warren-Averbach analysis (Fourier method).
◘ More tedious to use than Williamson-Hall:
 Mathematically more complex.
 Exact reflection shapes required.
◘ Reflection overlap causes problems.
 Not suitable for low symmetry structures.
◘ Different orders of the same reflection used.
◘ Gives also distribution of sizes and microstrains.
• Double-Voigt method.
◘ Works with reflections that are not pure Gaussian or Lorentzian.
◘ Not as sensitive to reflection overlap as Fourier methods.
◘ Gives also distribution of sizes and microstrains.

204
6.6.6. EXAMPLES
Ceramic material

205
• Y3Al5O12:Nd3+ (Nd:YAG) is the most common solid state laser material (: 1064 nm).
• Usually used as single crystals.
• Ceramics allow the preparation of larger entities and with higher Nd3+ content
(increased efficiency).
◘ Transparency required.
◘ Preparation from nanocrystalline powder: pressing and heating.
• The effect of pressure on the stresses in ceramic Nd:YAG pellets (Figure 136) was
studied with laboratory X-ray diffraction.

Cold pressing Heating, 450 oC

1 GPa 2 GPa 4 GPa 6 GPa 8 GPa

Figure 136. Y3Al5O12:Nd3+ nanoceramics treated at different pressures.

206
• No impurities were detected (Figure 137).
• No preferred orientation (To be discussed in chapter 6.7.)
was observed (Figure 138).
◘ Nanosized crystallites.
◘ Cubic (isotropic) structure.

Figure 137. X-ray powder diffraction

patterns of Y3Al5O12:Nd3+ nanoceramics
treated at different pressures.

Figure 138. Texture pole figures for the (420) and (611) reflections of 207
Y3Al5O12:Nd3+ nanoceramics treated at different pressures.
• Williamson-Hall plots (Figure 139) yielded crystallite sizes and microstrains.
◘ Isotropic microstrain.
◘ Strain increases with pressure.
• Microstrains affect emission line width (Figure 140).

Figure 140. Effect of microstrain

Figure 139. Williamson-Hall plots for on the emission line width of
Y3Al5O12:Nd3+ nanoceramics treated at Y3Al5O12:Nd3+ nanoceramics
different pressures. treated at different pressures. 208
• sin2 plots (Fig. 141):
◘ Heating initially decreases stresses, but increasing pressure increases stresses.
◘ Compressive stress ( < 0).
◘ Stress gradient.

Cu K1,2 radiation.
2: 87.3o
Step: 0.04°, time/step: 5 s.
-tilt: 0.0, ±26.6, ±39.2, and ±50.8o.

Figure 141. sin2 plot for the (10 4 0) reflection of Y3Al5O12:Nd3+ nanoceramics
treated at different pressures.
209
210
Layered material

211
• Cutting tools and other technical components are often coated with CVD/PVD
(Chemical/Physical Vapor Deposition) methods to improve their durability.
• The coatings are typically composed of multiple layers with different thickness and
composition.
• Surface durability is may be further improved by deliberate creation of stresses (e.g.
with grit blasting).
• In this work, a multilayered Al2O3/TiCN coating on WC/Co was studied.
◘ Composition: 2.5 m TiCN and 2.5 m Al2O3.
◘ X-ray diffraction was used to investigate the effect of grit blasting.
◘ Experimental setup:
 Co K and Cu K radiation.
 sin2: 0.05o at 0 - 89.5o.
 Reflections: Al2O3 (116) ja (024), TiCN (220).

212
• Without grit blasting, both layers show isotropic tensile stress ( > 0) (Figure 142).
• Grit blasting produces an isotropic compressive stress ( < 0) to both layers
(Figure 142).

Figure 142. Effect of grit blasting on the residual

stresses in Al2O3/TiCN coatings: sin2 plots. 213
• The results indicated considerable stresses (GPa range) in the coating layers
(Figure 143).

Figure 143. Stress distribution in the Al2O3/TiCN coatings.

214
Metal

215
• A cold-drawn steel wire (0.7 % carbon content) was studied.
◘ The wire diameter was drawn from the original 1.16 to 0.25 mm.
 Corresponds to a 2,7 GPa tensile strength.
◘ The effects of the drawing on the residual stresses was studied by X-ray diffraction.
◘ For sub-surface stress analysis, surface was removed by using a HF/H2O2/H2O solution.
◘ Experimental setup:
 Different radiation was used for different reflections (Figure 144).
 20-30 equally long wire pieces were glued next to each other on a plate (Figure 145).

Figure 144. Studied reflections

and the radiation used.

216
Figure 145. Setup with  = 0o.
• With  angle 0o (Figure 146), the surface shows compressive stress with small 
angles (Figure 147), but with wider angles the plot suggests nearly isotropic tensile
stress.
◘ Below the surface the stress distribution becomes even more complicated (Figure 152).

Figure 146. Figure 147. sin2 plots at = 0o for the original cold drawn
Diffraction at  = 0o. wire (left) and those with surface removed by chemical
treatment. 217
◘ The complication is connected with the increasing degree of texture below the surface
(Figure 148).

Figure 148. Texture plots for the original cold drawn wire (left) and those with
surface removed by chemical treatment as in Figure 147.

218
• With  angle 90o (Figure 149) the stresses weaken with increasing distance from the
surface (Figure 150).
◘ Isotropic tensile stress.

Figure 149. Diffraction at  = 90o.

Figure 150. sin2 plots at  = 90o for the

original cold drawn wire and those with
surface removed by chemical treatment.
219
• The authors finally used a non-trivial triaxial stress model with elastic anisotropy.

220
6.7. Texture analysis
• Each grain/crystallite in a polycrystalline material normally has an orientation
(crystallographic) different from that of its neighbors.
◘ Crystallographic orientation: how the atomic planes in a volume of crystal are
positioned relative to a fixed reference.
• The orientations of all the crystallites may be randomly distributed or they may
tend to cluster about some particular orientation or orientations.
• The latter condition is said to have preferred orientation or texture.
• As single crystals often have different properties in different directions, a
polycrystalline material with preferred orientation may also have directional
properties.
• Properties that depend on the average texture of a material include:
◘ Young’s modulus.
◘ Poisson’s ratio.
◘ Strength.
◘ Ductility.
◘ Toughness.
◘ Magnetic permeability.
◘ Electrical conductivity.
221
◘ Thermal expansion (in noncubic materials).
• Usually, directional properties are undesirable in polycrystalline materials, but
there are exceptions.
• E.g. the steel sheets used in transformer cores must undergo repeated cycles of
magnetization and demagnetization. Since iron crystals are best magnetized in
the [001] direction (Figure 151), the grains in these materials are preferred to
orientate in this direction.

Figure 151. Anisotropy of magnetization

in different crystal directions of iron.

222
Texture (macrotexture)
• The most established method for measuring textures is X-ray powder diffraction,
which gives a measure of the volume fraction of a particular family of planes,
which are orientated for diffraction.
• The texture thus obtained is an average value for the whole sampled volume,
typically comprising thousands of grains. Macrotexture.
• Preferred orientation is observed as (Figure 152):
◘ The increase of the intensities of reflections from atomic planes parallel to the sample
surface with excessive volume fraction.
◘ The decrease of the intensities of reflections from other atomic planes.
• Macrotexture can be studied by Rietveld or pole figure analyses.
BaSO4 Powder
(with random orientation)
(210) (211)

(111) (020)
(102) (112)
2000
Intensity / Counts

(011)

(201)
(002)
Figure 152. Comparison of the X-ray
powder diffraction pattern of BaSO4
1000
with extreme texture (a single crystal)
Single crystal and with random orientation (powder).
(i.e. excessive preferred orientation)

223
20 24 28 32
o
2
Microtexture
• The X-ray texture tells what volume fraction of the specimen has a particular
orientation. However, it does not tell how these grains are distributed
throughout the material.
• An approach that deals with the orientation statistics of a population of individual
grains, and usually encompasses the spatial location of these grains, has been
termed microtexture.
• If the spatial location plus the orientation of individual grains in a sample
population is known, access to new knowledge is obtained, since the
crystallographic and morphological aspects of structure evolution become fused.
• E.g. the following phenomena, which are all directly linked to microtexture (i.e.
individual grain orientation), can be explored:
◘ Local property effects.
◘ Interfacial parameters and properties.
◘ Morphological/geometrical grain parameters (size, shape and location).
◘ Orientation variations within individual grains: By classical definition, a grain has a
single, distinct orientation. However, in practice the orientation within a grain may vary
in a continuous or discontinuous manner, which can be termed orientation
perturbation. In particular, there may be perturbations close to interfaces and in
deformed structures reflecting the distribution of strain. 224
• Methods for microtexture analysis:
◘ Scanning electron microscopy (SEM).
 Electron backscatter diffraction (EBSD).
 Selected area scanning (SAC).
◘ Transmission electron microscopy (TEM).
 Selected area diffraction (SAD).
 Kikuchi diffraction.

Figure 153. Categorization of the mainstream techniques for texture determination

according to the radiation used as a probe.

225
Descriptors of orientation

• The orientation of crystallites can be defined by using different descriptors

(Figure 154).
• This chapter addresses those commonly used for X-ray based texture analysis, i.e.
macrotexture analysis.

Figure 154. Common

orientation descriptors.

Macrotexture
226
6.7.1. POLE FIGURE
• The direction of any 3D vector in a crystal (a crystallographic direction or the
normal to a crystal plane) can be described as a point on the unit reference
sphere, that is, a sphere with radius 1 notionally residing around the crystal
(Figure 155).
• The point of intersection of the normal to this plane with the reference sphere,
that is, its pole, is a measure for the arrangement of this plane in the crystal.
• Provided that the reference unit sphere is attached to an external frame, the
position of the pole on the sphere also provides information on the
crystallographic orientation of the crystal with respect to this frame, although the
crystal has still one degree of freedom by rotating around this particular axis.
◘ The position of a pole on the unit sphere with
regard to an external reference frame is
described by the two angles α and β.

Figure 155. Orientation of the basal plane (0001)

in a hexagonal crystal. However, since the crystal
can still rotate about the (0001) pole, for an
unequivocal definition of the orientation of the
crystal more information, here the position of the 227
(1010) pole, is required.
• Representation of the 3D orientation information on the unit sphere in a 2D
plane requires a projection of the sphere onto that plane (Figure 156).
• As mentioned earlier, one pole does not yield the entire orientation information,
since the crystal can still rotate about this particular pole (Figure 155), but
depending on the symmetry of the crystals, two or three poles may be needed to
describe the orientation unambiguously.
• Usually, the additional information is provided by other poles of the same family
of planes, i.e. different poles (hkl) of the family {hkl}.

Figure 156. (a) Crystal in the unit sphere; (b) projection of the {100} poles onto the
equator plane; (c) {100} pole figure and definition of the pole figure angles α and β for
the (100) pole. 228
6.7.2. POLE FIGURE MEASUREMENT
• For polycrystalline samples the intensity recorded at a certain sample orientation
is directly proportional to the number of crystallites, i.e. the volume fraction of
crystallites, which currently is in a reflection condition.
• Thus, the pole figure of a polycrystalline sample reflects the distribution of
orientations in that sample, i.e. its crystallographic texture.
• Texture can be measured using a special texture goniometer, which is similar to
that used for macrostress measurements (Figure 157). Both reflection and
transmission setups are possible.
• The data collection for a single reflection consists
of a series of measurements with
◘  fixed to the angle of the reflection.
◘  fixed to zero.
◘  and  scanned.
• The results then yield the orientation
matrix as a pole figure with the Miller
index notation.

229
Figure 157. Diffractometer axes.
• The measurements can be carried out either by changing the  and  angles
simultaneously or one at a time (Figure 158).
• Conventionally, angle steps of 5o and times per step of 1 s are used in the
measurements. This yields a total measuring time of less than 30 min per pole
figure.

Figure 158. (a) Spiral and (b)

circular, equal-angle scanning grids
for pole figure measurements (note
that for clarity both scanning
schemes are shown in steps of 20°).

Choice of reflections
• It is often necessary to use a number of different crystal planes to determine a
complicated texture.
• Requirements for the reflections:
◘ Strongly diffracting planes.
◘ Discrete reflections (no overlap).
◘ 2 angles: above 20 (less asymmetry) and below 50o (negligible peak shift due 230
to defocusing). In this range, the reflections are also usually the strongest.
Data errors and correction
• When the sample is tilted along the  axis, the irradiated area (as well as its
shape and volume) on the sample surface changes (Figure 159). Because of this,
the diffraction intensity will change even for a non-textured sample.

Figure 159. Shape of the irradiated spot in dependence on tilting angle  and Bragg angle
θ for a square-shaped incident beam (left). Effect of tilting angle  on signal width due to
defocusing (right). Note: here  = .

• The tilting on the  axis also moves a part of the sample surface off the focusing
circle. This leads to reflection broadening due to defocusing (Figure 159).
231
• The defocusing also causes reflection shifts even if there is no residual stress
present (see chapter 6.6). At the low 2 angles typical of texture analysis,
however, this effect is usually negligible.
• Both of these effects (area change & defocusing) yield erroneus counting values
for the detector at the selected angle and must therefore be accounted for
before analysis.
• Also, the background height usually changes with the tilting angle and thus needs
correction.
• After pole figure measurement and subsequent correction of the data with
respect to background intensity, defocusing error, and, if necessary, absorption,
the pole figure data are available as number of counts, or counts per second, per
pole figure point.
• For representation of the pole figures and for a subsequent evaluation, however,
it is necessary to normalize the intensities to standard units that are not
dependent on the experimental parameters (Equation 6.81).
Inormalized = 1/N * Icorrected (6.81)
where N = iIcorrected sin i/ isini

232
• Therewith, the integral over the full pole figure range becomes 1, which means
that the pole figure of a randomly oriented standard sample would be 1 at all
points.
• Pole figure regions with densities larger than 1 indicate that more lattice planes
are aligned in those directions than in a random sample, and vice versa.
Presentation of pole figures
• The corrected and normalized pole figures are presented as 2D contour plots or
3D plots (Figure 160).

Figure 160. Contouring

of a 3D pole density
distribution
to derive the standard
2D pole figure
presentation.

233
6.7.3. INTERPRETATION OF POLE FIGURES
• Sometimes the solutions may be fairly obvious, sometimes they can be compared
visually with known textures and sometimes they require common sense and
brain power.

Random texture
• A random texture means the texture for a sample with no preferred orientation.
Its crystallites are randomly distributed in orientation.
• There is statistically the same number of grains diffracting at any point in the pole
figure. Thus the diffracted intensity at any point in the pole figure is the
same.
• All the (hkl) planes are randomly distributed and thus all pole figures are
homogeneous (Figure 161).
• Each crystallite has three degrees of freedom (rotations around three orthogonal
axes) to orient.

234
  = 0o  = 22o  = 45o
Reflection
Intensity

Figure 161. With random orientation, there are always an equal number of atomic
planes diffracting regardless of the tilt angle  (top) or the hkl index. (100), (001)
and (110) pole figures of a sample with random texture (bottom). 235
Planar texture
• This kind of texture is exhibited by samples in which crystallites have two degrees
of freedom to orient.
• This means that the crystal can have different orientations with respect to two of
its axes, but the orientation with respect to the third axis is always fixed.
• This is called a planar texture, because it involves two axes with random
orientation.
• If e.g. the rotation axis is X and the direction family <100>* of an orthorhombic
crystal system. The <100>* directions are then distributed in the XZ plane. But,
since other directions are at random around <100>*, their pole figure remain
homogeneous
(Figure 162).

Figure 162. (100), (001) and (110)

pole figures of a sample with
planar texture (top). Degrees of
freedom projected on the XZ 236
plane (bottom).
• Planar textures with the rotation axis around Z are called cyclic-planar textures
(Figure 163).
• The rotation axes may also be other than the sample main axes (X, Y, Z). Then,
the resulting pole figures will be more complex than the examples given here.

Figure 163. (100), (001) and (110) pole figures of a sample with a cyclic planar texture. The
rotation axis of the cyclic-planar texture is with <100>* directions perpendicular to Z.

237
Fiber texture
• If the crystallites lose another degree of freedom in their orientation, they are
forced to align one of their <hkl>* with one direction of the sample, all other
directions being at random around <hkl>*.
• Such a texture with only one degree of freedom is called a fiber texture.
• If e.g. all the <001>* from an orthorhombic crystal system are located in the
direction  = 45° and  = 45°, the following pole figures result (Figure 164):

X
Figure 164. (100), (001) and (110) pole figures of a sample with a cyclic planar texture. The
rotation axis of the cyclic-planar texture is <001>* with  = 45° and  = 45°.

• If e.g. the <001>* from an orthorhombic crystal system are aligned with Z, cyclic
fiber texture is obtained (Figure 165). This is common for plate-like crystallites.
  = 0o  = 22o  = 45o
Reflection
Intensity

Figure 165. For plate-like crystallites with the (001) planes preferentially oriented normal to
the sample Z-axis, diffraction from the (001) plane is obtained only at tilt angle  = 0o (top).
(100), (001) and (110) pole figures of a sample with a cyclic fiber texture. The rotation axis
of the cyclic-planar texture is with <001>* directions perpendicular to Z (bottom).
3D texture
• Samples composed of crystallites which have no choice for different orientations
(0 degrees of freedom) will exhibit punctual poles for all their pole figures
(Figure 166).
• The pole figures will be similar to the ones of a single crystal. This texture is called
3D texture.

Figure 166. (100), (001) and (110) pole figures of a 3D texture of the orthorhombic crystal
system, with <001>* directions along Z and <100>* directions along X.

• For a real single crystal, the reflections will be of far higher intensity than for a
polycrystalline material with perfectly aligned crystals.
• With normalization, however, the pole figures of the two will be similar. 240
6.7.5. EXAMPLES
Pharmaceutical compact

241
242
243
244
245
246
247
Coating

248
249
250
251
6.8. Determination of particle size: SAXS
• The particle size may play an important role in applications with spatial or particle
mass restrictions, e.g. when using luminescent particles for
◘ In vivo imaging.
◘ Biotechnical/medical assays.
• A number of different methods are available for particle size analysis, e.g.:
◘ Sieving (usually for particle sizes of 3 m and larger).
◘ Optical microscopy (a few m and larger)
◘ Sedimentation: based on measurement of sedimentation velocity (30 nm and larger).
◘ Static light scattering, i.e. laser diffraction (100 nm and larger).
◘ Dynamic light scattering (1 nm – 10 m).
◘ Electron microscopies (SEM, TEM).
◘ Acoustic methods.
◘ Atomic force microscopy.
• Regular X-ray powder diffraction usually covers diffraction angles starting from
4 degrees in 2, which corresponds to interplanar distances of 22 Å and higher
for CuK radiation (according to Bragg’s law).
• Smaller scattering angles down 0.001o to are covered by small angle scattering
(SAS) methods SAXS (Small Angle X-ray Scattering) and SANS (Small Angle
Neutron Scattering) (Figures 167 & 168).
• In these low angles, scattering by particle or pore boundaries with separations in
the range of 1 – 100 nm can be observed. Thus, the SAS techniques can be used
for particle (and pore) size determination. Shape information may also be
obtained.

Figure 167. Comparison of regular X-ray powder diffraction (XRD), i.e. wide 253
angle X-ray scattering, and SAXS.
Figure 168. Comparison of size regions reachable by regular X-ray diffraction, regular
SAS and special SAS techniques.
254
6.8.1. SAXS SETUP
• Because of the very small angles measured, SAXS requires a special setup with
high angular resolution, strict beam collimation and the detector further away
from the sample than in regular X-ray powder diffraction.
• This may be achieved within a regular powder diffractometer (shorter sample to
detector distance) (Figure 169) or a dedicated SAXS apparatus (longer distance)
(Figure 170).
• Only transmission mode is possible.

Detector
Sample X-ray tube
stage

Figure 169. A regular X-ray powder diffractometer with a point detector converted to
SAXS mode. 255
Figure 170. Bruker NANOStar SAXS apparatus with a 2D detector.

256
• SAXS setup requirements:
◘ Monochromatic radiation (X-ray mirror may be used).
◘ High intensity.
◘ Well collimated beam (slits): decreases intensity but improves resolution.
◘ Minimal axial divergence (soller slits).
◘ Small background around the beam.
• SAXS measurements can be done for solids, liquids and their combinations (gel,
suspension etc.).
• The samples may be placed between e.g. glass plates or in a capillary tube (Figure
171) with the maximum sample thickness in the order of 1 mm.

Figure 171. Sample holders

for SAXS.
257
6.8.2. GENERAL ASPECTS OF SAXS
• Due to the multi-particle nature of the samples studied by SAS as well as their
more or less random orientation all the scattering data is averaged to one
dimension.
• The three-dimensional structure is represented by a one-dimensional function
describing the dependency of the scattering intensity I(q) on the scattering angle
(equation 6.82):
I(q) = A * A  fifj * sin(qrij) / (qrij) (6.82)
f: atomic scattering factor, rij: distance between atoms i and j
• In SAS, the data is usually presented in the scattering vector (q) scale (equation
6.83):
q = (4 / ) * sin( / 2) (6.83)
• The amplitude A of the radiation scattered from a sample is proportional to the
Fourier transform of the electron density. The amplitude can not be measured,
but the intensity I(q), which is the square of the amplitude (equation 6.82), can be
measured.

258
• The electron density can not be solved exactly from the scattering intensity, but
the scattering intensity can be calculated for known electron densities: the
inverse scattering problem (Figure 172).
• The inverse scattering problem needs to be solved before anything useful may be
obtained from the SAS data.

Scattering Function
Distance Distribution Function

Inverse Scattering Problem

Figure 172. Reproducing the structure from the scattering curve (dashed arrows): the
inverse scattering problem.
259
• To solve the structure from the scattering data, a series of calculations have to be
carried out to find out what kind of scattering curves result from the models of
the expected structures (Figure 173).
• Monte Carlo and Simulated Annealing can be used.
• The models are modified based on the differences between the
observed and calculated scattering data.
• Finally, after several iterations, the structural model or a few models describing
the structure within their statistical accuracy,
are found.
• In a general case, it is assumed that the scatterers
are statistically isotropic with no long range order
and that they are embedded dilutely in a
homogeneous matrix so that the average distances
between the particles is much more than their sizes.
• These assumptions apply for particles in a solution
or inhomogenities in a solid.

Figure 173. Calculated scattering functions of a sphere

(−) , prolate ellipsoid (-.-) and oblate ellipsoid (--). 260
6.8.3. DATA ANALYSIS
• The central peak scattered intensity gets broader as the domain size (particles,
pores) decreases (Figure 174), as in regular wide angle diffraction.

• For the analysis, the SAXS signal is

divided to two ranges (Figure 175):
◘ Guinier (low q).
◘ Porod (high q).

Figure 174. Effect of domain size on

the SAXS signal breadth.

Figure 175. Scattering in the SAS range

(note: h = q). 261
6.8.3.1. Accounting for instrumental contribution
• The first procedure in SAXS data analysis is to remove the instrumental
contribution from the sample. This includes:
◘ Background removal (constant, due e.g. to a glass sample holder).
◘ Desmearing.
• Smearing can result if the radiation is not strictly monochromatic and if the shape
of the beam is not point-like.
• The desmearing (Figure 176) can be carried out if the spectrum of the used
radiation and the shape (and dimensions) of the beam are known.

desmeared

smeared

Figure 176. Effect of smearing.

262
6.8.3.2. Guinier law
• The Guinier approximation for the scattering intensity at low q values states that
the shape of the intensity curve is Gaussian (equation 6.49):
I(q) = I(0) * exp[-(qRg)2 / 3] (6.84)
Rg: radius of gyration (domain size); For spherical particles Rg = (3/5)1/2 R, where R is the
radius of the sphere
• The R value can then be obtained by the slope (R2) of the linear part in a
ln(I) vs. q2 curve (Figure 177).
• The Guinier law holds for differently shaped particles (sphere, rod, plate,…
provided that they are not very long or very flat.

Figure 177. The linear fit in the Guinier

approximation. 263
• The Guinier law assumes a dilute system and identical particles.
• If the experimental data does not follow the law (when q2 < 0.1 nm2) (Fig. 178):
◘ The system is aggregated. Then the intensity may increase strongly towards q = 0.
◘ The system is too dense. Then the intensity may diminish at small q, because of
interparticle interactions.
◘ The particle size varies.

Possible aggregation or effect of direct beam

Possible interparticle
interactions Figure 178. Deviations from
Guinier law at low angles.

264
6.8.3.3. Porod law
• The Porod law states that SAXS intensity shows a power law behaviour at large q
values (Figure 179).
• The Porod approximation is written as (equation 6.85):
I(q)  4 * (-0)2 * S / q4 (6.85)
: electron density, S: total surface area
• Gives information on the particle shape, e.g.:
◘ Sheets: I(q) ≈ constant /q2.
◘ Thin rods: I(q) ≈ constant /q1.
◘ Fractals: I(q) ≈ constant /qa (a gives the fractal dimension of the system).
• Also information on particle volume and surface area.

Figure 179. Calculated scattering functions for a

particle with different domain sizes according to
the Porod law.
265
6.8.3.4. Distance distribution function
• Distance distribution p(r) gives the relative number of distances of two points
inside the particle as a function of the distance r (Figure 180).
• It is obtained from the intensity as (equation 6.86):
p(r) = (1/2π)3 ∫ dq 4πq2 I(q) (sinqr/qr). (6.86)

Figure 180. Distance distribution functions calculated for a cube and a sphere with the
same Rg (left) as well as sphere (−) , prolate ellipsoid (-.-) and oblate ellipsoid (--) with
the same Rg (right) . 266
6.8.3.5. Evaluation of data from interacting systems
• When the concentration of the scattering particles is high, interparticle
interference distorts the scattering curve at small q values.
• It may be possible (if the particle size/shape does not depend on concentration)
to perform experiments in a concentration series and to extrapolate the obtained
R values to zero concentration (Figure 181).

Figure 181. Extrapolation to zero concentration.

267
• If the interparticle contributions are not taken into account, transforming the the
whole q range yields an incorrect pair distribution p(r). The interaction effects
cause strong oscillations in the p(r) function (Figure 182). These effects may be
eliminated by cutting off the lowest q values.
◘ Allows the determination of the p(r) of an inhomogeneous spherical particle by indirect
Fourier transform.
◘ As the method requires cutting off a part of the Guinier region, false R values will be
obtained for all but spherical particles.

Figure 182. Cut-off method. 268

• The generalized indirect Fourier transformation (GIFT) approach uses the
optimization of different parameters in the scattering function to find the
minimum in the parameter surface. This method is well suited for homogeneous
and inhomogeneous spheres and cylinders.

Figure 183. Basic procedure of GIFT. 269

6.8.3.6. Examples
Kinetics: Polymer-SiO2 core-shell materials

270
271
272
273
274
275
Kinetics & mechanism: Ag nanoparticles

276
277
278
Particle sizing: Fe3O4

279
280
281
Particle sizing: Na(Y,Yb)F4

TEM XRD (Scherrer)

25 nm 27 nm x 24 nm
(hexagon width x plate thickness)

50 nm 30 nm x 28 nm

1000 nm 150 nm x 80 nm

282
6.9. Characterization of thin layers & surfaces
Grazing Incidence Diffraction
• X radiation has a large penetration depth into any matter.
• Thin films are difficult to analyze due to their small diffracting volumes, which
result in low diffracted intensities compared to the substrate and background.
• This combination of low diffracted signal and high background make it very
difficult to identify the phases present.
• Accordingly, special techniques must be used when analyzing thin films.
• By increasing the path length of the incident X-ray beam through the film, the
intensity from the film can be increased so that conventional phase identification
analysis can be carried out.
• The most common technique for analyzing films in
the range 20 – 100 nm is to use a grazing (typically
below 3o) incidence angle (Figure 184) combined with
a parallel beam geometry.
 Grazing Incidence Diffraction (GID).

Figure 184. GID uses a narrow incident beam 283

and produces a broad diffracted beam.
• With grazing incidence, the penetration depth is reduced by three orders of
magnitude from regular diffraction, i.e. from 1 – 10 m to 1 – 10 nm.
• GID offers:
◘ Phase analysis on very thin layers.
◘ Surface sensitivity down to nm scale.
◘ Depth profiling of the phase composition of layered samples.
◘ Crystallography of thin films.
• At grazing angles, it is also possible to use X-ray reflection (XRR) and grazing
incidence small angle X-ray scattering techniques (Figure 185).

Figure 185. X-ray techniques

at grazing angles. 284
6.9.1. PRINCIPLES
• With low enough incident angle, the incoming X-ray beam is totally reflected (no
diffraction).
• Above this critical angle (C), i.e. when  > C, a part of the X-rays penetrate the
sample causing diffraction, the rest are reflected (Figure 186).
• The critical angle can be calculated as follows (Equations 6.87 & 6.88):
C = (2)1/2 (6.87)
 = (2e2/2mc2) * N(Z+f’) (6.88)
where N: number of unit cells scattering,
Z: atomic number, f’: real resonant
scattering term.

Figure 186. Evolution of X-ray penetration depth and 285

reflectivity at grazing incidence angles.
• Above the critical angle, the reflecting amplitude is obtained as (Equation 6.89):
(6.89)

• The transmitting (penetrating) amplitude is obtained as (Equation 6.90):

(6.90)

in these equations:

Figure 187. Wave vectors at grazing incidence angles above the critical angle.
286
• Below the critical angle, the incident beam experiences total external reflection
and an exponentially fading evanescent wave propagates at the surface (Figure
188).
• The evanescent waves penetrate the sample within a depth of a few nanometers.
Thus, there is always a small portion of the beam that penetrates the sample.
• Incident angles below the critical angle can thus be exploited to probe the surface
of the sample.

Figure 188. Total

reflection causing
evanescent waves.

• An ordered 2D system (surface) gives rise to rod-like Bragg reflections (instead of

cone-like) that contain information on the electron density along the z-axis of the
ordered objects.
• The total cross section for scattering from a 2D system is in general very small
and high X-ray flux from synchrotron sources is required. 287
• The intensity of a GID reflection has a maximum at the critical angle of the
material (Figure 189). The amplification obtained is ca. fourfold.

Figure 189. Enhancement of the

transmission function at the critical
angle.

288
• At the very lowest incident angles, GID does not give diffraction from atomic
planes parallel to the surface.
 Not suitable for study of materials with strong preferred orientation parallel
to the surface.
• Instead, the evanescent waves probe layers perpendicular to the surface, while at
incident angles higher than the critical angle other planes also give diffraction
(Figure 190).
• The information obtained normal to the surface (qspec) is from total reflectivity.
This is used in the method X-ray reflectivity (XRR).
◘ Gives information parallel to qspec and can thus not distinguish between different in-
plane (surface) structures.

Figure 190. The direction of information at

289
grazing incidence angles.
6.9.2. LABORATORY GID SETUP
• The incident and diffracted beams are made nearly parallel by means of a narrow
slit or Göbel Mirrors on the incident beam and a long Soller slit on the detector
side (Figure 191).

Figure 191. A typical laboratory GID setup. 290

• The stationary incident beam makes a very small angle with the sample surface
(typically 0.3 - 3°), which increases the path length of the X-ray beam through
the film.
• This helps to increase the diffracted intensity (diffracting volume increases
proportionally), while at the same time, reduces the diffracted intensity from the
substrate (Figure 192).
• Overall, there is a dramatic increase in the film signal-to-background ratio.

Figure 192. Comparison of the

diffraction patterns for a 100 nm
Mo layer on a 1400 nm AlN layer
deposited on a Si substrate
obtained with Bragg-Brentano
(blue) and GID (red) setups.

291
• During the collection of the diffraction pattern, only the detector rotates through
the angular range, thus keeping the incident angle, the beam path length, and the
irradiated area constant.
• Also 1D and 2D detectors can be used.
• The long Soller slit on the receiving side allows only those beams that are nearly
parallel to arrive at the detector. This has an added advantage of reducing
sensitivity to sample displacement from the rotation axis.

6.9.3. DEPTH-DEPENDENT GID

• GID can be used to determine the phase composition of layers on a substrate.
• The angle of incidence is fixed so that the degree of penetration into the sample,
by the X-rays, is kept constant throughout the measurement
• At lowest incidence angles the X-rays penetrate only the uppermost layers of a
sample, whereas at higher incidence angles they penetrate deeper into the
sample.
• Thus, by changing the incident angle, successive layers on a sample surface can
be sampled using GID.

292
 • Example: layered Zn – CuGaInSe – CdSe/Mo material (Figure 193).
◘ At low incidence angles (0.3 & 0.45o) the X-ray beam only samples the topmost Zn layer.
◘ At 0.6 and 1.0o, also the CuGaInSe layer is detected.
◘ At 2.0o, the pattern contains signals from all layers.

Figure 193. GID patterns of a layered Zn – CuGaInSe – CdSe/Mo material measured at 293
different incidence angles.
6.10. Analysis of short-range order
• With increasing complexity of materials, the traditional crystallographic structure
solution approach is no longer sufficient to understand properties of materials on
an atomic scale.
• The limitation of traditional structure refinements is the fact that it only yields
the long range average structure of the material.
• If we consider e.g. a structure which has 70 % atoms and 30 % vacancies, the
ordering of these can be completely different (Figure 194).

Figure 194. Simulated structures with random order (left) and short-range 294
chemical ordering (right).
• The analysis of the diffraction patterns for these two materials (Figure 195) using
the Bragg reflections will give the same results, since the Bragg peak positions
and intensities are the same.
• However, the background of the diffraction pattern of the material with short-
range order shows diffuse (wide) features not present for the randomly ordered
case (Figure 195).

Figure 195. Diffraction patterns of the two structures in figure 194. The top figure 295
shows a zoom in on the backgrounds of the two patterns
• While Bragg scattering contains information about the long range average
structure, diffuse scattering contains information about in general two-body
correlations, in this example about the chemical short-range ordering.
• The atomic arrangements hold the key to understanding and possibly predicting
the properties of materials. E.g. the particular ordering shown in figure 194, could
be determining the properties of that material.
• To study the local structure, one must go beyond the analysis of Bragg scattering
or beyond traditional crystallography. For this, the total scattering pattern, which
includes Bragg and diffuse scattering, needs to be analyzed.

Figure 196. 2D single crystal diffraction patterns of the two structures in figure 194: 296
random (left) and with short range order (right).
6.10.1. THE PAIR DISTRIBUTION FUNCTION
• The pair distribution function (PDF) gives the probability of finding an atom at a
given distance r from another atom. In other words, it can be understood as a
bond length distribution (Figure 197).

Figure 197. A simple atomic arrangement (left) and the bond length distribution for
the same arrangement (right).
297
• The PDF is obtained via a Fourier transform of the total scattering pattern
(Figure 198) and it yields the local structure on length scales within as well as
beyond the unit cell size.

Figure 198. Generation of

the pair distribution
function.
• The study of the PDF on different length scales allows one to obtain information
about the nano-scale structure of materials.
• Consider the PDF from a sample that has some kind of domain structure with a
characteristic domain size of x.
◘ Looking at the PDF on a length-scale << x the pairs of atoms contributing will
predominantly lie within the same domain as each other and the resulting structure will
be the intra-domain structure.
◘ Studying the same PDF at a distance r > x/2 most of the pairs of atoms giving rise to
features in the PDF lie in different domains and the result is the average of the different
local structures.
• Example: The C60 can be considered as a microstructured material where each
bucky ball is a single domain with a radius of 7.1 Å.
◘ Sharp peaks are observed in the PDF coming from the characteristic C-C pairs on the ball
up to 7.1 Å (Figure 199).
◘ Thereafter only broad, featureless structure exists.
◘ The sharp peaks are the intramolecular structure and the broad feature the inter-
molecular structure (or ball-ball correlations).

Figure 199. The C60 molecule and

its pair distribution function.
299
• For the example structures in figure 194 the PDFs are rather similar (Figure 200),
but the difference curve reveals that the structure with short-range order has
e.g. more atom-atom pairs at the distance of 2.5 Å.

Figure 200. Pair distribution functions for the two structures in figure 194. 300
Applications
• Materials with disorder.
◘ Semiconductors.
◘ Relaxor ferroelectrics.
• Nanomaterials.
◘ Nanoparticles.
◘ “Objects” with nanoscale dimensions (tubes, belts, sheets, …).
• Amorphous phases (Liquids, metallic glasses).
• Other complex materials.
◘ Precursor phases from “soft” chemistry.
◘ Layered materials with turbostratic disorder.

301
6.10.2. OBTAINING THE EXPERIMENTAL PDF
• The PDF is obtained from the powder diffraction data via a sine Fourier transform
of the normalized scattering intensity S(Q) (Equation 6.91, Figure 201):

(6.91)

where (r): microscopic pair density, 0: average number density and Q: magnitude of the
scattering vector. For elastic scattering Q = 4 sin /  with 2 being the scattering angle
and  the wavelength of the radiation used.

Figure 201. The equations for obtaining the pair distribution function. 302
• The trouble with the Fourier transform is that we would like to measure our data
up to Q infinity which of course is not possible. The cutoff at finite Q decreases
the real space resolution of the PDF.
• E.g. for ZnSe0.5Te0.5 the difference in bond lengths for Zn-Se and Zn-Te is 0.14 Å.
◘ The PDF was calculated by using Q = 17 and 40 Å-1 as the cutoff limit.
◘ This had a clear effect on the resolution for the distances at 2.5 Å resulting in the loss of
possibility for differentiating the Zn-Se and Zn-Te distances (Figure 202).

Figure 202. Pair distribution function for ZnSe0.5Te0.5 obtained with two
different cutoff limits. 303
• The cutoff of Q = 17 Å-1 corresponds to using MoK radiation (Table 15), which
has the highest energy conventionally available for laboratory diffraction.
• Even though for some applications the use of a laboratory X-ray source might
yield useful PDF data, in many cases high energies are required to be able to
access high values of Qmax. In the case of X-rays this requires the use of high
energy synchrotron radiation.
Table 15. Qmax values for different laboratory and
synchrotron X-ray wavelengths.
Qmax / Å-1
/Å 2max = 160o 120o 90o
K edge
Cr 2.28970 5.4 4.8 3.9
Fe 1.93604 6.4 5.6 4.6
Co 1.78897 6.9 6.1 5.0
Cu 1.54056 8.0 7.1 5.8
Mo 0.70930 17.4 15.3 12.5
Ag 0.55940 22.1 19.5 15.9
Synchrotron
0.40000 30.9 27.2 22.2
0.30000 41.3 36.3 29.6
0.20000 61.9 54.4 44.4 304
0.12370 100.0 88.0 71.8
• With the laboratory (X-ray tube) wavelengths, the resolution decreases clearly for
Cu (Qmax = 8 Å-1) and other anodes with even longer wavelengths (Figure 203).

Figure 203. Pair distribution functions for

carbamazebine with different cutoff limits.

305
• Also the Q resolution (Q) has an effect on the PDF (Figure 204).
• The Q resolution results in an exponential dampening of the PDF peaks as a
function of r , i.e. exp-(rQ)2/2.
• Using high Q-resolution data for PDF
analysis allows one to access medium range
distances, i.e. 1 - 10 nm (10 - 100 Å).

Figure 204. PDF of Ni obtained from GLAD instrument at IPNS (filled circles) and NPD at
LANSCE (solid line). Note: the displayed range extends to r = 100 Å. The scattering data
were terminated at Qmax = 35 Å-1 for both data sets. 306
• Compared with the EXAFS technique (to be discussed in chapter 7), which is
capable of determining only the first and second nearest neighbor distances, the
PDF method has an advantage of describing the short- as well as medium-range
local structure.
• Having done the experiment, the remaining challenge is to obtain the properly
corrected and normalized structure function S(Q) which is then Fourier
transformed.
• For X-ray diffraction data a few programs designed for this are available, e.g.
PDFGetX and RAD.

307
6.10.3. MODELLING THE PDF
• After having obtained an experimental PDF, the question arises how to analyze it.
• The simplest analysis is to obtain bond length information and coordination
numbers by fitting the near neighbor peaks of the PDF (single peak fitting), but in
most cases the comparison to or the refinement of a structural model is desired.
• The PDF is simply the bond length distribution of the material weighted by the
respective scattering powers of the contributing atoms.
• The PDF can be calculated from a structural model using the relation (Equation
6.92):
(6.92)

where the sum goes over all pairs of atoms i and j within the model crystal separated by rij.
The scattering power of atom i is bi and <bi> is the average scattering power of the sample.

308
• Full profile refinement of the PDF based on a structural model can be carried out.
◘ Parameters such as lattice parameters, anisotropic atomic displacement parameters,
position and site occupancies can then be refined.
◘ This is similar to the results of a Rietveld refinement, but the structural model obtained
is strictly only valid for length scales corresponding to the r-range used for the
refinement.
◘ This opens up the possibility to study the local structure on different length scales by
varying the r-range refined.
• This approach works very well in situations where the model structure can be
described with relatively few atoms.
• In other cases the Reverse Monte Carlo technique (RMC) can be used to match a
structural model containing many atoms to an experimental PDF.
• The RMC algorithm works as follows:
◘ The PDF is calculated from the chosen crystal starting configuration and a goodness-of-
fit parameter 2 (comparing the observed and calculated PDF curves) is computed.
◘ The RMC simulation proceeds with the selection of a random site within the crystal.
◘ The system variables associated with this site, such as occupancy or displacement, are
changed by a random amount, and then the model PDF and the 2 are recalculated.
◘ The change of the 2 before and after the generated move is computed. Every move
which improves the fit (2 < 0) is accepted.
◘ Also bad moves (2 > 0) are accepted, but with a lower probability.
◘ All accepted moves can be used as new starting points. 309
◘ This procedure is repeated until a (global) minimum is reached.
6.10.4. EXAMPLES
Strain analysis of microcrystalline ZnSe1-xTex

310
311
312
313
314
315
Gold nanoparticles

316
317
318
319
Amorphous (fumed) SiO2 with laboratory X-rays

320
7. X-RAY ABSORPTION SPECTROSCOPY
• X-ray absorption spectroscopy (XAS) measures inner shell transitions induced by
the absorption of X-rays (E: 1 - 200 keV / 0.06    12 Å).
• The process is the photoelectric effect (Figure 13) and it goes as follows
(Figure 205):
1. Incoming X-ray photon interacts with a deep-core electron.
2. Deep-core electron is promoted to some unoccupied state above the Fermi energy
(highest occupied state), propagates away, and leaves behind a core-hole.
3. A short time later (1 or 2 fs), a higher-lying electron decays into the core-hole and emits
a photon.

n: 3 / M edges 1 2 3
n: 2 / L edges

n: 1 / K edge

Figure 205. The basic process in X-ray absorption spectroscopy. 321

• Both X-ray absorption and emission (Figure 206) can be monitored by XAS.
• Usually:
◘ Absorption for matrix elements (high concentration).
◘ Emission (actually excitation measurements) for dopants (low concentration).

Figure 206. K edge X-ray absorption (left) and some possible emission transitions (right).
322
• Each element has a characteristic set of absorption and fluorescence energies
(see chapter 2.1.).
• Thus, XAS can be used for the identification of elements and their valence states.
• Usually, K (K emission) or LIII (L emission) edges are used.

Table 16. X-ray absorption and emission energies for europium.

Edge Energy / eV Line Transition Energy / eV
K 48519 K1 LIII  K 41542
LI 8052 K2 LII  K 40902
LII 7617 L1 MV  LIII 5846
LIII 6977 L2 MIV  LIII 5817

MI 1800 L1 MIV  LII 6456

• As the absorption or emission intensity is measured as a function of X-ray energy,

i.e. tunable X-ray energy is required, the measurements must be carried out at a
synchrotron.
323
 Yb
Tm
10000
Dy
9000

8000
Eu
Pr Nd
E / eV

7000 Ce

6000
Ba
5000

4000 Ca E(Sr): 16105 eV

Figure 207. L and K (Ca and Sr) absorption energies for selected alkaline earth and rare
earth elements.

324
The X-ray absorption spectrum
• The X-ray absorption spectrum can be divided to five features or regions
(Figure 208).

XANES
2
White line
Absorbance / Arb. Unit

EXAFS

NEXAFS

Eu2O3
Eu(LIII) edge
Pre-edge Fluorescence mode
296 K
0
7000 7200 7400
Energy / eV

Figure 208. X-ray absorption spectrum measured at the EuLIII edge of Eu2O3.

325
• Pre-edge.
◘ Electronic transitions from a core level to unoccupied or partially occupied energy
levels.
◘ Possible information on coordination geometry.
• Edge.
◘ Electronic transitions with high probability, also to continuum states.
◘ Information on oxidation state.
• White line.
◘ Transitions to unoccupied levels close to continuum.
• Post-edge.
◘ Electronic transitions to continuum states.
◘ Photoelectrons ejected.
◘ Information on interatomic distances.
• Techniques:
◘ XANES (X-ray Absorption Near Edge Structure): pre-edge, edge, white line.
◘ NEXAFS (Near Edge X-ray Absorption Fine Structure): beginning of post-edge.
◘ EXAFS (Extended X-ray Absorption Fine Structure): post-edge.
326
7.1. XANES
• The absorption edge energies are element dependent. Thus, the XANES spectra
can be used for elemental identification.
• Moreover, the absorption edge energy is sensitive to the oxidation state of the
measured element.
◘ The energy of an edge increases as the oxidation state of the absorber increases.
◘ This can be explained using an electrostatic model, since atoms with a higher oxidation
state should have a higher charge, thus requiring more energetic X-rays to eject core
electrons.
• The energy of an absorption edge is not well defined. It can be taken as the
energy at half-height or, more commonly, as the maximum in the first derivative
with respect to energy.

327
• Edge spectra frequently have unresolved transitions
superimposed on the rising edge (Figure 209).
These will affect any attempt to define a unique

Absorbance / Arb. Unit

edge energy. 6

• The signal in the XANES region is much stronger

than in the EXAFS region. Thus, very low
concentrations down to < 10 ppm are 4
BaTiO3
detectable, which makes the method very
sensitive to impurities. Ti(K) Edge

• The XANES may contain all the same information 2

IV
Ti
obtainable from the EXAFS region, but at present
the models describing the XANES region are not
good enough for the extraction of the complete 0
4950 5000
information.
Energy / eV
Figure 209. XANES spectrum: TiK edge of BaTiO3.
Benefits of XANES
• (Usually) non-destructive (unlike ICP or AAS).
• Quantification of element concentrations is possible (although not as sensitively
as in ICP or AAS). 328
• Relative quantification of different valence states possible.
7.1.1. d TRANSITION METALS
Pre-edge features
• For the K edge of a 3d transition metal,
the pre-edge features (Figure 209) arise from the 1s  3d transitions, and are
observed for every metal that has an open 3d shell.
• Although the 1s  3d transition is forbidden by dipole selection rules, it is
nevertheless observed due both to 3d + 4p mixing and to direct quadrupolar
coupling.
• The sensitivity to the 3d + 4p mixing means that the intensity of the 1s  3d
transition can be used as a probe of geometry, with the intensity increasing as
the site is progressively distorted from a centrosymmetric environment.
• The intensity change may be accompanied with merging or splitting of the signal
(Figure 210). Thus, also the number of pre-edge peaks may give information on
the coordination.
• On the other hand, the intensity is also dependent on bond strength (Figure 211).
• The analogous 1s  4d transition for the 4d transition metals is generally not
observed.

329
Figure 210. TiK edge XANES spectra for compounds with four-, five- and six-coordinated (from
top to bottom) titanium (left). Effect of degree of long-range order on the pre-edge feature of
Ti (right).

330
Figure 211. The correlation between the intensity of the vanadium 1s  3d pre-edge feature
and the bond strength for different coordination polyhedra.
331
• The intensity of the 1s  3d transition as well as its energy can sometimes also
give information on the valence state of the absorbing element (Figure 212).

V2+ VIV
d3 d1

V3+ VV
d2 d0

V3+ / IV
d2 / 1

Figure 212. VK edge XANES spectra for vanadium oxides with different oxidation 332
states for V (left). Effect of vanadium oxidation state on XANES features (right).
Other features
• For first transition series metals, the allowed 1s  4p transition is sometimes
observed on the K edges.
• This transition can sometimes be used to deduce geometry, as well. E.g. it is
intense for square–planar complexes but weak for tetrahedral ones (Figure 213).
• On the LIII and LII edges, 2p  4d transitions can be observed for the 4d
elements.
• These transitions are relatively sharp,
and allowed, thus making them intense
(white lines).
• Similar spectroscopic advantages
(narrow lines, allowed transitions)
are found for the L edge studies of the
3d transition metals. However, in this
case the very low edge energy is
experimentally challenging, requiring
the use of ultra-high vacuum for the
sample.

333
Figure 213. NiK edge XANES spectra for two four-coordinated complexes.
7.1.2. 4f ELEMENTS
• For the rare earth elements, the white line by the L III (and LII) absorption edge can
be used as a fingerprint of the oxidation state as they are separated by ca. 8 eV
between the different valences (Figure 214).
• In analogy to the d transition metals, the white
lines by the LIII and LII absorption edges are due
to the 2p  5d transitions.
2+
Eu
3
• There are no other XANES features, but due 3+

Absorbance / Arb. Unit

Eu
to the high intensity of this white line, it is
difficult to define the exact energy of
the absorption edge. 2

2+
Sr2MgSi2O7:Eu
1
Eu(LIII) Edge

Figure 214. XANES spectrum: EuLIII 0

6950 7000
edge of Sr2MgSi2O7:Eu.
Energy / eV 334
7.2. EXAFS
7.2.1. EXAFS FINE STRUCTURE
• The ejected photoelectrons form waves, which are scattered back from the
neighboring atoms (Figure 215).
• Depending on the distance between the absorbing atom and its backscattering
neighbors as well as the photoelectron wavelength, the outgoing and
backscattering waves are either in-phase or out-phase.
• As the wavelength is changed during the measurement, alternating constructuve
and destructive interference results. This is observed as the oscillating fine
structure in the EXAFS region.
• The physical origin of EXAFS is thus electron scattering observed by spectroscopic
means.

Figure 215. Photoelectron waves going out of

the absorbing atom (Eu, red) and backscattered
from a neighboring atom (O, green). 335
• This immediate scattering from the neighboring atoms (Figure 216) is called
single scattering and it is the main contributor to the EXAFS fine structure.
• Multiple scattering (Figure 216) is also possible.
◘ Less intensive than single scattering due to its lower probability.
◘ Abundant in the NEXAFS range and present in the low EXAFS range.
• Multiple scattering is extremely angle dependent.
◘ For scattering angles less than ca. 150o, multiple scattering is weak and can often be
neglected.
◘ For angles between 150 and 180o, multiple scattering must be considered.
• The angle dependence of multiple scattering means that EXAFS can, at least in
principle, provide direct information about bond angles.

Absorber
Backscatterers

Figure 216. Single (dashed arrows) and multiple scattering (solid arrows).
336
Factors affecting the fine structure

• The atomic number of the backscatterer affects the fine structure amplitude
(Figure 217).
◘ Heavy atoms have a strong scattering power resulting in high EXAFS amplitude.
◘ Light atoms show a low EXAFS amplitude due to their weak scattering power.

Figure 217. Effect of backscatterer atomic number on the fine structure amplitude. 337
• The bond length (distance between absorber and backscatterer) affects the fine
structure wavelength (Figure 218).
◘ Long bonds give short modulation.
◘ Short bonds give long modulation.
• The coordination number affects
the fine structure amplitude
(Figure 218).
◘ More bonds give higher amplitude.
◘ Less bonds give lower amplitude.
 The structural environment of
the absorbing atom can be
calculated from EXAFS data.

Figure 218. Effect of bond length and

coordination number on the fine structure
modulation wavelength and amplitude. 338
7.2.2. THE EXAFS EQUATION
• The EXAFS range is typically defined to start 20 - 30 eV above the absorption
edge.
• For a single absorber-scatterer pair the alternating interference will give rise to
sinusoidal oscillations in the absorption coefficient if the energy is given in units
proportional to the inverse photoelectron wavelength, i.e. the photoelectron
wave vector k defined as (Equation 7.1):
k = [2me(E – E0) / h2]1/2 (7.1)
me: electron mass, h: Planck’s constant, E: energy, E0: photoelectron binding energy
• This can be approximated as (Equation 7.2):
k = [0.2625(E – E0)]1/2 (7.2)
• EXAFS, (k), can be expressed as the fractional modulation in the X-ray
absorption coefficient (Equation 7.3):
(k) = ( - 0) / 0 (7.3)
: observed absorption coefficient, 0: the absorption that would be observed in the
absence of EXAFS effects.
• Since 0 cannot be directly measured, it is approximated, typically by fitting a
smooth spline function through the data.
• Division by 0 normalizes the EXAFS oscillations “per atom” and thus the EXAFS
represents the average structure around the absorbing atoms.
• For quantitative analyses, the EXAFS equation can be written as follows with the
sum taken over all of the scattering atoms near the absorber (Equation 5.4):

(7.4)

Basic amplitude Damping Phase shift

340
• The amplitude and phase (the photoelectron wave undergoes a phase shift when
passing through the potential of the absorbing and scattering atom) parameters
contain the information necessary to identify the scattering atom.
• E.g. sulfur and oxygen introduce phase shifts ij(k) that differ by approximately .
• Unfortunately, both the energy dependence of the photoelectron scattering and
phase shift depend only weakly on scatterer identity, and thus it is difficult to
identify the scatterer with precision.
• This means that e.g. O and N, or S and Cl, typically cannot be distinguished, while
O and S can.
• The EXAFS amplitude falls off as 1/R2. This reflects the decrease in photoelectron
amplitude per unit area as one moves further from the photoelectron source (i.e.
from the absorbing atom).
• The main consequence of this damping is that the EXAFS information is limited to
atoms in the near vicinity of the absorber.
• The Debye–Waller factor (j), is the root-mean-square deviation in absorber-
scatterer distance. This damping is always present due to zero-point thermal
motion, and may, for polyatomic systems, also occur as a consequence of
structural disorder.
• As a consequence of the damping terms, EXAFS oscillations are typically
341
only observed for atoms within approximately 5 Å of the absorbing atom.
7.2.3. DATA HANDLING AND ANALYSIS
Spline removal
• The EXAFS data handling starts with the spline removal, i.e. the extraction of the
contribution from the absorption edge (Figure 219).

Figure 219. A spline (red) used for

background removal.

342
• After spline removal, the EXAFS is obtained. It is presented in the k scale and
usually with e.g. k2 weighting, since the EXAFS decays fast (Figure 220).

Figure 220. EXAFS after spline removal (top).

The same presented with k2 weighting
(bottom).

343
Fourier transform
• Although the EXAFS equation (Equation 7.4) provides a complete description of
the EXAFS oscillations, it is not a particularly convenient form for visualizing the
information content of an EXAFS spectrum.
• Fourier transformation can be used to decompose a k-space signal into its
different constituent frequencies. This yields a pseudo-radial distribution function
(Figure 221).
◘ In practice, the starting k value is usually set to 3 Å-1 to exclude the multiple scattering
effects present at lower k.
◘ The ending k value should be set as high as possible, since the resolution of the distance
distribution increases with increasing kmax. For the L edges k > 12 Å-1 is seldom reached,
but for the K edges k values exceeding 20 Å-1 are common.
◘ The usable k range may be limited by the presence of other absorption edges of the
studied element or those of other elements present in the sample.

344
Figure 221. EXAFS data (left) and its Fourier transform (right) for a THF solution
of CuCN2LiCl. The Fourier transform clearly shows three distinct peaks,
reflecting the presence of three distinct absorber–scatterer interactions, as
indicated above the Fourier transform.

345
• The Fourier transform is useful for obtaining a qualitative understanding of a
system.
• However, Fourier transforms are subject to several potential artifacts and cannot
be used for quantitative data analysis.
◘ Depending on the resolution of the data, multiple shells of scatterers do not necessarily
give rise to multiple peaks in the Fourier transform.
◘ Two peaks may appear to be well resolved despite the fact that they have substantial
overlap. If the Fourier components from the two different scatterers interfere
destructively, this leads to a false minimum in the modulus (Figure 222).

Figure 222. Individual Fourier

transforms for two scattering shells
with different distances from the
absorber (bottom). Fourier transform
of the sum of the EXAFS of the two
shells (top).
346
Structural modelling
• When the pseudo-radial distribution has been obtained from Fourier transform, a
structural model (e.g. from crystallographic data or known interatomic distances
and angles) can be constructed and fit to the Fourier transform.
• Next, the structural parameters can be
refined to obtain a good fit between the
experimental and calculated radial
distributions (Figure 223).
• The results then give refined values for
interatomic distances, coordination
numbers and the Debye-Waller factors.

Figure 223. A refined fit of the two first

coordination spheres in FeO. The
results are shown in R (top) and k
(bottom) space.
347
7.2.4. BENEFITS AND REQUIREMENTS
Benefits of EXAFS
• Probes local structure (< 1 nm).
• Gives element-specific local structural information (unlike total X-ray scattering,
see chapter 6.10.).
• The EXAFS signal is dominated by the core hole lifetime (1 - 2 fs). It is thus very
fast and may be sensitive to different dynamical behaviour.
• Practically all elements can be studied (unlike in Mössbauer spectroscopy).
• Powders (both crystalline and amorphous), ceramics, liquids, suspensions, etc.
can be measured with the original state preserved.

348
Requirements for structural probing
• Good quality spectrum (low noise compared to fine structure amplitude).
• No overlapping absorption edges: at least 500 eV of undisturbed fine structure is
needed for calculations.
• K edges should always be used, if possible, because their signals are clearly
stronger than those of the L edges.
• Difficulties for studies of low-concentration elements (e.g. dopants in
luminescent materials or transition metals in minerals):
◘ The signal intensity is proportional to concentration.
◘ Absorption (or emission) of a matrix element may create a high background.
◘ If multiple dopants are present, their absorption edges may be too close to each other
for calculations.

349
7.3. EXPERIMENTAL SETUP
• The typical setup for XANES/EXAFS measurements contains three ionization
chamber detectors (Figure 224):
◘ I0 for the measurement of initial beam intensity.
◘ I1 for transmitted intensity.
◘ I2 for measuring a reference.
• Optionally, a fluorescence detector (If) and a cryostat may be installed.
• Transmission and fluorescence signals
may be obtained simultaneously.

Fluorescence
Double crystal detector
monochromator

I0 I1 I2
Sample Reference
350
Figure 224. A typical XANES/EXAFS setup.
• The transmission signal is obtained as (E) = ln(I0/I1). It shows a decrease with
increasing energy (Figure 225).
• The fluorescence signal is (E) = If/I0. It shows an increase with increasing energy
(Figure 219).

Fluorescence

Transmission

Figure 225. Comparison of the X-ray

absorption spectrum of the same sample
obtained in fluorescence (top) and
transmission mode (bottom).

351
Fluorescence detection
• For fluorescence detection, energy dispersive detectors are typically used.
• Then, the desired emission is chosen as a region of interest to discriminate all
other energies (Figure 226).
◘ The integrated area of the peak will change in the same way as the absorption factor
when scanning over the absorption edge.

MnK edge as an example

K edge energy: 6537.7 eV
Kα emission energy: 5898 eV

Figure 226. Selection of a region of interest for energy dispersive

fluorescence detection (left) and a Vortex energy dispersive fluorescence 352
detector (right).
• If a non energy dispersive detector (e.g. a PIPS) is used for detection, unwanted
emission may be suppressed by using filters.
• E.g. a Mn filter can be used to filter out the FeK emission from FeK (Figure 227).

Figure 227. The principle of using a filter to remove unwanted emissions.

353
• With the use of filters, there is always the possibility of emission from the filter,
as well. This can be removed by using soller slits (Figure 228).

Figure 228. The principle of using a soller slit to remove filter emission.

354
Sample holders
• For powder materials, the sample is either pressed to a disc or used as such.
• The sample holder typically contains multiple holes for the samples (Figure 229).
The discs can then be attached by frames and the powders by sealing them with
Kapton tape to their holes.
• These same holders can be used for other kinds of solid samples as well.

Figure 229. Different kinds of sample holders

for EXAFS/XANES measurements of solid
materials.
355
• If cryostats (liquid N2 or He) are used for controlled cooling of the samples, the
holders are attached to the end of a cold finger (Figure 230).

Figure 230. Sample holders

with cold fingers.

• The holders can be made of metals (e.g. Al or Cu) or polymers (e.g. polyacetal).
• Metallic holders will provide better thermal conductivity, but they also
potentially provide contamination signals, if the beam hits the holder to any
extent.
• Even if the metallic holder material itself has no disturbing emissions, its own
impurities can have a strong effect (e.g. Fe in Al).
• For liquid samples, cuvettes with Kapton windows are used.
356
7.4. EXAMPLES
Persistent luminescence material

357
358
359
360
361
Glassy material

362
363
364
365
8. X-RAY FLUORESCENCE SPECTROSCOPY
8.1. General features
• X-ray fluorescence spectroscopy (XRF) is an analytical method to determine the
chemical composition of all kinds of materials.
• The materials can be in solid, liquid or powder form.
• XRF is fast, accurate and non-destructive, and usually it requires only minimal
sample preparation.
• Applications include metal, oil, polymer, plastic and food industries as well as
mining, mineralogy, geology, environmental analysis of water & waste, pharmacy
and materials research.
• The precision and reproducibility of XRF analysis is very high. Very high accuracy
can be obtained when good standard specimens are available, but good accuracy
can be achieved also without specific standards.
• XRF is based on the photoelectric effect (see Figures 13 and 205) caused by the X-
ray irradiation of the sample.
• Each element in the sample will give its own characteristic fluorescence spectrum
(see chapter 2.1.) and thus the spectrum observed for a sample is a combination
of the spectra of the individual elements (Figure 231). 366
 Figure 231. Example of an XRF spectrum.

• The elemental composition is obtained by deconvoluting the total spectrum and

taking into account the calibration for each element. Generally the emission line
intensities of individual elements are proportional to their concentration.
• Detectable concentrations range from ppm levels upwards.
• The detectable element range extends to U, but there are limits for the lightest
detectable element, which is typically Na or Be depending on the setup (for
example air absorbs low-energy X-rays). 367
8.2. Special features to be taken into account
8.2.1. SECONDARY FLUORESCENCE
• Because the radiation emitted by the elements in the samples are X-rays (primary
fluorescence), they may be absorbed by other elements in the sample causing
characteristic emission from them (secondary fluorescence) (Figure 232).

Figure 232. Primary and secondary X-ray fluorescence.

• A spectrometer will measure the sum of primary and secondary fluorescence, and
it is impossible to distinguish between these two contributions. 368
• The contribution of secondary fluorescence could be up to 20 %.
• It is possible that also tertiary and even higher order radiation occurs, but in
almost all practical situations these are negligible.

8.2.2. ABSORPTION AND ANALYSIS DEPTH

• With thick enough samples, X-radiation produced in the deeper layers will not be
able to leave the sample.
• This depth limit depends on the material and the energy of X-radiation (Figure
233).

Figure 233. Analysis depths for three different X-ray lines in various materials.
• When a sample is measured, only the atoms within the analysis depth are
analyzed.
• If samples with various thicknesses are analyzed, the thickness must be taken into
account. 369
8.2.3. COHERENT AND COMPTON SCATTERING
• A part of the incoming X-rays are scattered by the sample instead of producing
characteristic radiation.
• The scattering may be coherent (Figure 10) or Compton scattering (Figure 12).
• Light elements show high Compton scattering intensity, because they have many
loosely bound electrons.
• Compton scattering is not observed for the heaviest elements (Figure 234).
• A peak due to Compton scattering could thus be present in an XRF spectrum on
the high-energy side of the peak of the X-ray source.

Figure 234. Compton and Rayleigh (coherent) scattering for light (Perspex material)
and heavy (metallic Pb) elements. 370
8.3.3. BREMSSTRAHLUNG
• If the spectral output of the X-ray source is not filtered, bremsstrahlung (Figure
15) may be observed, because it also scatters from the sample.
• With proper filtering, the bremsstrahlung is suppressed and signal-to-noise ratio
is improved (Figure 235).
• However, often bremsstrahlung is needed to excite the lightest elements
selectively to make them detectable.
Pb lines
Ag lines

Figure 235. XRF spectra measured from a Pb target using the unfiltered and 371
filtered Ag tube.
8.4. XRF spectrometers
• The basic concept for the spectrometers is to have a source, a sample and a
detection system (Figure 236). The source is usually an X-ray tube.
• XRF spectrometer systems can be divided into two main groups: energy
dispersive (EDXRF) and wavelength dispersive (WDXRF).

Figure 236. Basic XRF

construction (top left)
and XRF spectrometers
in different shapes and
sizes
372
8.4.1. EDXRF
• In the simplest configuration, the X-ray tube irradiates the sample directly and the
fluorescence coming from the sample is measured with an energy dispersive
detector (Figure 237). In this configuration there is usually the possibility to use a
filter to remove e.g. bremsstrahlung.
• This configuration is also called a 2D geometry, because the X-ray path is on one
plane, i.e. it is two-dimensional.
• An alternative is to place a secondary target between the tube and sample
resulting in indirect excitation of the sample (Figure 237).
Indirect (3D)
Direct (2D)

Figure 237. Basic configurations of EDXRF spectrometers. 373

• The secondary target emits its characteristic X-rays which are used to excite the
sample. Using such secondary sources (targets) helps achieving optimal excitation
for all elements.
• The advantage of the indirect configuration is that the scattered radiation from
the tube cannot reach the detector. This results in a very low background level
(Figure 238) enabling the detection of weak peaks, i.e. low concentrations.
• The disadvantage is that with this extra step in the excitation process, excitation
energy and power is lost.
• The indirect configuration is also called a 3D geometry, because the X-ray path is
in two perpendicular planes, i.e. it is three-dimensional.

Direct (2D) Indirect (3D)

Figure 238. Comparison of XRF spectra obtained from a soil sample with
EDXRF having direct and indirect excitation. 374
8.4.2. WDXRF
• The first part of a WDXRF spectrometer is equivalent to an EDXRF spectrometer
with 2D optics and without a secondary target, but the detection system is
different.
• In WDXRF, the detection system is a set of collimators, a diffraction crystal and a
detector (Figure 239).
• The X-rays coming from the sample fall on the crystal, which diffracts X-rays with
different wavelengths to different directions (angles). The detector then measures
the intensities at different angles.
• The detector may be mounted on a goniometer which moves it through the
angular range to cover the whole wavelength range. XRF spectrometers with
goniometers are called sequential spectrometers.
• Another possibility is that there are multiple detectors that only measure a
specific wavelength (i.e. specific element). XRF spectrometers with such fixed
detection arrays are called simultaneous spectrometers.
• Combined systems with both moving and fixed detectors are also manufactured.

375
Figure 239. Construction of a sequential WDXRF spectrometer with 2D optics
and direct excitation. The diffracting crystals can be changed to reach
different wavelength ranges

376
8.4.3. COMPARISON OF EDXRF AND WDXRF SPECTROMETERS
• Both EDXRF and WDXRF have their advantages and disadvantages (Figure 240)
affecting their usefulness in applications.

Figure 240. Comparison of the properties of EDXRF and WDXRF devices.

377
8.4.4. X-RAY TUBES
• Most conventional WDXRF spectrometers use a high-power (2 – 4 kW) X-ray
tube.
• EDXRF spectrometers use either a high-power tube for indirect construction or
low-power (0.5 – 1.0 kW) tube direct construction.
• Typical anode materials have high emission energies so as to be able to excite
some lines from all elements:
– Mo (17.5 keV for K emission)
– Rh (20.2 keV)
– Ag (22.2 keV)
– W (59 keV)
– Au (68.8 keV)
• To achieve the strongest fluorescence from an element, the excitation energy
should be just above the binding energy of the studied element. Thus, the
characteristic lines excite best the heavier elements while for the lighter elements
bremsstrahlung gives the best excitation.
• Sometimes the emission of heavier elements in a sample makes the detection of
emission lines from the lighter elements difficult. This problem can be solved by
making the acceleration voltage of the X-ray tube small enough not to produce
any characteristic lines but only bremsstrahlung (see chapter 2.1.). 378
8.4.5. SECONDARY TARGETS
• The secondary targets are classified to three types: fluorescent, Barkla and Bragg.
• Fluorescent targets use the X-ray fluorescence of the elements in the target to
excite the sample. These targets also scatter radiation from the tube, but the
fluorescence dominates.
• Spectrometers can be equipped with a set of fluorescence targets and optimum
fluorescence is achived by selecting the correct target. Fluorescence targets are
usually made of heavier elements.
• Barkla targets use scattered tube radiation to excite the sample. These targets
also fluoresce but the energy and intensity of these lines is too low to excite the
sample. Barkla targets are made of light materials like Al2O3 or B4C, because
these give the strongest scattering.
• Barkla targets are used to obtain polarized excitation, which improves the
sensitivity for detection. Generally, they are used to measure heavier elements
(usually with absorption edges above 8 eV).
• Bragg targets are crystals that reflect only one specific energy in a certain
direction. Thus, it is possible to select only a single tube line with no other
radiation as the excitation source.
379
• Bragg targets can also be used for polarized excitation.
8.4.6. DETECTORS
• EDXRF mainly uses a solid state detector. This is a wide-range detector that
measures all elements from Na to U.
• WDXRF uses gas-filled detectors to measure from Be to Cu and scintillation
detectors from Cu to U.
• These detectors produce an electric pulse when an X-ray photon enters the
detector and the intensity of this pulse is proportional to the energy of the
incoming photon.
• The pulses are amplified and then counted by a multichannel analyzer (MCA).
• The basic designs for these detector types are shown in figure 241.
• A simple comparison of the properties of these detector types was made in
chapter 2.4

380
 Solid state

Gas-filled

Scintillation

Figure 241. XRF detector types. 381

8.4.7. FILTERS
• Filters are used not only to suppress bremsstrahlung (see chapter 8.3.3.) but also
emission lines of the X-ray tube (Figure 242). Thus, the filters improve the signal-
to-noise ratio.

Figure 242. Effect of filtering. Left: getting rid of X-ray tube’s (Rh) lines and
background. Right: reduction of background.
• Commonly used filter materials are aluminum and brass (CoZn alloy) with a
thickness between 100 and 1000 m.
• If the X-ray emission intensity from the sample is too high for the detector, a filter
can be used to absorb a part of the exciting radiation to prevent the detector
382
from saturating.
8.5. XRF analysis
8.5.1. SAMPLE PREPARATION
• Most XRF spectrometers are designed to measure samples that are circular disks
with a radius between 5 and 50 mm.
• The sample is placed in a cup and the cup is placed in the spectrometers.
• Special supporting films (Figure 243) allow the measurement of loose powders
and liquids.
• Some devices also allow the measurements of solid objects without a cup.

Figure 243. XRF sample cups and films. 383

Solids
• Solids require only a minimum amount of preparartion. Often, cleaning and
polishing is enough.
• Metals oxidize when exposed to air so they are often polished to remove the
oxide from the surface.

Powders

• Powders can be placed on a supporting film and measured directly.

• Another technique is to press them into a tablet. Sometimes a binding materials
needs to be added to improve the quality of the tablet.

Beads
• Powders together with a binding flux additive (e.g. Li2B4O7) can also be melted
into a glass sample called a bead. This bead can be measured directly.
• The sample may decompose and parts of it may evaporate because of the
heating.
• The possible decomposition as well as the elements of the flux must be taken into
account in the analysis of the XRF results from beads.
384
Liquids
• Liquids are poured in to special cups with supporting films.
• Liquids cannot be measured in vacuum and if it is considered that air absorbs too
much of the X-rays, a He gas sphere can be used.
Materials on filters
• Filters used to filter air or liquid can be analyzed with XRF, but the filter contains
only a small amount of the studied material.
• No specific preparation is needed and the sample can be measured directly.

8.5.2. MEASUREMENT
• XRF is a very sensitive technique and thus samples must be clean. Even
fingerprints on a sample can affect the results.
• For accurate results, the spectrometer is tuned (tube voltage & detector settings
etc.) to the elements to be analyzed. Wrong results could be obtained otherwise.
• To achieve the maximum intensity for a studied line, the excitation energy should
be just above that of the analytical line. This can be achieved by adjusting the
tube voltage accordingly.
385
8.5.3. QUALITATIVE ANALYSIS
• Qualitative analysis in XRF is primarily based on the determination of emission
line positions and a search from a database with emission energies for all
elements.
• This involves the procedures of peak search and peak match:
– Peak search uses a mathematical technique to find the peaks in a spectrum.
– Peak match determines the elements to which the peak profiles belong.
• Line intensities are also needed, because all elements have multiple X-ray
emission lines, whose intensity ratios are constant. For example the K/K
emission intensity ratio is always constant at 10. Thus, if the K line is observed,
there should also be the K line present and its intensity should be 10 % of the K.
• In EDXRF the intensity is defined as the peak area (integral intensity), whereas in
WDXRF it is defined as the height of the peak with backgrond subtracted.
• Deconvolution is used for the determination of intensities of individual lines.
Deconvolution also involves the fitting of background, because the backround
level influences the intensity strongly.
• In deconvolution:
– The backround is modelled and stripped out.
– The measured spectrum is fitted to theoretical line profiles by least squares fitting until the sum
of all the profiles (lines) gives the best fit with the measured spectrum.
 Figure 244. Deconvolution of two partly overlapping lines: background
modelling and profile fit.

• There may be a lot of emission lines in one sample, which makes the fit
complicated. However, from theory, it is possible to calculate the ratios between
groups of line intensities (e.g. the K/K intensity ratio is always 10). This reduces
the number of free parameters in a deconvolution making the fit easier.

387
8.5.4. QUANTITATIVE ANALYSIS
• In quantitative analysis, the net intensities are converted into concentrations.
• The usual procedure is to calibrate the spectrometer by measuring one or
more reference materials. The calibration determines the relationship between
the concentrations of the elements and the intensity of the fluorescence lines of
those elements.
• Ideally, the intensity of an analytical line is linearly proportional to the
concentration of the analyte.
• Across a limited range, the curve describing the relation between intensity (R)
and concentration (C) can be approximated to a straight line given by (Eq. 8.1):
C = D + E*R (8.1)
where D and E are constants determined by linear reggression.
• A better fit and wider range could be obtained by adding more terms, e.g.:
C = D + E*R + F*R2 (8.2)
where D, E and F are constants obtained by parabolic fitting.
• Once the values of D, E (and F) are known, concentration C for the studied
element can be calculated based on the observed line intensity R.
• Regardless of the model used for the fitting, the concentration range is still
limited (Figure 245) and the calibration curves can only be used for samples that
are similar to the standards (references) used.

Figure 245. Examples of calibration with linear and parabolic fitting.

Matrix effects and correction for them

• In addition to the concentration of the studied element, the line intensity
depends also on the presence and concentrations of other elements, because
these other elements can attenuate (due to absorption) or enhance (due to
secondary emission) the intensity of an analytical line.
• Matrix correction models use an additional term M (matrix correction factor) to
correct for these effects:
C = D + E*R*M (8.3)
• There are many ways to define the matrix correction factor M, e.g.
– Fundamental parameter (FP) model, which is based on the physics of the X-rays. This is also
called the Sherman method.
– Compton model is an empirical model taking into account Compton scattering.
– De Jongh model, which will be used as an example below.
• The De Jongh model uses the following equation:

(8.4)

where the  are numeric values that indicate how much element j attenuates or
enhances the intensity of analyte i.
- Because the sum of all concentrations is in a specimen equals 1, one element can be eliminated,
indicated by j  e. The element being eliminated is often the major component like Fe in steel.
- The values of the  are calculated by theory using the FP model. It is also possible to determine
the  values experimentally by measuring a set of different standards to allow the
determination of , D and E.
- If  are determined by the FP model, at least two standards are required to determine D and E.

• If the matrix effects are taken into account, the dependence of line intensity on
concentration can be made rather linear (Figure 246), which makes quantitative
analysis easier and more reliable.
No matrix correction

With matrix correction

Figure 246. Effect of
matrix correction on the
calibration curve of Ni in
steel.

391
Drift correction
• The stability and reproducibility of spectrometers is high, but long-term intensity
drift is unavoidable.
• The detector and X-ray tube degrade over time and the response of the other
components can also change with time.
• Because of this drift, calibration is not valid after a certain period of time.
• The setup must be either recalibrated or one can measure special monitor
samples whose relative change in intensity is applied to the samples being
analyzed.
Thin and layered samples
• For thick samples, the intensity corrected for matrix effects is linearly
proportional to the concentration and does not depend on the thickness of the
sample. This is because the sample is so thick that a portion of the X-rays does
not escape the sample because of absorption. Once the sample is thicker than
this analytical depth, changes in sample thickness do not affect emission line
intensities.
• For thin samples, radiation can reach the detector even from the bottom of the
sample. Thus, a small change in thickness will cause a noticeable difference in
emission intensity. 392
• The Compton correction model will correct for such finite thickness. This is
because the Compton line and the analytical line have approximately the same
thickness dependency. By dividing the intensity of the analytical line by the
intensity of the Compton line the thickness dependency vanishes:
– The matrix correction factor M for Eq. 8.3. equals Rline/RCompton.
• If the thickness of the sample is known, it can be taken into account in the FP
calculations, which then makes it possible to determine the concentration of the
analyte even from a thin sample.

8.5.5. STANDARDLESS ANALYSIS

• Calibration can be made independent of the compounds studied by following e.g.
these steps:
– The standards are enterd as elements and/or compounds but all concentrations are converted to
element concentrations by using the chemical formulae of the compound and the atomic
weights of the elements.
– The calibration is done per element, so the concentrations of the elements are calibrated against
their intensities.
– The concentration of the elements is determined for an unknown sample.
– The element concentrations are converted to compound concentration using the chemical
sample together with its chemical formula and the atomic weights of the elements.
• All types of materials can then be used as long as they contain the elements of
interest.
• Because the standards can be chosen independently of the sample to be
analyzed, they are called reference samples rather than standards.
• The qualitative and quantitative analysis can be made fully automated making it
possible to analyze all kinds of materials with a single calibration.
• Although material-specific calibration gives highest precision and accuracy,
standardless analysis is also well suited for all types of materials (Figure 247).

Figure 247. Comparison of quantitative analysis with conventional calibration and

394
standardless analysis.
LITERATURE
Brandt, C.G. and Kinneging, A.J., A Practical Guide to Quantitative Phase Analysis, PANalytical B.V., Almelo,
The Netherlands, 2005.
Brouwer, P., Theory of XRF – Getting Acquainted with the Principles, PANalytical B.V., Almelo, The
Netherlands, 2003.
Chateigner, D., QTA Internet Course, 2001, http://qta.ensicaen.ismra.fr/.
Chung, F.C. and Smith, D.K. (eds.), Industrial Applications of X-ray Diffraction, Marcel Dekker, New York NY,
USA, 2000.
Cullity, B.D. and Stock, S.R., Elements of X-Ray Diffraction, 3rd ed., Prentice Hall, Upper Saddle River NJ,
USA, 2001.
Dykhne, T., Taylor, R., Florence, A. and Billinge, S.J.L., Data Requirements for the Reliable Use of Atomic Pair
Distribution Functions in Amorphous Pharmaceutical Fingerprinting, Pharm. Res. 28 (2011) 1041-1048.
Engler, O. and Randle, V., Introduction to Texture Analysis, 2nd ed., CRC, Boca Raton FL, USA, 2010.
Fuoss, P.H. and Brennan, S., Surface Sensitive X-Ray Scattering, Ann. Rev. Mater. Sci. 20 (1990) 365-390.
Gateshki, M., te Nijenhuis, H. and Fransen, M., Structural Characterization of Nanocrystalline and
Amorphous Materials Using High Energy X-ray Diffraction on a Laboratory System: Pair Distribution
Function Analysis, Panalytical, 2009.
Glatter, O. and Kratky, O. (Eds.), Small-Angle X-Ray Scattering, Academic Press, London, Great Britain,
1982.
Hodeau, J.-L., Favre-Nicolin, V., Bos, S., Renevier, H., Lorenzo, E., and Berar, J.-F., Resonant Diffraction,
Chem. Rev. 101 (2001) 1843-1867.
Jalilehvand, F., Structure of Hydrated Ions and Cyanide Complexes by X-Ray Absorption Spectroscopy, PhD
Thesis, Royal Institute of Technology, Stockholm, Sweden, 2000.
Jenkins, R., X-Ray Fluorescence Spectrometry, 2nd ed., Wiley, New York NY, USA, 1999.
Klug, H.P. and Alexander, L.E., X-Ray Diffraction Procedures for Polycrystalline and Amorphous Materials,
Wiley, New York NY, USA, 1959.
Noyan, I.C. and Cohen, J.B., Residual Stress, Measurement by Diffraction and Interpretation, Springer,
Berlin, Germany, 1987.
Penner-Hahn, J.E., X-Ray Absorption Spectroscopy, in Comprehensive Coordination Chemistry II, Vol. 2, pp.
159-186.
Pielaszek, R., FW1/5/4/5M Method for Determination of the Grain Size Distribution from Powder
Diffraction Line Profile, J. Alloys Compd. 382 (2004) 128-132.
Proffen, T., Billinge, S.J.L., Egami, T., and Louca, D., Structural Analysis of Complex Materials Using the
Atomic Pair Distribution Function – A Practical Guide, Z. Kristallogr. 218 (2003) 132-143.
Scardi, P., Leoni, M., and Delhez, R., Line Broadening Analysis Using Integral Breadth Methods: A Critical
Review, J. Appl. Crystallogr. 37 (2004) 381-390.
Thompson, A.C., Attwood, D.T., Gullikson, E.M., Howells, M.R., Kortright, J.B., Robinson, A.L., Underwood,
J.H., Kim, K.-J., Kirz, J., Lindau, I., Pianetta, P., Winick, H., Williams, G.P., and Scofield, J.H., X-Ray Data
Booklet, 2nd Ed., Center for X-Ray Optics and Advanced Light Source, Lawrence Berkeley National
Laboratory, Berkeley CA, USA, 2001.
Withers, P.J. and Bhadeshia, H.K.D.H., Overview, Residual Stress, Part 1 – Measurement Techniques,
Mater. Sci. Technol. 17 (2001) 355-365.
Withers, P.J. and Bhadeshia, H.K.D.H., Overview, Residual Stress, Part 2 – Nature and Origins, Mater. Sci.
Technol. 17 (2001) 366-375.
Young, R.A. (ed.), The Rietveld Method, IUCr Monographs on Crystallography 5, Oxford University Press,
Guildford, Great Britain, 1993. 396