Professional Documents
Culture Documents
) 56
VI. Fluorescence Yields and CosterKronig Transition Probabilities
VII. Coefcients for Calculating the Photoelectric Absorption
Cross Sections t (Barns=Atom) Via lnln Representation
VIII. Coefcients for Calculating the Incoherent Collision
Cross Sections s
c
(Barns=Atom) Via the lnln Representation
IX. Coefcients for Calculating the Coherent Scattering
Cross Sections s
R
(Barns=Atom) Via the lnln Representation
X. Parameters for Calculating the Total Mass Attenuation
Coefcients in the Energy Range 0.11000 keV [Via Eq. (78)]
XI. Total Mass Attenuation Coefcients for Low-Energy Ka Lines
XII. Correspondence Between Old Siegbahn and New IUPAC
Notation X-ray Diagram Lines
References
Copyright 2002 Marcel Dekker, Inc.
2 Wavelength-Dispersive X-ray Fluorescence
Jozef A. Helsen and Andrzej Kuczumow
I. Introduction
II. Fundamentals of Wavelength Dispersion
III. Layout of a Spectrometer
IV. Qualitative and Quantitative Analysis
V. Chemical Shift and Speciation
VI. Instrumentation
VII. Future Prospects
References
3 Energy-Dispersive X-ray Fluorescence Analysis Using X-ray Tube Excitation
Andrew T. Ellis
I. Introduction
II. X-ray Tube Excitation Systems
III. Semiconductor Detectors
IV. Semiconductor Detector Electronics
V. Summary
References
4 Spectrum Evaluation
Piet Van Espen
I. Introduction
II. Fundamental Aspects
III. Spectrum Processing Methods
IV. Continuum Estimation Methods
V. Simple Net Peak Area Determination
VI. Least-Squares Fitting Using Reference Spectra
VII. Least-Squares Fitting Using Analytical Functions
VIII. Methods Based on the Monte Carlo Technique
IX. The Least-Squares-Fitting Method
X. Computer Implementation of Various Algorithms
References
5 Quantication of Innitely Thick Specimens by XRF Analysis
Johan L. de Vries and Bruno A. R. Vrebos
I. Introduction
II. Correlation Between Count Rate and Specimen Composition
III. Factors Inuencing the Accuracy of the Intensity Measurement
IV. Calibration and Standard Specimens
V. Converting Intensities to Concentration
VI. Conclusion
References
Copyright 2002 Marcel Dekker, Inc.
6 Quantication in XRF Analysis of Intermediate-Thickness Samples
Andrzej A. Markowicz and Rene E. Van Grieken
I. Introduction
II. Emission-Transmission Method
III. Absorption Correction Methods Via Scattered Primary Radiation
IV. Quantitation for Intermediate-Thickness Granular Specimens
References
7 Radioisotope-Excited X-ray Analysis
Stanislaw Piorek
I. Introduction
II. Basic Equations
III. Radioisotope X-ray Sources and Detectors
IV. X-ray and g-ray Techniques
V. Factors Affecting the Overall Accuracy of XRF Analysis
VI. Applications
VII. Future of Radioisotope-Excited XRF Analysis
VIII. Conclusions
Appendix: List of Companies that Manufacture Radioisotope-Based
X-ray Analyzers and Systems
References
8 Synchrotron Radiation-Induced X-ray Emission
Keith W. Jones
I. Introduction
II. Properties of Synchrotron Radiation
III. Description of Synchrotron Facilities
IV. Apparatus for X-ray Microscopy
V. Continuum and Monochromatic Excitation
VI. Quantitation
VII. Sensitivities and Minimum Detection Limits
VIII. Beam-Induced Damage
IX. Applications of SRIXE
X. Tomography
XI. EXAFS and XANES
XII. Future Directions
References
9 Total Reection X-ray Fluorescence
Peter Kregsamer, Christina Streli, and Peter Wobrauschek
I. Introduction
II. Physical Principles
III. Instrumentation
IV. Chemical Analysis
V. Surface Analysis
VI. Thin Films and Depth Proles
VII. Synchrotron Radiation Excitation
Copyright 2002 Marcel Dekker, Inc.
VIII. Light Elements
IX. Related Techniques
References
10 Polarized Beam X-ray Fluorescence Analysis
Joachim Heckel and Richard W. Ryon
I. Introduction
II. Theory
III. Barkla Systems
IV. Bragg Systems
V. Barkla-Bragg Combination Systems
VI. Secondary Targets
VII. Conclusion
References
11 Microbeam XRF
Anders Rindby and Koen H. A. Janssens
I. Introduction and Historical Perspective
II. Theoretical Background
III. Instrumentation for Microbeam XRF
IV. Collection and Processing of m-XRF Data
V. Applications
References
12 Particle-Induced X-ray Emission Analysis
Willy Maenhaut and Klas G. Malmqvist
I. Introduction
II. Interactions of Charged Particles with Matter,
Characteristic X-ray Production, and
Continuous Photon Background Production
III. Instrumentation
IV. Quantitation, Detection Limits, Accuracy, and Precision
V. Sample Collection and Sample and Specimen Preparation
for PIXE Analysis
VI. Applications
VII. Complementary Ion-Beam-Analysis Techniques
VIII. Conclusions
References
13 Electron-Induced X-ray Emission
John A. Small, Dale E. Newbury, and John T. Armstrong
I. Introduction
II. Quantitative Analysis
III. Microanalysis at Low Electron Beam Energy
IV. Analysis of Samples with Nonstandard Geometries
V. Spatially Resolved X-ray Analysis
References
Copyright 2002 Marcel Dekker, Inc.
14 Sample Preparation for X-ray Fluorescence
Martina Schmeling and Rene E. Van Grieken
I. Introduction
II. Solid Samples
III. Fused Specimen
IV. Liquid Specimen
V. Biological Samples
VI. Atmospheric Particles
VII. Sample Support Materials
References
Copyright 2002 Marcel Dekker, Inc.
Contributors
John T. Armstrong, Ph.D. National Institute of Standards and Technology,
Gaithersburg, Maryland
Johan L. de Vries, Ph.D.* Eindhoven, The Netherlands
Andrew T. Ellis, Ph.D. Oxford Instruments Analytical Ltd., High Wycombe,
Buckinghamshire, England
Joachim Heckel, Ph.D. Spectro Analytical Instruments, GmbH & Co. KG, Kleve, Ger-
many
Jozef A. Helsen, Ph.D. Catholic University of Leuven, Leuven, Belgium
Koen H. A. Janssens, Ph.D. University of Antwerp, Antwerp, Belgium
Keith W. Jones, Ph.D. Brookhaven National Laboratory, Upton, New York
Peter Kregsamer, Dr. techn., Dipl. Ing. Atominstitut, Vienna, Austria
Andrzej Kuczumow, Ph.D. Lublin Catholic University, Lublin, Poland
Willy Maenhaut, Ph.D. Ghent University, Ghent, Belgium
Klas G. Malmqvist, Ph.D. Lund University and Lund Institute of Technology, Lund,
Sweden
Andrzej A. Markowicz, Ph.D. Vienna, Austria
Dale E. Newbury, Ph.D. National Institute of Standards and Technology, Gaithersburg,
Maryland
Copyright 2002 Marcel Dekker, Inc.
Stanislaw Piorek, Ph.D.
{
Niton Corporation, Billerica, Massachusetts
Anders Rindby, Ph.D. Chalmers University of Technology and University of Go tebo rg,
Go tebo rg, Sweden
Richard W. Ryon, B.A. Lawrence Livermore National Laboratory, Livermore, Califor-
nia
Martina Schmeling, Ph.D. Loyola University Chicago, Chicago, Illinois
John A. Small, Ph.D. National Institute of Standards and Technology, Gaithersburg,
Maryland
Christina Streli, Ph.D. Atominstitut, Vienna, Austria
Piet Van Espen, Ph.D. University of Antwerp, Antwerp, Belgium
Rene E. Van Grieken, Ph.D. University of Antwerp, Antwerp, Belgium
Bruno A. R. Vrebos, Dr. Ir. Philips Analytical, Almelo, The Netherlands
Peter Wobrauschek, Ph.D. Atominstitut, Vienna, Austria
Copyright 2002 Marcel Dekker, Inc.
1
X-ray Physics
Andrzej A. Markowicz
Vienna, Austria
I. INTRODUCTION
In this introductory chapter, the basic concepts and processes of x-ray physics that relate
to x-ray spectrometry are presented. Special emphasis is on the emission of the continuum
and characteristic x-rays as well as on the interactions of photons with matter. In the
latter, only major processes of the interactions are covered in detail, and the cross sections
for dierent types of interactions and the fundamental parameters for other processes
involved in the emission of the characteristic x-rays are given by the analytical expressions
and=or in a tabulated form. Basic equations for the intensity of the characteristic x-rays
for the dierent modes of x-ray spectrometry are also presented (without derivation).
Detailed expressions relating the emitted intensity of the characteristic x-rays to the
concentration of the element in the specimen are discussed in the subsequent chapters of
this handbook dedicated to specic modes of x-ray spectrometry.
II. HISTORY
X-rays were discovered in 1895 by Wilhelm Conrad Ro ntgen at the University of
Wu rzburg, Bavaria. He noticed that some crystals of barium platinocyanide, near a dis-
charge tube completely enclosed in black paper, became luminescent when the discharge
occurred. By examining the shadows cast by the rays. Ro ntgen traced the origin of the rays
to the walls of the discharge tube. In 1896, Campbell-Swinton introduced a denite target
(platinum) for the cathode rays to hit; this target was called the anticathode.
For his work x-rays, Ro ntgen received the rst Nobel Prize in physics, in 1901. It was
the rst of six to be awarded in the eld of x-rays by 1927.
The obvious similarities with light led to the crucial tests of established wave optics:
polarization, diraction, reection, and refraction. With limited experimental facilities,
Ro ntgen and his contemporaries could nd no evidence of any of these; hence, the des-
ignation x (unknown) of the rays, generated by the stoppage at anode targets of the
cathode rays, identied by Thomson in 1897 as electrons.
The nature of x-rays was the subject of much controversy. In 1906, Barkla found
evidence in scattering experiments that x-rays could be polarized and must therefore by
waves, but W. H. Braggs studies of the produced ionization indicated that they were
Copyright 2002 Marcel Dekker, Inc.
corpuscular. The essential wave nature of x-rays was established in 1912 by Laue,
Friedrich, and Knipping, who showed that x-rays could be diracted by a crystal (copper
sulfate pentahydrate) that acted as a three-dimensional diraction grating. W. H. Bragg
and W. L. Bragg (father and son) found the law for the selective reection of x-rays. In
1908, Barkla and Sadler deduced, by scattering experiments, that x-rays contained com-
ponents characteristic of the material of the target; they called these components K and L
radiations. That these radiations had sharply dened wavelengths was shown by the
diraction experiments of W. H. Bragg in 1913. These experiments demonstrated clearly
the existence of a line spectrum superimposed upon a continuous (White) spectrum. In
1913, Moseley showed that the wavelengths of the lines were characteristic of the element
of the which the target was made and, further, showed that they had the same sequence as
the atomic numbers, thus enabling atomic numbers to be determined unambiguously for
the rst time. The characteristic K absorption was rst observed by de Broglie and in-
terpreted by W. L. Bragg and Siegbahn. The eect on x-ray absorption spectra of the
chemical state of the absorber was observed by Bergengren in 1920. The inuence of the
chemical state of the emitter on x-ray emission spectra was observed by Lindh and
Lundquist in 1924. The theory of x-ray spectra was worked out by Sommerfeld and
others. In 1919, Stenstro m found the deviations from Braggs law and interpreted them as
the eect of refraction. The anomalous dispersion of x-ray was discovered by Larsson in
1929, and the extended ne structure of x-ray absorption spectra was qualitatively in-
terpreted by Kronig in 1932.
Soon after the rst primary spectra excited by electron beams in an x-ray tube were
observed, it was found that secondary uorescent x-rays were excited in any material ir-
radiated with beams of primary x-rays and that the spectra of these uorescent x-rays were
identical in wavelengths and relative intensities with those excited when the specimen was
bombarded with electrons. Beginning in 1932, Hevesy, Coster, and others investigated in
detail the possibilities of uorescent x-ray spectroscopy as a means of qualitative and
quantitative elemental analysis.
III. GENERAL FEATURES
X-rays, or Ro ntgen rays, are electromagnetic radiations having wavelengths roughly
within the range from 0.005 to 10 nm. At the short-wavelength end, they overlap with
g-rays, and at the long-wavelength end, they approach ultraviolet radiation.
The properties of x-rays, some of which are discussed in detail in this chapter, are
summarized as follows:
Invisible
Propagated in straight lines with a velocity of 3610
8
m=s, as is light
Unaected by electrical and magnetic elds
Dierentially absorbed while passing through matter of varying composition,
density, or thickness
Reected, diracted, refracted, and polarized
Capable of ionizing gases
Capable of aecting electrical properties of liquids and solids
Capable of blackening a photographic plate
Able to liberate photoelectrons and recoil electrons
Capable of producing biological reactions (e.g., to damage or kill living cells and to
produce genetic mutations)
Copyright 2002 Marcel Dekker, Inc.
Emitted in a continuous spectrum whose short-wavelength limit is determined only
by the voltage on the tube
Emitted also with a line spectrum characteristic of the chemical elements
Found to have absorption spectra characteristic of the chemical elements
IV. EMISSIONOF CONTINUOUS RADIATION
Continuous x-rays are produced when electrons, or other high-energy charged particles,
such as protons or a-particles, lose energy in passing through the Coulomb eld of a
nucleus. In this interaction, the radiant energy (photons) lost by the electron is called
bremsstrahlung (from the German bremsen, to brake, and Strahlung, radiation; this term
sometimes designates the interaction itself). The emission of continuous x-rays nds a
simple explanation in terms of classic electromagnetic theory, because, according to this
theory, the acceleration of charged particles should be accompanied by the emission of
radiation. In the case of high-energy electrons striking a target, they must be rapidly
decelerated as they penetrate the material of the target, and such a high negative accel-
eration should produce a pulse of radiation.
The continuous x-ray spectrum generated by electrons in an x-ray tube is char-
acterized by a short-wavelength limit l
min
, corresponding to the maximum energy of the
exciting electrons:
l
min
=
hc
eV
0
(1)
where h is Plancks constant, c is the velocity of light, e is the electron charge, and V
0
is the
potential dierence applied to the tube. This relation of the short-wavelength limit to the
applied potential is called the DuaneHunt law.
The probability of radiative energy loss (bremsstrahlung) is roughly proportional to
q
2
Z
2
T=M
2
0
, where q is the particle charge in units of the electron charge e, Z is the atomic
number of the target material, T is the particle kinetic energy, and M
0
is the rest mass of
the particle. Because protons and heavier particles have large masses compared to the
electron mass, they radiate relatively little; for example, the intensity of continuous x-rays
generated by protons is about four orders of magnitude lower than that generated by
electrons.
The ratio of energy lost by bremsstrahlung to that lost by ionization can be
approximated by
m
0
M
0
2
ZT
1600m
0
c
2
(2)
where m
0
the rest mass of the electron.
A. Spectral Distribution
The continuous x-ray spectrum generated by electrons in an x-ray tube (thick target) is
characterized by the following features:
1. Short-wavelength limit, l
min
[Eq. (1)]; below this wavelength, no radiation is
observed.
Copyright 2002 Marcel Dekker, Inc.
2. Wavelength of maximum intensity l
max
, approximately 1.5 times l
min
; however,
the relationship between l
max
and l
min
depends to some extent on voltage,
voltage waveform, and atomic number.
3. Total intensity nearly proportional to the square of the voltage and the rst
power of the atomic number of the target material.
The most complete empirical work on the overall shape of the energy distribution
curve for a thick target has been of Kulenkamp (1922, 1933), who found the following
formula for the energy distribution;
I(v) dv = i aZ v
0
v ( ) bZ
2
dv (3)
where I(n) dn is the intensity of the continuous x-rays within a frequency range
(n; n dv); i is the electron current striking the target, Z is the atomic number of the
target material, n
0
is the cuto frequency (=c=l
min
) above which the intensity is zero, and
a and b are constants independent of atomic number, voltage, and cuto wavelength. The
second term in Eq. (3) is usually small compared to the rst and is often neglected.
The total integrated intensity at all frequencies is
I = i(a
/
ZV
2
0
b
/
Z
2
V
0
) (4)
in which a
/
= a(e
2
=h
2
)=2 and b
/
= b(e=h). An approximate value for b
/
=a
/
is 16.3 V; thus,
I = a
/
iZV
0
(V
0
16:3Z) (5)
The eciency E of conversion of electric power input to x-rays of all frequencies is
given by
Eff =
I
V
0
i
= a
/
Z(V
0
16:3Z) (6)
where V
0
is in volts. Experiments give a
/
= (1:2 0:1) 10
9
(Condon, 1958).
The most complete and successful eorts to apply quantum theory to explain all
features of the continuous x-ray spectrum are those of Kramers (1923) and Wentzel
(1924). By using the correspondence principle, Kramers found the following formulas for
the energy distribution of the continuous x-rays generated in a thin target:
I(v) dv =
16p
2
AZ
2
e
5
3
3
_
m
0
V
0
c
3
dv; v < v
0
I(v) dv = 0; v > v
0
(7)
where A is the atomic mass of the target material. When the decrease in velocity of the
electrons in a thick target was taken into account by applying the ThomsonWhiddington
law (Dyson, 1973), Kramers found, for a thick target,
I(v) dv =
8pe
2
h
3
3
_
lm
0
c
3
Z(v
0
v) dv (8)
where l is approximately 6. The eciency of production of the x-rays calculated via
Kramers law is given by
Eff = 9:2 10
10
ZV
0
(9)
which is in qualitative agreement with the experiments of Kulenkamp (Stephenson,
1957); for example,
Copyright 2002 Marcel Dekker, Inc.
Eff = 15 10
10
ZV
0
(10)
It is worth mentioning that the real continuous x-ray distribution is described only ap-
proximately by Kramers equation. This is related, inter alia, to the fact that the derivation
ignores the self-absorption of x-rays and electron backscattering eects.
Wentzel (1924) used a dierent type of correspondence principle than Kramers, and
he explained the spatial distribution asymmetry of the continuous x-rays from thin targets.
An accurate description of continuous x-rays is crucial in all x-ray spectrometry
(XRS). The spectral intensity distributions from x-ray tubes are of great importance for
applying fundamental mathematical matrix correction procedures in quantitative x-ray
uorescence (XRF) analysis. A simple equation for the accurate description of the actual
continuum distributions from x-ray tubes was proposed by Tertian and Broll (1984). It is
based on a modied Kramers law and a rened x-ray absorption correction. Also, a strong
need to model the spectral Bremsstrahlung background exists in electron-probe x-ray
microanalysis (EPXMA). First, tting a function through the background portion, on
which the characteristic x-rays are superimposed in an EPXMA spectrum, is not easy;
several experimental tting routines and mathematical approaches, such as the Simplex
method, have been proposed in this context. Second, for bulk multielement specimens, the
theoretical prediction of the continuum Bremsstrahlung is not trivial; indeed, it has been
known for several years that the commonly used Kramers formula with Z directly sub-
stituted by the average
"
Z =
P
i
W
i
Z
i
(W
i
and Z
i
are the weight fraction and atomic number
of the ith element, respectively) can lead to signicant errors. In this context, some im-
provements are oered by several modied versions of Kramers formula developed for a
multielement bulk specimen (Statham, 1976; Lifshin, 1976; Sherry and Vander Sande, 1977;
Smith and Reed, 1981). Also, a new expression for the continuous x-rays emitted by thick
composite specimens was proposed (Markowicz and Van Grieken, 1984; Markowicz et al.,
1986); it was derived by introducing the compositional dependence of the continuum x-rays
already in the elementary equations. The new expression has been combined with known
equations for the self-absorption of x-rays (Ware and Reed, 1973) and electron back-
scattering (Statham, 1979) to obtain an accurate description of the detected continuum
radiation. A third problem is connected with the description of the x-ray continuum gen-
erated by electrons in specimens of thickness smaller than the continuum x-ray generation
range. This problem arises in the analysis of both thin lms and particles by EPXMA.
A theoretical model for the shape of the continuous x-rays generated in multielement
specimens of nite thickness was developed (Markowicz et al., 1985); both composition and
thickness dependence have been considered. Further renements of the theoretical approach
are hampered by the lack of knowledge concerning the shape of the electron interaction
volume, the distribution of the electron within the interaction volume, and the anisotropy of
continuous radiation for dierent x-ray energies and for dierent lm thickness.
B. Spatial Distribution and Polarization
The spatial distribution of the continuous x-rays emitted by thin targets has been in-
vestigated by Kulenkamp (1928). The author made an extensive survey of the intensity at
angles between 22
and 150
for 37.8 kV to 65
as the theory of Scherzer predicts, it can be seen from Figure 2 that for a thinner foil, a
lower intensity at 0
.
The continuous radiation from thick targets is characterized by a much smaller
anisotropy than that from thin targets. This is because in thick targets the electrons are
rarely stopped in one collision and usually their directions have considerable variation.
The use of electromagnetic theory predicts a maximum energy at right angles to the in-
cident electron beam at low voltages, with the maximum moving slightly away from
perpendicularity toward the direction of the elctron beam as the voltage is increased. In
general, an increase in the anisotropy of the continuous x-rays from thick targets is ob-
served at the short-wavelength limit and for low-Z targets (Dyson, 1973).
Figure1 Intensity of continuous x-rays as a function of direction for different voltages. (Curve (a)
is repeated as dotted line.) (From Stephenson, 1957.)
Copyright 2002 Marcel Dekker, Inc.
Continuous x-ray beams are partially polarized only from extremely thin targets; the
angular region of polarization is sharply peaked about the photon emission angle
y = m
0
c
2
=E
0
, where E
0
is the energy of the primary electron beam. Electron scattering in
the target broadens the peak and shifts the maximum to larger angles. Polarization is
dened by (Kenney, 1966)
P(y; E
0
; E
n
) =
dsl(y; E
0
; E
n
) ds|(y; E
0
; E
n
)
dsl(y; E
0
; E
n
) ds|(y; E
0
; E
n
)
(11)
where an electron of energy E
0
radiates a photon of energy E
n
at angle y; dsl(y; E
0
; E
n
)
and ds|(y; E
0
; E
n
) are the cross sections for generation of the continuous radiation with
the electric vector perpendicular (l) and parallel (|) to the plane dened by the incident
electron and the radiated photon, respectively. Polarization is dicult to observe, and only
thin, low-yield radiators give evidence for this eect. When the electron is relativistic
before and after the radiation, the electrical vector is most probably in the l direction.
Practical thick-target Bremsstrahlung shows no polarization eects whatever (Dyson,
1973; Stephenson, 1957; Kenney, 1966).
V. EMISSIONOF CHARACTERISTICX-RAYS
The production of characteristic x-rays involves transitions of the orbital electrons of
atoms in the target material between allowed orbits, or energy states, associated with
ionization of the inner atomic shells. When an electron is ejected from the K shell by
electron bombardment or by the absorption of a photon, the atom becomes ionized and
the ion is left in a high-energy state. The excess energy the ion has over the normal state of
the atom is equal to the energy (the binding energy) required to remove the K electron to a
state of rest outside the atom. If this electron vacancy is lled by an electron coming from
an L level, the transition is accompanied by the emission of an x-ray line known as the Ka
line. This process leaves a vacancy in the L shell. On the other hand, if the atom contains
sucient electrons, the K shell vacancy might be lled by an electron coming from an M
level that is accompanied by the emission of the Kb line. The L or M state ions that remain
may also give rise to emission if the electron vacancies are lled by electrons falling from
further orbits.
Figure 2 Intensity of continuous x-rays as a function of direction for different thicknesses of the
A1 target together with theoretical prediction. (From Stephenson, 1957.)
Copyright 2002 Marcel Dekker, Inc.
A. Inner Atomic Shell Ionization
As already mentioned, the emission of characteristic x-ray is preceded by ionization of
inner atomic shells, which can be accomplished either by charged particles (e.g., electrons,
protons, and a-particles) or by photons of sucient energy. The cross section for ion-
ization of an inner atomic shell of element i by electrons is given by (Bethe, 1930; Green
and Cosslett, 1961; Wernisch, 1985)
Q
i
= pe
4
n
s
b
s
ln U
UE
2
c;i
(12)
where U = E=E
c;i
is the overvoltage, dened as the ratio of the instantaneous energy of the
electron at each point of the trajectory to that required to ionize an atom of element i, E
c;i
is the critical excitation energy, and n
s
and b
s
are constants for a particular shell:
s = K: n
s
= 2; b
s
= 0:35
s = L: n
s
= 8; b
s
= 0:25
The cross section for ionization Q
i
is a strong function of the overvoltage, which shows a
maximum at U 34 (Heinrich, 1981; Goldstein et al., 1981).
The probability (or cross section) of ionization of an inner atomic shell by a charged
particle is given by (Merzbacher and Lewis, 1958)
s
s
=
8pr
2
0
q
2
f
s
Z
4
Z
s
(13)
where r
0
is the classic radius of the electron equal to 2.818610
15
m, q is the particle
charge, Z is the atomic number of the target material, f
s
is a factor depending on the wave
functions of the electrons for a particular shell, and Z
s
is a function of the energy of the
incident particles.
In the case of electromagnetic radiation (x or g), the ionization of an inner atomic
shell is a result of the photoelectric eect. This eect involves the disappearance of a ra-
diation photon and the photoelectric ejection of one electron from the absorbing atom,
leaving the atom in an excited level. The kinetic energy of the ejected photoelectron is
given by the dierence between the photon energy hn and the atomic binding energy of the
electron E
c
(critical excitation energy). Critical absorption wavelengths (Clark, 1963) re-
lated to the critical absorption energies (Burr, 1974) via the equation l(nm) =1.24=E(ke V)
are presented in Appendix I. The wavelenghts of K, L, M, and N absorption edges can also
be calculated by using simple empirical equations (Norrish and Tao, 1993).
For energies far from the absorption edge and in the nonrelativistic range, the cross
section t
K
for the ejection of an electron from the K shell is given by (Heitler, 1954)
t
K
=
32
2
_
3
pr
2
0
Z
5
(137)
4
m
0
c
2
hv
7=2
(14)
Equation (14) is not fully adequate in the neighborhood of an absorption edge; in this
case, Eq. (14) should be multiplied by a correction factor f(X) (Stobbe, 1930):
f(X) = 2p
D
hv
1=2
e
4X arccot X
1 e
2pX
(15)
where
Copyright 2002 Marcel Dekker, Inc.
X =
D
hv D
1=2
(15a)
with
D
1
2
(Z 0:3)
2
m
0
c
2
(137)
2
(15b)
When the energy of the incident photon is of the order m
0
c
2
or greater, relativistic
cross sections for the photoelectric eect must be used (Sauter, 1931).
B. Spectral Series in X-rays
The energy of an emission line can be calculated as the dierence between two terms, each
term corresponding to a denite state of the atom. If E
1
and E
2
are the term values re-
presenting the energies of the corresponding levels, the frequency of an x-ray line is given
by the relation
v =
E
1
E
2
h
(16)
Using the common notations, one can represent the energies of the levels E by means
of the atomic number and the quantum numbers n, l, s, and j (Sandstro m, 1957):
E
Rh
=
(Z S
n;l
)
2
n
2
a
2
(Z d
n;l; j
)
2
n
3
1
l
1
2
3
4n
!
a
2
(Z d
n;l; j
)
4
n
3
j( j 1) l(l 1) s(s 1)
2l(l
1
2
)(l 1)
(17)
where S
n;l
and d
n;l; j
are screening constants that must be introduced to correct for the eect
of the electrons on the eld in the atom, R is the universal Rydberg constant valid for all
elements with Z>5 or throughout nearly the whole x-ray region, and a is the ne-
structure constant given by
a =
2pe
2
hc
(17a)
The theory of x-ray spectra reveals the existence of a limited number of allowed
transitions; the rest are forbidden. The most intense lines create the electric dipole ra-
diation. The transitions are governed by the selection rules for the change of quantum
numbers:
Dl = 1; Dj = 0 or 1 (18)
The j transition 0 ?0 is forbidden.
According to Diracs theory of radiation (Dirac, 1947), transitions that are forbidden
as dipole radiation can appear as multipole radiation (e.g., as electric quadrupole and
magnetic dipole transitions). The selection rules for the former are
Dl = 0 or 2; Dj = 0; 1; or 2 (19)
The j transitions 0 ?0,
1
2
?
1
2
, and 0 1 are forbidden.
The selection rules for magnetic dipole transitions are
Copyright 2002 Marcel Dekker, Inc.
Dl = 0; Dj = 0 or 1 (20)
The j transition 0 ?0 is forbidden.
The commonly used terminology of energy levels and x-ray lines is shown in Figure 3.
A general expression relating the wavelength of an x-ray characteristic line with the
atomic number of the corresponding element is given by Moseleys law (Moseley, 1914):
1
l
= k(Z s)
2
(21)
where k is a constant for a particular spectral series and s is a screening constant for the
repulsion correction due to other electrons in the atom. Moseleys law plays an important
role in the systematizing of x-ray spectra. Appendix II tabulates the energies and wave-
lengths of the principal x-ray emission lines for the K, L, and M series with their ap-
proximate relative intensities, which can be dened either by means of spectral line peak
intensities or by area below their intensity distribution curve. In practice, the relative
Figure 3 Commonly used terminology of energy levels and x-ray lines. (From Sandstro m, 1957.)
Copyright 2002 Marcel Dekker, Inc.
intensities of spectral lines are not constant because they depend not only on the electron
transition probability but also on the specimen composition.
Considering the K series, the Ka fraction of the total K spectrum is dened by the
transition probability p
Ka
, which is given by (Schreiber and Wims, 1982)
p
Ka
=
I(Ka
1
Ka
2
)
I(Ka
1
Ka
2
) I(Kb
1
Kb
2
)
(22)
Wernisch (1985) proposed a useful expression for the calculation of the transition prob-
ability p
Ka
for dierent elements:
p
Ka;i
=
1:052 4:39 10
4
Z
2
i
; 11 _ Z
i
_ 19
0:896 6:575 10
4
Z
i
; 20 _ Z
i
_ 29
1:0366 6:82 10
3
Z
i
4:815 10
5
Z
2
i
; 30 _ Z
2
i
_ 60
8
<
:
(23)
For the L level, split into three subshells, several electron transitions exist. The transition
probability p
La
, dened as the fraction of the transitions resulting in La
1
and La
2
radiation
from the total of possible transitions into the L
3
subshell, can be calculated by the ex-
pression (Wernisch, 1985)
p
La;i
=
0:944; 39 _ Z
i
_ 44
4:461 10
1
5:493 10
2
Z
i
7:717 10
4
Z
2
i
3:525 10
6
Z
3
i
; 45 _ Z
i
_ 82
8
<
:
(24)
Radiative transition probabilities for various K and L x-ray lines (West, 198283) are
presented in detail in Appendix III. The experimental results, together with the estimated
95% condence limits, for the relative intensity ratios for K and L x-ray lines for selected
elements with atomic number from Z=14 to 92 have been reported by Stoev and Dlouhy
(1994). The values are in a good agreement with other published experimental data.
Because the electron vacancy created in an atom by charged particles or electro-
magnetic radiation has a certain lifetime, the atomic levels E
1
and E
2
[Eq. (16)] and the
characteristic x-ray lines are characterized by the so-called natural widths (Krause and
Oliver, 1979). The natural x-ray linewidths are the sums of the widths of the atomic levels
involved in the transitions. Semiempirical values of the natural widths of K, L
1
, L
2
and L
3
levels, Ka
1
and Ka
2
x-ray lines for the elements 10 _ Z _ 110 are presented in Appendix
IV. Uncertainties in the level width values are from 3% to 10% for the K shell and from
8% to 30% for the L subshell. Uncertainties in the natural x-ray linewidth values are from
3% to 10% for Ka
1;2
. In both cases, the largest uncertainties are for low-Z elements
(Krause and Oliver, 1979).
C. X-ray Satellites
A large number of x-ray lines have been reported that do not t into the normal energy-
level diagram (Clark, 1955; Kawai and Gohshi, 1986). Most of the x-ray lines, called
satellites or nondiagram lines, are very weak and are of rather little consequence in ana-
lytical x-ray spectrometry. By analogy to the satellites in optical spectra, it was supposed
that the origin of the nondiagram x-ray lines is double or manyfold ionization of an atom
through electron impact. Following the ionization, a multiple electron transition results in
emission of a single photon of energy higher than that of the characteristic x-rays. The
majority of nondiagram lines originate from the dipole-allowed deexcitation of multiply
Copyright 2002 Marcel Dekker, Inc.
ionized or excited states and are called multiple-ionization satellites. A line where the
initial state has two vacancies in the same shell, notably the K shell, is called a hypersa-
tellite. In practice, the most important nondiagram x-ray lines occur in the K
a
series; they
are denoted as the Ka
3
;a
4
doublet, and their origin is a double electron transition. The
probability of a multiple-electron transition resulting in the emission of satellite x-ray lines
is considerably higher for low-Z elements than for heavy and medium elements. For
instance, the intensity of the AlKa
3
satellite line is roughly 10% of that of the AlKa
1
; a
2
characteristic x-rays.
Appendix V tabulates wavelengths of the K satellite lines. A new class of satellites
that are inside the natural width of the parent lines was observed by Kawai and Gohshi
(1986). The origin of these satellites, called parasites or hidden satellites, is multiple
ionization in nonadjacent shells.
D. Soft X-ray Emission-Band Spectra
In the soft x-ray region, the characteristic emission spectra of solid elements include
continuous bands of width varying from 1 to 10 electron volts (eV); the same element in
vapor form produces only the usual sharp spectral lines. The bands occur only when an
electron falls from the outermost or valency shell of the atom, the levels of which are
broadened into a wide band when the atoms are packed in a crystal lattice. Investigation
of the emission-band spectra is of great signicance in understanding the electronic
structure of solid metals, alloys, and complex coordination compounds.
E. Auger Effect
It has already been stated that the excess of energy an atom possesses after removing one
electron from an inner shell by whatever means may be emitted as characteristic radiation.
Alternatively, however, an excited atom may return to a state of lower energy by ejecting
one of its own electrons from a less tightly bound state. The radiationless transition is
called the Auger eect, and the ejected electrons are called Auger electrons (Auger, 1925;
Burhop, 1952). Generally, the probability of the Auger eect increases with a decrease in
the dierence of the corresponding energy states, and it is the highest for the low-Z
elements.
Because an excited atom already has one electron missing (e.g., in the K shell) and
another electron is ejected in an Auger process (e.g., from the L shell), the atom is left in a
doubly-ionized state in which two electrons are missing. This atom may return to its
normal state by single- or double-electron jumps with the emission of diagram or satellite
lines, respectively. Alternatively, another Auger process may occur in which a third
electron is ejected from the M shell.
The Auger eect also occurs after capture of a negative meson by an atom. As the
meson changes energy levels in approaching the nucleus, the energy released may be either
emitted as a photon or transferred directly to an electron that is emitted as a high-energy
Auger electron (in the keV range for hydrogen and the MeV range for heavy elements).
Measurements of the energy and intensity of the Auger electrons are applied ex-
tensively in surface physics studies (Auger electron spectroscopy).
F. FluorescenceYield
An important consequence of the Auger eect is that the actual number of x-ray photons
produced from an atom is less than expected, because a vacancy in a given shell might be
Copyright 2002 Marcel Dekker, Inc.
lled through a nonradiative transition. The probability that a vacancy in an atomic shell
or subshell is lled through a radiative transition is called the uorescence yield. The
application of this denition to the K shell of an atom is straightforward, and the uor-
escence yield of the K shell is
o
K
=
I
K
n
K
(25)
where I
K
is the total number of characteristic K x-ray photons emitted from a sample and
n
K
is the number of primary K shell vacancies.
The denition of the uorescence yield of higher atomic shells is more complicated,
for the following two reasons:
1. Shells above the K shell consist of more than one subshell; the average
uorescence yield depends on how the shells are ionized.
2. CosterKronig transitions occur, which are nonradiative transitions between the
subshells of an atomic shell having the same principal quantum number (Fink,
1974; Bambynek et al., 1972).
In case CosterKronig transitions are absent, the uorescence yield of the ith sub-
shell of a shell, whose principal quantum number is indicated by X(X = L; M; . . .), is
given as
o
X
i
=
I
X
i
n
X
i
(26)
An average or mean uorescence yield " o
X
for the shell X is dened as
" o
X
=
X
k
i=1
N
X
i
o
X
i
(27)
where N
X
i
is the relative number of primary vacancies in the subshell i of shell X:
N
X
i
=
n
X
i
P
k
i=1
n
X
i
;
X
k
i=1
N
X
i
= 1 (28)
The summations in Eqs. (27) and (28) extend over all k subshells of shell X. For the de-
nition of the average uorescence yield, the primary vacancy distribution must be xed;
that is, CosterKronig transitions must be absent. It is noteworthy that " o
X
generally is not
a fundamental property of the atom, but depends both on the atomic subshell uorescence
yields o
X
i
and on the relative number of primary vacancies N
X
i
characteristic of the method
used to ionize the atoms.
In the presence of CosterKronig transitions, which modify the primary vacancy
distribution by the transfer of ionization from one subshell with a given energy to a
subshell with less energy, the average uorescence yields can be calculated by using two
alternative approaches. In the rst, the average uorescence yield " o
X
is regarded as a
linear combination of the subshell uorescence yields o
X
i
with a vacancy distribution
modied by CosterKronig transitions:
" o
X
=
X
k
i=1
V
X
i
o
X
i
;
X
k
i=1
V
X
i
> 1 (29)
Copyright 2002 Marcel Dekker, Inc.
where V
X
i
is the relative number of vacancies in the subshell i of shell X, including
vacancies shifted to each subshell by CosterKronig transitions. The V
X
i
values can be
expressed in terms of the relative numbers N
X
i
of primary vacancies and the CosterKronig
transition probability for shifting a vacancy from a subshell X
i
to a higher subshell X
j
,
denoted as f
X
ij
(Bambynek et al., 1972):
V
X
1
= N
X
1
V
X
2
= N
X
2
f
X
12
N
X
1
V
X
3
= N
X
3
f
X
23
N
X
2
( f
X
13
f
X
12
f
X
23
)N
X
1
(30)
In an alternative approach, the mean uorescence yield " o
X
is a linear combination of
the relative numbers of primary vacancies N
X
i
:
" o
X
=
X
k
i=1
N
X
i
n
X
i
(31)
where n
X
i
represents the total number of characteristic x-rays that result per primary
vacancy in the X
i
subshell. The transformation relations between the coecients n
X
i
and
the subshell uorescence yields o
X
i
follow from Eqs. (29) through (31) and are given in
Fink (1974) and Bambynek et al. (1972).
Among the uorescence yield o
X
i
, the Auger yield a
X
i
, and the CosterKronig
transition probabilities f
X
ij
, the following relationship must hold (Krause, 1979):
o
X
i
a
X
i
X
k
i=1
f
X
ij
= 1 (32)
The mean Auger yield " a
X
is given by
" a
x
=
X
k
i=1
V
X
i
a
X
i
(33)
The values of the K, L, and M shell uorescence yields, the CosterKronig transition
probabilities, as well as the Auger yields are given in Appendix VI.
Although, in principle, the K shell uorescence yield o
k
can be calculated theoreti-
cally, experimental data are applied in practice. The following semiempirical equation, due
to Burhop (1952), gives values correct to a few percent between Z = 23 and Z = 57 and
less accurate values outside these limits:
o
K
1 o
K
1=4
= 6:4 10
2
3:40 10
2
Z 1:03 10
6
Z
3
(34)
The uorescence yield for the K series can also be calculated from a dierent equation:
o
K
1 o
K
1=4
= 0:217 0:03318Z 1:14 10
6
Z
3
(35)
which gives quite good agreement with the experimental values for almost all elements.
Based on a critical review of the available experimental and theoretical data, a recom-
mended set of o
K
; " o
L
; and " o
M
values together with the following analytical expressions
were given by Hubbell (1989) and Hubbell et al. (1994):
Copyright 2002 Marcel Dekker, Inc.
o
K
(1 _ Z _ 100) =
X
3
i=0
C
i
Z
i
" #
4
1
X
3
i=0
C
i
Z
i
" #
4
8
<
:
9
=
;
1
(36)
with
C
0
= 0:0370 0:0052
C
1
= 0:03112 0:00044
C
2
= (5:44 0:11) 10
5
C
3
= (1:25 0:07) 10
6
" o
L
(3 _ Z _ 36) = 1:939 10
8
Z
3:8874
" o
L
(37 _ Z _ 100) =
X
3
i=0
C
i
Z
i
" #
4
1
X
3
i=0
C
i
Z
i
" #
4
8
<
:
9
=
;
1
(37)
with
C
0
= 0:17765
C
1
= 0:00298937
C
2
= 8:91297 10
5
C
3
= 2:67184 10
7
" o
M
(13 _ Z _ 100) = 1:29 10
9
(Z 13)
4
(38)
Other useful expressions for the calculation of the uorescence yields
o
K
(12 _ Z _ 42) and o
L3
(38 _ Z _ 79) have been proposed by Hanke et al. (1985),
based on literature and experimental data:
o
K
= 3:3704 10
1
6:0047 10
2
Z
3:3133 10
3
Z
2
3:9215 10
5
Z
3
(39)
o
L3
= 4:41 10
2
4:7559 10
3
Z
1:1494 10
4
Z
2
1:8594 10
7
Z
3
(40)
For the N shell, the best source of " o
N
data is a theoretical work of McGuire (1974)
which provides o
N1
; o
N2
, and o
N3
values for 25 elements over the range 38 _ Z _ 103
and o
N4
; o
N5
, and o
N6;7
values for 20 elements over the range 50 _ Z _ 103.
The average uorescence yield " o
N
can be calculated from (Hubbell, 1989)
" o
N
=
X
7
i=1
1
32
N
N
i
o
N
i
(41)
where N
N
i
are the numbers of electrons in each N
i
subshell.
A comparison of the total x-ray yields for bulk samples (including both the prob-
ability of ionization and the uorescence yield) in terms of photons per steradian per
incident quantum for electrons, protons, and x-ray photons is shown in Figure 4.
Copyright 2002 Marcel Dekker, Inc.
G. Fine Features of X-ray Emission Spectra (Valence or Chemical Effects)
Because characteristic x-ray emission is a process in which the innermost electrons in the
atom are concerned, it is reasonable to suppose that the external, or valence, electrons
have little or no eect on the x-ray emission lines. However, this is not fully true for K lines
of low-Z elements and L or M lines of higher-Z elements, where the physical state and
chemical combination of the elements aect the characteristic x-rays (Clark, 1955). The
changes in ne features of x-ray emission spectra with chemical combination can be
classied into three groups: (1) shifting in wavelength (Kallithrakas-Kontos, 1996),
(2) distortion of line shape, and (3) intensity changes (Kawai et al., 1993; Rebohle et al.,
1996). Wavelength shifts to both longer and shorter wavelengths result from energy-level
changes due to electrical shielding or screening of the electrons when the valence electrons
are drawn into a bond. Generally, the so-called last or highest-energy member of a given
series is most aected by chemical combination; maximum energy shifts are of the order of
a few electron volts. Distortion of an x-ray emission line shape gives some indication of the
energy distribution of the electrons occupying positions in or near the valence shell. The
changes in the characteristic x-ray intensity are a result of alterations in excitation
Figure 4 Total x-ray yields for excitation by electrons, protons, and primary x-ray photons as a
function of energy of the exciting quantum. (From Birks, 1971a.)
Copyright 2002 Marcel Dekker, Inc.
probabilities of the electrons undergoing transitions. Certain x-ray lines or bands appear
or disappear with chemical combinations. In the case of the K series, the most noticeable
chemical eects on x-ray emission are seen in spectra from low-Z elements (4 _ Z _ 17).
The L series shows as large or even larger changes with chemical combination of the
elements than K series. The valence eects in L spectra have been observed for elements of
the rst transition series and others nearby in the periodic table.
Because the ne features of x-ray emission spectra may be applied to determine how
each element is chemically combined in the sample (speciation), the valence eects found
numerous applications in such elds as physics of solids and surface or near-surface
characterization.
VI. INTERACTIONOF PHOTONS WITHMATTER
Interactions of photons with matter, by which individual photons are removed or deected
from a primary beam of x or g radiation, may be classied according to the following:
The kind of target, such as electrons, atoms or nuclei, with which the photon
interacts
The type of event, such as absorption, scattering, or pair production, that takes
place
These interactions are thought to be independent of the origin of the photon (nuclear
transition for g-rays versus electronic transition for x-rays); hence, we use the term
photon to refer to both g- and x-rays here.
Possible interactions are summarized in Table 1 (Hubbell, 1969), where t is the total
photoelectric absorption cross section per atom (t = t
K
t
L
) and s
R
and s
C
are
Rayleigh and Compton collision cross sections, respectively.
The probability of each of these many competing independent processes can be
expressed as a collision cross section per atom, per electron, or per nucleus in the absorber.
The sum of all these cross sections, normalized to a per atom basis, is then the probability
s
tot
that the incident photon will have an interaction of some kind while passing through a
very thin absorber that contains one atom per square centimeter of area normal to the
path of the incident photon:
s
tot
= t s
R
s
C
(42)
The total collision cross section per atom s
tot
, when multiplied by the number of
atoms per cubic centimeter of absorber, is then the linear attenuation coecient m
+
per
centimeter of travel in the absorber:
m
+
1
cm
= s
tot
cm
2
atom
g
cm
3
N
0
A
atoms
g
(43)
where r is the density of the medium and N
0
is Avogadros number (6.02252610
23
atoms=g atom). The mass attenuation coecient m (cm
2
=g) is the ratio of the linear at-
tenuation coecient and the density of the material.
It is worth mentioning that the absorption coecient is a much more restricted
concept than the attenuation coecient. Attenuation includes the purely elastic process in
which the photon is merely deected and does not give up any of its initial energy to the
absorber; in this process, only a scattering coecient is involved. In a photoelectric in-
teraction, the entire energy of the incident photon is absorbed by an atom of the medium.
Copyright 2002 Marcel Dekker, Inc.
Table1 Classication of Photon Interactions
Scattering
Type of interaction Absorption Elastic (coherent) Inelastic (incoherent) Multiphoton eects
Interaction with atomic
electrons
Photoelectric effect
a
~Z
4
low energy
t
~Z
5
high energy
Rayleigh scattering
a
s
R
~Z
2
Compton scattering
a
s
C
~Z
Two-photon Compton
scattering
~Z
Interaction with nucleus
or bound nucleons
Nuclear photoelectric effect:
reactions (g, n) (g, p),
photofission
~Z
Nuclear coherent
scattering
(g, g)
~Z
2
Nuclear Compton
scattering
(g, g
/
)
~Z
(E_10 MeV)
Interaction with electrical field
surrounding charged particles
1. Electronpositron pair production
in field of nucleous,
Delbru ck scattering
~Z
4
~Z
2
(E_1.02 MeV)
2. Electronpositron pair production
in electron field
~Z
(E_2.04 MeV)
3. Nucleonantinucleon pair
production
(E_3 GeV)
Interactions with mesons Photomeson production
(E_150 MeV)
Coherent resonant
scattering (g, g)
a
Major effects of photon attenuation in matter, which are of great importance in practical x-ray spectrometry.
Source: From Hubbell, 1969.
Copyright 2002 Marcel Dekker, Inc.
In the Compton eect, some energy is absorbed and appears in the medium as the kinetic
energy of a Compton recoil electron; the balance of the incident energy is not absorbed
and is present as a Compton-scattered photon. Absorption, then, involves the conversion
of incident photon energy into the kinetic energy of a charged particle (usually an elec-
tron), and scattering involves the deection of incident photon energy.
For narrow, parallel, and monochromatic beams, the attenuation of photons in
homogeneous matter is described by the exponential law:
I = I
0
e
m
+
t
(44)
where I is the transmitted intensity, I
0
is the incident intensity, and t is the absorber
thickness in centimeters.
If the absorber is a chemical compound or a mixture, its mass attenuation coecient
can be approximately evaluated from the coecients m
i
for the constituent elements
according to the weighted average:
m =
X
n
i=1
W
i
m
i
(45)
where W
i
is the weight fraction of the ith element and n is the total number of the elements
in the absorber. The mixture rule [Eq. (45)] ignores changes in the atomic wave function
resulting from changes in the molecular, chemical, or crystalline environment of an atom.
Above 10 keV, errors from this approximation are expected to be less than a few percent
(except in the regions just above absorption edges), but at very low energies (10100 eV),
errors of a factor of 2 can occur (Deslattes, 1969).
For situations more complicated than the narrow-beam geometry, the attenuation is
still basically exponential, but it is modied by two additional factors. The rst of these,
sometimes called a geometry factor, depends on the source absorber geometry. The other
factor, often called the buildup factor, takes into account secondary photons produced in
the absorber, mainly as the result of one or more Compton scatters, which nally reach the
detector. The determination of the buildup factor, dened as the ratio of the observed
eect to the eect produced only by the primary radiation, constitutes a large part of g-ray
transport theory (Evans, 1963).
In subsequent sections, only major eects of photon attenuation are discussed
in detail.
A. Photoelectric Absorption
In the photoelectric absorption described partially in Sec. V.A, a photon disappears and
an electron is ejected from an atom. The K shell electrons, which are the most tightly
bound, are the most important for this eect in the energy region considered in XRS. If the
photon energy drops below the binding energy of a given shell, however, an electron from
that shell cannot be ejected. Hence, a plot of t versus photon energy exhibits the char-
acteristic absorption edges.
The mass photoelectric absorption coecient tN
0
=A at the incident energy E (keV)
can approximately be calculated based on Walters equations (Compton and Allison,
1935):
Copyright 2002 Marcel Dekker, Inc.
tN
0
A
=
30:3Z
3:94
AE
3
for E > E
K
0:978Z
4:30
AE
3
for E
L1
< E < E
K
0:78Z
3:94
AE
3
for E
M1
< E <
L3
8
>
<
>
:
(46)
Based on available experimental and theoretical information for approximately
10,000 combinations of Z and E covering 87 elements and the energy range 1 keV to
1 MeV, the following lnln polynomials for the photoeect cross section t
j
have been tted
in incident photon energy between each absorption-edge region (Hubbell et al., 1974):
ln t
j
=
P
1;2;or 3
i=0
A
ij
ln[E (keV)[
i
(47)
In this polynomial, the total photoeect cross section t
j
represents one of the following
sums:
t
1
= t
M
t
N
t
O
or t
N
t
O
;
E
M1
< E < E
L3
or E < E
M5
t
2
= t
L
t
M
t
N
t
O
; E
L1
< E < E
K
t
3
= t
K
t
L
t
M
t
N
t
O
; E > E
K
(48)
The values of the tted coecients A
ij
for the lnln representation are given in Appendix
VII (McMaster et al., 1969). In multiple-edge regions (e.g., between L
1
and L
3
edge en-
ergies), the photoelectric absorption cross sections are also obtained via Eq. (47) by using
the following constant jump ratios j (t just above an absorption edge divided by t just
below that absorption edge):
j = 1:16 for E
L2
< E < E
L1
j = 1:64 = 1:16 1:41 for E
L3
< E < E
L2
j = 1:1 for E
M2
< E < E
M1
j = 1:21 = 1:1 1:1 for E
M3
< E < E
M2
j = 1:45 = 1:1 1:1 1:2 for E
M4
< E < E
M3
j = 2:18 = 1:1 1:1 1:2 1:5 for E
M5
< E < E
M4
(49)
Simple expressions for calculating the values of the energies of all photoabsorption edges
are given in Sec. VI.D.
The experimental ratio of the total photoelectric absorption cross section t to the
K shell component t
K
can be tted with an accuracy of ~23% by the equation (Hubbell,
1969)
t
t
K
1 0:01481 ln
2
Z 0:000788 ln
3
Z (50)
Based on the tables of McMaster et al. (1969), Poehn et al. (1985) found a useful
approximation for the calculation of the jump ratios (called also jump factors) for the
K shell ( j
K
) and L
3
subshell ( j
L3
):
j
K
= 1:754 10 6:608 10
1
Z 1:427 10
2
Z
2
1:1 10
4
Z
3
for 11 _ Z _ 50
j
L3
= 2:003 10 7:732 10
1
Z 1:159 10
2
Z
2
5:835 10
5
Z
3
for 30 _ Z _ 83
(51)
Copyright 2002 Marcel Dekker, Inc.
As already mentioned [Eq. (49)], the values of the jump factors at the L
2
and L
1
ab-
sorption edges are constant for all elements and equal to 1.41 and 1.16, respectively.
Tabulated values for the photoelectric absorption cross sections for the elements
1 _ Z _ 100 in the energy range of 1 keV to 100 MeV are also available in the work of
Storm and Israel (1970), which provides the photon cross sections for all major interaction
processes as well as the atomic energy levels, K and L x-ray line energies, weighted average
energies for the K and L x-ray series, and relative intensities for K and L x-ray lines.
When the apparently sharp x-ray absorption discontinuities are examined at high
resolution, they are found to contain a ne structure that extends in some cases to about a
few hundred electron volts above the absorption edge. The ne structure very close to an
absorption edge (less than or equal to 50 eV above the edge) is generally referred to as the
Kossel structure and is designated as XANES (x-ray absorption near-edge structure).
Peaks and trenches in this region, which can dier by a factor of 2 or more from the
smoothly extrapolated data, can be described in terms of transitions of the (very low en-
ergy) ejected electrons to unlled discrete energy states of the atom (or molecule), rather
than to the continuum of states beyond a characteristic energy (Sandstro m, 1957; Ko-
ningsberger and Prins, 1988; Behrens, 1992b). Superimposed on the Kossel structure is the
so-called Kronig structure [extended x-ray absorption ne structure (EXAFS)], which
usually extends to about 300 eV above the absorption edge (occasionally to nearly 1 keV
above an edge). The Kronig structure can be described in terms of interference eects on
the de Broglie waves of the ejected electrons by the molecular or crystalline spatial ordering
of neighboring atoms (Hasnain, 1991; Behrens, l992a). The oscillations of the absorption
coecient are of the order of 50% in the energy region 5060 eV above an absorption edge
and of the order of 15% in the region beyond 200 eV above the edge.
Modulations of the absorption coecient in the energy region above an absorption
edge can be described theoretically in terms of the electronic parameters (Lee and Pendry,
1975). Through a Fourier transform relationship, the modulations are closely related to
the radial distribution function around the element of interest (Sayers et al., 1970). Be-
cause both the Kossel and the Kronig ne structures can vary in magnitude and in energy
displacement of the features, depending on the molecular, crystalline, or thermal en-
vironment of the atom, they can be applied for local structural analysis of various ma-
terials, including powders, disordered solids, and liquid and amorphous substances
(Lagarde, 1983, Behrens, 1992a, 1992b; Koningsberger and Prins, 1988).
B. Compton Scattering
Compton scattering (Compton, 1923a, 1923b) is the interaction of a photon with a free
electron that is considered to be at rest. The weak binding of electrons to atoms may be
neglected, provided the momentum transferred to the electron greatly exceeds the mo-
mentum of the electron in the bound state. Considering the conservation of momentum
and energy leads to the following equations:
hn =
hn
0
1 g(1 cos y)
(52)
T = hn
0
hn = hn
0
g(1 cos y)
1 g(1 cos y)
(53)
tan j =
1
1 g
cot
y
2
(54)
Copyright 2002 Marcel Dekker, Inc.
with
g =
hn
0
m
0
c
2
where hn
0
and hn are the energies of the incident and scattered photon, respectively, y is the
angle between the photon directions of travel before and following a scattering interaction,
and T and f are the kinetic energy and scattering angle of the Compton recoil electron,
respectively.
For f=180
; at
higher incident photons energies, the Compton scattering becomes predominantly forward.
The dierential KleinNishina scattering cross section ds
S
KN
=dO for unpolarized
radiation, dened as the ratio of the amount of energy scattered in a particular direction to
the energy of incident photons, is given by
ds
s
KN
dO
=
hn
hn
0
ds
KN
dO
cm
2
electron sr
(60)
Copyright 2002 Marcel Dekker, Inc.
The average (or total) collision cross section s
KN
gives the probability of any Compton
interaction by one photon while passing normally through a material containing one
electron per square centimeter:
s
KN
=
Z
p
0
ds
KN
dO
2p sin y dy
= 2pr
2
0
1 g
g
2
2(g 1)
1 2g
ln(1 2g)
g
!
ln(1 2g)
2g
1 3g
(1 2g)
2
( )
cm
2
electron
(61)
Again, at the low-energy limit, this cross section reduces to the classic Thomson cross
section:
s
Th
=
8
3
pr
2
0
= 0:6652 10
24
cm
2
electron
(62)
At extremely high energies hn
0
m
0
c
2
, Eq. (61) reduces to
s
KN
= pr
2
0
1
g
ln 2g
1
2
cm
2
electron
(63)
The average (or total) scattering cross section, dened as the total scattered energy in
photons of various energies hn, is given by
s
s
KN
=
Z
p
0
ds
s
KN
dO
2p sin y dy
= pr
2
0
ln(1 2g)
g
3
2(1 g)(2g
2
2g 1)
g
2
(1 2g)
2
8g
2
3(1 2g)
3
" #
cm
2
electron
(64)
The usual KleinNishina theory that assumes that the target electron is free and at
rest cannot be directly applicable in some cases. Departures from it occur at low energies
because of electron-binding eects and, at high energies, because of the possibility of
emission of an additional photon (double Compton eect) and radiative corrections as-
sociated with emission and reabsorption of virtual photons; these corrections are discussed
in the work of Hubbell (1969).
The total incoherent (Compton) collision cross section per atom s
C
, involving the
binding corrections by applying the so-called incoherent scattering function S(x, Z), can
be calculated according to
s
C
=
1
2
r
2
0
Z
1
1
[1 g(1 cos y)[
2
(
1 cos
2
y
g
2
(1 cos y)
2
1 g(1 cos y)
" #
ZS(x; Z)
)
2pd(cos y)
cm
2
atom
(65)
where x = sin(y=2)=l is the momentum transfer parameter and l is the photon wavelength
(in angstroms).
Copyright 2002 Marcel Dekker, Inc.
The values of the incoherent scattering function S(x, Z) and the incoherent collision
cross section s
C
are given by Hubbell et al. (1975). A useful combination of analytical
functions for calculating S(x, Z) has recently been proposed by Szalo ki (1996):
S(x; Z) =
s
1
(x; Z) =
X
3
i=1
d
i
[exp(x) 1[
i
; 0 _ x _ x
i
s
2
(x; Z) = [Z s
1
(x; Z) t
2
[g
1
(x) t
2
g
2
(x) s
1
(x
1
; Z) x
1
< x
8
>
<
>
:
(66)
where
g
1
(x) = 1 exp[t
1
(x
1
x)[
and
g
2
(x) = 1 exp[t
3
(x
1
x)[
The parameters for the calculation of the S(x, Z), including the critical value of x
1
, are
given by Szalo ki (1996) for all elements (Z=1100) and any values of x. The average
deviation between the calculated [Eq. (66)] and tabulated data (Hubbell et al., 1975) is
slightly above 1%.
The incoherent collision cross sections s
C
can also be calculated by using lnln
polynomials already dened by Eq. (47) (by simply substituting t
j
with s
C
and taking
i =3). The values of the tted coecients for the lnln representation for s
C
valid in the
photon energy range 1 keV to 1 MeV are given in Appendix VIII.
To complete this subsection, it is worth mentioning the Compton eect for polarized
radiation. The dierential collision cross section (ds
KN
=dO)
pp
for the plane-polarized ra-
diation scattered by unoriented electrons has also been derived by Klein and Nishina. It
represents the probability that a photon, passing through a target containing one electron
per square centimeter, will be scattered at an angle y into a solid angle dO in a plane
making an angle b with respect to the plane containing the electrical vector of the incident
wave:
ds
KN
dO
pp
=
r
2
0
2
hn
hn
0
2
hn
0
hn
hn
hn
0
2 sin
2
y cos
2
b
cm
2
electron sr
(67)
The cross section has its maximum value for b=90
hn
0
hn
hn
hn
0
sin
2
y
hn
0
hn
hn
hn
0
cos y
!
cm
2
electron sr
(68)
The rst term is the usual KleinNishina formula for unpolarized radiation. The sign
for the additional term applies to right circularly polarized photons.
Copyright 2002 Marcel Dekker, Inc.
C. Rayleigh Scattering
Rayleigh scattering is a process by which photons are scattered by bound atomic electrons
and in which the atom is neither ionized nor excited. The incident photons are scattered
with unchanged frequency and with a denite phase relation between the incoming and
scattered waves. The intensity of the radiation scattered by an atom is determined by
summing the amplitudes of the radiation coherently scattered by each of the electrons
bound in the atom. It should be emphasized that, in Rayleigh scattering, the coherence
extends only over the Z electrons of individual atoms. The interference is always con-
structive, provided the phase change over the diameter of the atom is less than one-half a
wavelength; that is, whenever
4p
l
r
a
sin
y
2
< 1 (69)
where r
a
is the eective radius of the atom.
Rayleigh scattering occurs mostly at the low energies and for high-Z materials, in the
same region where electron binding eects inuence the Compton scattering cross section.
The dierential Rayleigh scattering cross section for unpolarized photons is given by
(Pirenne, 1946)
ds
R
dO
=
1
2
r
2
0
(1 cos
2
y) [ F(x; Z)[
2
cm
2
atom sr
(70)
where F(x; Z) is the atomic form factor,
F(x; Z) =
Z
0
r(r)4pr
sin[(2p=l)rs[
(2p=l)rs
dr (71)
where r(r) is the total density, r is the distance from the nucleus, and s = 2 sin(y=2). The
atomic form factor has been calculated for Z<26 using the Hartree electronic distribution
(Pirenne, 1946) and for Z>26 using the FermiThomas distribution (Compton and
Allison, 1935).
At high photon energies, Rayleigh scattering is conned to small angles; at low
energies, particularly for high-Z materials, the angular distribution of the Rayleigh-scat-
tered radiation is much broader. A useful simple criterion for judging the angular spread
of Rayleigh scattering is given by (Evans, 1958).
y
R
= 2 arcsin
0:0133Z
1=3
E(Mev)
(72)
where y
R
is the opening half-angle of a cone containing at least 75% of the Rayleigh-
scattered photons. In the forward direction, [F(x; Z)[
2
= Z
2
, so that Rayleigh scattering
becomes appreciable in magnitude and must be accounted for in any g- or x-ray scattering
experiments.
The total coherent (Rayleigh) scattering cross section per atom s
R
can be calcu-
lated from
Copyright 2002 Marcel Dekker, Inc.
s
R
=
1
2
r
2
0
Z
1
1
(1 cos
2
y)[F(x; Z)[
2
2p d(cos y)
=
3
8
s
Th
Z
1
1
(1 cos
2
y)[F(x; Z)[
2
d(cos y)
cm
2
atom
(73)
The values of the atomic form factor F(x, Z) and the coherent scattering cross section s
R
are given in the work of (Hubbell et al., 1975). Recently, Szalo ki (1996) proposed a useful
combination of analytical functions to calculate F(x, Z):
F
1
(x; Z) =
f
11
(x; Z) = a exp(b
1
x) (Z a) exp(cx); 0 _ x _ x
1
f
12
(x; Z) = f
11
(x
1
; Z) exp(b
2
(x
1
x)[; x
1
_ x _ x
2
f
13
(x; Z) = f
12
(x
2
; Z) exp[b
3
(x
2
x)[; x
2
_ x _ x
3
f
14
(x; Z) = f
13
(x
3
; Z)
x
x
3
h i
b
4
; x
3
_ x _ x
4
8
>
>
>
<
>
>
>
:
(74)
where 1 _ Z _ 7
F
2
(x; Z) =
f
21
(x; Z) = a exp(b
1
x) (Z a) exp(cx); 0 _ x _ x
1
f
22
(x; Z) = f
21
(x
1
; Z) exp[b
2
(x
1
x)[; x
1
_ x _ x
2
f
23
(x; Z) = f
22
(x
2
; Z) exp[b
3
(x
2
x)[; x
2
_ x _ x
3
8
<
:
(75)
where 8 _ Z _ 100
The parameters for the calculation of F(x, Z), including the critical values of x
1
, x
2
,
and x
3
, are given by Szalo ki (1996) for all elements (Z=1100) and the momentum
transfer x from 0 to 15 A
71
. The average deviation between the calculated [Eqs. (74) and
(75)] and tabulated data (Hubbell et al., 1975) is less than 2%.
The simplest method for calculating the coherent scattering cross section s
R
consists
in applying the lnln representation [see Eq. (47) with s
R
instead of t
j
and i =3]. The
values of the tted coecients for lnln polynomials for calculating s
R
in the photon
energy range 1 keV to 1 MeV are given in Appendix IX.
D. Total Mass Attenuation Coefficient
An extensive review of current tabulations of x-ray attenuation coecients has been given
by Hubbell (1984). Dierences between various compilations of total mass attenuation
coecients result from uncertainties in our knowledge of partial cross sections for the
interaction of photons with matter as a function of elemental atomic number Z and
photon energy E. Present discrepancies are disturbing, to say the least, frequently
amounting to 510% in the photon energy region below 10 keV and rising to as much as
30% near an absorption edge.
Hubbell (1982) has tabulated mass attenuation coecients and mass energy ab-
sorption coecients for photon energies from 1 keV to 20 MeV for 40 elements ranging
from hydrogen (Z=1) to uranium (Z=92) and for 45 mixtures and compounds of do-
simetric interest.
The uncertainty ranges for the total mass attenuation coecient values in the tabu-
lation of McMaster et al. (1969) have been estimated by Hubbell et al. (1974). These ranges
of uncertainties fall into four categories. Category I (uncertainty below 2%) applies over
the energy region 640 keV (except near absorption edges) for the following elements: C,
Copyright 2002 Marcel Dekker, Inc.
Mg, Al, Ti, Fe, Ni, Cu, Zn, Zr, Mo, Pd, Ag, Cd, Sn, La, Gd, Ta, W, Pt, Au, Pb, Th, and U.
In this category, the photon energy region above 100 keV is also included for all elements in
which incoherent scattering comprises more than 90% of the total cross section. Category
II (uncertainty of 25%) applies to the energy region 26 keV for all elements, 640 keV for
all elements not specied in category I, and above 40 keV except for the scattering-domi-
nated region specied in category I. In category III (uncertainty of 515%), the authors
(Hubbell et al., 1974) included (1) the elements hydrogen, helium, and lithium, (2) the energy
region 12 keV for elements, and (3) the regions containing K, L, M, and N absorption
edges, and the ne-structure regions extending from 200 eV to 1 keV above each of these
regions. The experimental uncertainties in cases 1 and 2 greatly exceed 15%. Category IV
(uncertainty above 15%) applies to the photon energy region about 200 eV above an ab-
sorption edge (Kossel and Kronig ne-structure regions) for all elements.
Based on the tables published by McMaster et al. (1969), Wernisch et al. (1984)
developed an algorithm for the calculation of the total mass attenuation coecient valid
for the photon energy range from 1 to 50 keV and for 73 elements (11 _ Z _ 83). The
authors have applied the simple expression
m
cm
2
g
= He
dk ln E
(76)
Values of H, d, and k have been obtained from least-squares ts applied to the data
published by McMaster et al. (1969); they are given in Figure 5.
The values of the edge energies E
K
, E
L1
, E
L2
, E
L3
, E
M1
, E
M2
, E
M3
, E
M4
, and E
M5
can
simply be calculated from (Wernisch et al., 1984)
E
i
= r
i
s
i
Z t
i
Z
2
n
i
Z
3
keV (77)
the parameters r
i
, s
i
, t
i
, and n
i
for various absorption edges i are given in Table 2.
Another exible semiempirical scheme to calculate the total mass attenuation coef-
cient m for a very wide photon energy range (0.11000 keV) has been proposed by Orlic
et al. (1993)
m = exp[ p
1
p
2
(ln l) p
3
(ln l)
2
p
4
(ln l)
3
[ s
KN
ZN
0
A
1
(78)
where s
KN
is the average incoherent collision cross section gives by Eq. (61). The values of
the tting parameters p
1
, p
2
, p
3
, and p
4
are constant for each element and within the energy
regions dened by two adjacent absorption edges or for energies beyond the K absorption
edge; the parameters are given in Appendix X.
The experimental and theoretical values of total interaction cross sections [s
tot
,
Eq. (42)] and the mass attenuation coecients m for the elements (1 _ Z _ 98) in the
energy range 4.924.9 keV have been collected and compared by Arndt et al. (1992).
Most of the available tabulations of x-ray attenuation coecients do not include the
photon energy region below 1 keV that corresponds to the energies of characteristic K
x-rays of light elements (Z<11). Experimental data in this energy region are incomplete
and it should not be assumed that the accuracy of the available tabulated values is better
than ~15% (Veigele, 1974; Appendix XI).
E. Diffraction, Refraction, and Dispersion
When a beam of monochromatic x-rays falls onto a crystal lattice, a regular periodic
arrangement of atoms, a diracted beam only results in denite directions. The phe-
nomenon of x-ray diraction at an ordered array of atoms (or molecules) can also be
Copyright 2002 Marcel Dekker, Inc.
Figure 5 Definition of the energy ranges (a) and values of the parameters H, d, and k (b) applied
for the calculation of the total mass attenuation coefficients according to Eq. (76). (From Wernisch
et al., 1984.)
Copyright 2002 Marcel Dekker, Inc.
interpreted as a reection of an incident x-ray beam by the interior planes of a crystal
(Bragg reection). By elementary calculation of the dierence in path between two co-
herent rays, W. L. Bragg found the reinforcement condition for reection (known as
Braggs equation or law):
nl = 2d sin
y
n
2
(79)
where n is the order of reection, d is the interplanar spacing, and y
n
=2 is the angle of
reection (or Bragg angle) dened as the angle between the reecting plane of the crystal
and the incident or reected beam. The rst-order reection (n =1) is normally strongest,
and the reected intensity decreases as n increases.
Braggs law as given in Eq. (79) is only a rst approximation, as the refraction in the
crystal interferes with the angle of reection. Because the refractive index of x-rays is
slightly less than unity, the deviations from Braggs law, Eq. (79), were not observed in the
early years until methods were found for precise measurements of x-ray wavelengths. The
refraction is accounted for by ascribing a slightly dierent value d
n
of the lattice constant
to each order of reection; the simple Braggs law [Eq. (79)] can thus be written as
(Sandstro m, 1957)
nl = 2d 1
4d
2
n
2
d
l
2
sin
y
n
2
(80)
where d = 1 n
/
for small photon absorption (l <1 A
) and n
/
is the refractive index for
x-rays. d is a small positive number of the order 10
5
for heavy elements and 10
6
for light
elements at l =1 A
and is proportional to l
2
.
Because the value of d is positive, total reection occurs back into air when an x-ray
beam meets a surface at a large enough angle of incidence. Provided no absorption occurs,
the critical glancing angle y
tr
(tr, total reection) is dened by (Sandstro m, 1957)
sin y
tr
=
2d
_
(81)
Table2 Values of the Parameters r
i
,s
i
,t
i
, and n
i
Applied for Calculating the Energies of Absorption
Edges via Eq. (77)
Applicability range
i r
i
s
i
t
i
n
i
%
a
Z
min
Z
max
K 1.304610
1
2.633610
3
9.718610
3
4.144610
5
3.5 11 63
L
1
4.506610
1
1.566610
2
7.599610
4
1.792610
5
2.2 28 83
L
2
6.018610
1
1.964610
2
5.935610
4
1.843610
5
2.3 30 83
L
3
3.390610
1
4.931610
2
2.336610
3
1.836610
6
1.9 30 83
M
1
8.645 3.977610
1
5.963610
3
3.624610
5
0.4 52 83
M
2
7.499 3.459610
1
5.250610
3
3.263610
5
0.4 55 83
M
3
6.280 2.831610
1
4.117610
3
2.505610
5
0.4 55 83
M
4
4.778 2.184610
1
3.303610
3
2.115610
5
0.4 60 83
M
5
2.421 1.172610
1
1.845610
3
1.397610
5
0.4 61 83
a
Standard deviation of calculated energies [Eq. (77)] relative to the energies from the tables of McMaster et al.
(1969).
Source: From Wernisch et al., 1984.
Copyright 2002 Marcel Dekker, Inc.
For l =0.1 nm, the value of y
tr
is of the order 10
3
for light elements and 5610
3
for
heavy elements, increasing in proportion to l
The refractive index n
/
for a medium containing one type of atom can be calculated
from (Hirsch, 1962)
n
/
= 1
Nl
2
2p
e
2
m
0
c
2
F(0) (82)
where N is the number of atoms per unit volume and F(0) is the atomic scattering factor at
zero scattering angle. Equation (82) shows that n
/
depends on the wavelength; this phe-
nomenon is called dispersion.
The anomalous dispersion causes the quantity d=l
2
, Eq. (80), to vary slightly with
the wavelength. The variations become important only in the neighborhood of the ab-
sorption edges of the constituents of the crystal. Dividing the quantity d=l
2
into one
normal part [(d=l
2
)
n
[ and one anomalous part [(d=l
2
)
a
[, the theory of anomalous dis-
persion leads to an expression of Braggs law that can be written as (Sandstro m, 1957)
nl = 2d 1
4d
2
n
2
d
l
2
n
4d
2
n
2
d
l
2
a
!
sin
y
n
2
(83)
Combined with Braggs law in its uncorrected form, Eq. (79), this expression becomes
(Sandstro m, 1957)
l = 2d 1
4d
2
n
2
d
l
2
n
!
sin
y
n
=2
n
4d
2
n
2
d
l
2
a
l (84)
The values of (d=l
2
)
a
giving a correction for anomalous dispersion can be determined
experimentally.
The theory of anomalous dispersion has been applied by Sparks (1974, 1975) to
explain the inelastic angular-independent scattering from elements having an absorption-
edge energy just above the energy of the incident x-rays. The observed intensity of the
inelastically scattered radiation was found to be dependent on the nearness of the energy
of an absorption edge to the energy of the incident x-rays. The energy of the inelastic peaks
is shifted from the incident energy by the binding energy of the most tightly bound shell
from which electrons could be photoejected by the incident radiation.
F. X-ray Raman Scattering
Immediately after the discovery of Raman scattering in the visible-wavelength region,
a similar eect concerning x-ray radiation was experimentally examined (Davis and
Mitchell, 1928; Krishnan, 1928). X-ray Raman scattering appears as a band spectrum
having a short-wavelength edge corresponding to a denite energy loss equal to the K
electron-binding energy E
K
of the element. This inelastic eect was observed, for example,
when CrKa and CuKa radiation was scattered by solids of light elements, such as lithium,
beryllium , boron, and by graphite (Suzuki, 1966; Suzuki et al., 1970; Suzuki and Naga-
sawa, 1975). The shape of the Raman band is similar to that of the soft x-ray K absorption
spectrum of the solids.
Mizuno and Ohmura (1967) have found the following two conditions for x-ray
Raman scattering:
4p" a
l
0
< 1 (85)
Copyright 2002 Marcel Dekker, Inc.
and
hv
0
E
K
(86)
where a" is the mean radius of charge distribution of the K electrons and l
0
is the wave-
length of the incident x-rays. The intensity of the x-ray Raman scattering I(y, l) is given by
(Suzuki and Nagasawa, 1975)
I(y; l) (1 cos
2
y) T
1
(l)
4p" a
l
0
sin
y
2
!
2
T
2
(l)
4p" a
l
0
sin
y
2
!
2
( )
(87)
where T
1
(l) and T
2
(l) factors are related to the dipole and multipole transitions, respec-
tively. Although according to this equation a slight displacement of the peak position of
the Raman band is expected with the scattering angle, the peak position does not shift in
most experiments.
In general, x-ray Raman scattering gives information about the unoccupied states
above the Fermi level of the relevant solids. Moreover, this kind of inelastic scattering by
electrons in solids should sometimes be taken into account in x-ray spectrum evaluation.
VII. INTENSITYOF CHARACTERISTICX-RAYS
This section provides some of the necessary background information for subsequent
chapters dealing with quantitative x-ray analysis. Derivation of any relationship between
excitation source intensity and measured characteristic x-rays is sometimes complex and is
presented in detail in many relevant books, such as those by Jenkins et al. (1981) and
Tertian and Claisse (1982) on XRF analysis and by Goldstein et al. (1981) and Heinrich
(1981) on EPXMA.
A. Photon Excitation
When continuous (polychromatic) radiation is used to excite the characteristic x-rays of
element i in a completely homogeneous sample of thickness T (cm) and when enhancement
eects are neglected, the intensity of the uorescent radiation I
i
(E
i
) is described by
I
i
(E
i
) dO
1
dO
2
=
dO
1
dO
2
4p
e(E
i
)
sin C
1
Z
E
max
E
c;i
a
i
(E
0
)
1 exp[rT(m(E
0
) csc C
1
m(E
i
) csc C
2
)[
m(E
0
) csc C
1
m(E
i
) csc C
2
I
0
(E
0
) dE
0
(88)
with
a
i
(E
0
) = W
i
t
/
i
(E
0
)o
i
p
i
1
1
j
i
(89)
where dO
1
and dO
2
are the dierential solid angles for the incident (primary) and emerging
(characteristic) radiation, respectively; e(E
i
) is the intrinsic detector eciency for recording
a photon of energy E
i
; E
c;i
and E
max
are the critical absorption energy of element i and the
maximum energy in the excitation spectrum; r is the density of the specimen (in g=cm
3
);
C
1
and C
2
are the eective incidence and takeo angles, respectively; m(E
0
) and m(E
i
) are
the total mass attenuation coecients (in cm
2
=g) for the whole specimen [Eq. (45)] at
Copyright 2002 Marcel Dekker, Inc.
energies E
0
and E
i
, respectively; I
0
(E
0
) dE
0
is the number of incident photons per second
per steradian in the energy interval E
0
to E
0
dE
0
; W
i
is the weight fraction of the ith
element; and t
/
i
(E
0
) is the total photoelectric mass absorption coecient for the ith element
at the energy E
0
(in cm
2
=g).
Because large solid angles O
1
and O
2
are used for the excitation and characteristic
x-rays in practical spectrometers, Eq. (88) should also be integrated over these nite solid
angles. Such calculations can often be omitted; however, and for a given measurement
geometry, an experimentally determined geometry factor G can be applied.
As seen from Eq. (88), the intensity of characteristic x-rays is modied by the eects
of primary [m(E
0
)] and secondary [m(E
i
)] absorption in the specimen; this is a major source
of the so-called matrix eects in XRF analysis.
If the excitationsource is monochromatic (emits only one energy), Eq. (88) simplies to
I
i
(E
i
) = G
e(E
i
)a
i
(E
0
)I
0
(E
0
)
sin C
1
1 exp[rT(m(E
0
) csc C
1
m(E
i
) csc C
2
)[
m(E
0
) csc C
1
m(E
i
) csc C
2
(90)
The enhancement eect, consisting of an extra excitation of the element of interest by the
characteristic radiation of some matrix elements, modies the equations for the intensity
I
i
(E
i
). In the case of monochromatic photon excitation, a factor, 1 H
i
, should be in-
cluded in Eq. (90), where H
i
is the enhancement term dened as
H
i
=
1
2m
i
(E
0
)
X
m
k=1
W
k
o
k
1
1
j
k
m
i
(E
k
)m
k
(E
0
)
ln(1 m(E
0
)=[m(E
k
) sin C
1
[)
m(E
0
)= sin C
1
ln(1 m(E
i
)=[m(E
k
) sin C
2
[)
m(E
i
)= sin C
2
!
(91)
where m
i
(E
0
) and m
i
(E
k
) are the total mass attenuation coecients for the ith element at
the energy of incident radiation (E
0
) and characteristic radiation of the element k (E
k
),
respectively, m
k
(E
0
) is the total mass attenuation coecient for the element k at the
energy E
0
, and m(E
k
) is the total mass attenuation coecient for the whole specimen at
the energy E
k
.
1. Thin-SampleTechnique
If the total mass per unit area of a given sample m = rT is small, Eq. (90) simplies to
I
thin
i
(E
i
) =
G
sin C
1
e(E
i
)a
i
(E
0
)I
0
(E
0
)m (92)
The relative error resulting from applying Eq. (92) instead of Eq. (90) does not exceed 5%
when the total mass per unit area satises the condition
m
thin
_
0:1
m(E
0
) csc C
1
m(E
i
) csc C
2
(93)
A major feature of the thin-sample technique is that the intensity of the characteristic
x-rays, I
thin
i
, depends linearly on the concentration of the ith element (or on its mass per
unit area); it is equivalent to the fact that, in the thin-sample technique, matrix eects can
safely be neglected.
Copyright 2002 Marcel Dekker, Inc.
2. Thick-SampleTechnique
By the term thick sample, we mean a sample whose mass per unit area (thickness) is
greater than the so-called saturation mass (thickness). The saturation thickness is dened
as a limiting value above which practically no further increase in the intensity of the
characteristic radiation is observed as the sample thickness is increased.
If the total mass per unit area of a given sample is suciently large, Eq. (90) sim-
plies to
I
thick
i
(E
i
) =
Ge(E
i
)a
i
(E
0
)I
0
(E
0
)
m(E
0
) sin C
1
=(sin C
2
)m(E
i
)
(94)
The relative error resulting from applying Eq. (94) instead of Eq. (90) does not exceed 1%
when the total mass per unit area satises the condition
m
thick
_
4:61
m(E
0
) csc C
1
m(E
i
) csc C
2
(95)
B. Electron Excitation
For a thin foil (dened as a foil of thickness such that the beam electron undergoes only one
scattering act), the intensity of the characteristic x-rays I
tf
i
(in photons per second) is given by
I
tf
i
= K
i
I
0
Q
i
o
i
N
0
1
A
i
rTW
i
(96)
where K
i
is a factor depending on the measurement geometry and detection eciency and
the other symbols have the following dimensions: I
0
=electrons=s, Q
i
=ionizations per
electron for 1 atom=cm
2
, o
i
=photons=ionization. N
0
=atoms=mol, 1=A
i
=mol=g,
r=g=cm
3
, and T=cm.
For a bulk target, Eq. (96) must be integrated over the electron path, taking into
account the loss of energy by the electron beam:
I
bulk
i
= I
0
o
i
K
i
N
0
W
i
A
i
Z
E
c;i
E
0
Q
i
(E)
dE=d(rx)
dE (97)
where dE=d(rx) is the electron energy loss per unit of distance traveled in a material, given
by the Bethe equation (Heinrich, 1981; Bethe and Ashkin, 1953):
dE
d(rx)
= 78; 500
1
E
X
n
j=1
W
j
Z
j
A
j
ln 1:166
E
J
j
keV=g cm
2
(98)
where J
j
is the mean ionization potential of element j(J
j
= 9:76Z
j
58:5Z
0:19
j
eV)
(Goldstein et al., 1981). Strictly speaking, the Bethe equation is only valid for the electron
energy E > 6:3J
j
. Below this limit, the energy loss of the electrons should be described
either by the modication of Rao-Sahib and Wittry (1972) [Eq. (99)] or by the modi-
cation of Love et al. (l978) [Eq. (l00)]:
Copyright 2002 Marcel Dekker, Inc.
dE
d(rx)
= 0:6236 10
5
1
E
_
X
n
j=1
W
j
Z
j
A
j
J
j
p keV=g cm
2
(99)
dE
d(rx)
=
X
n
j=1
W
j
Z
j
A
j
J
j
1
1:18 10
5
(E=J
j
)
1=2
1:47 10
6
E=J
j
keV=g cm
2
(100)
However, to describe the intensity of characteristic x-rays emitted, three eects should be
considered additionally (Heinrich, 1981; Goldstein et al., 1981; Love and Scott, 1981):
1. Absorption of characteristic x-rays within the specimen
2. Electron backscattering
3. Secondary uorescence by characteristic x-rays and=or bremsstrahlung con-
tinuum produced by an electron beam
C. Particle Excitation
For a thin, uniform, homogeneous target, the intensity of characteristic x-rays I
tt
i
is given
by the simple formula
I
tt
i
= K
i
I
0
N
0
s
s;i
(E
0
)o
i
1
A
i
m
i
(101)
where m
i
is the areal density of the element with atomic number Z
i
and atomic mass A
i
.
For a thick, homogeneous target, the intensity of characteristic x-rays, I
thick
i
, from
the element i of concentration W
i
can be calculated by (Campbell and Cookson, 1984)
I
thick
i
=
K
i
o
i
N
0
A
i
I
0
W
i
X
0
E
0
s
s;i
(E)T
i
(E)
S(E)
dE (102)
where S(E) is the stopping power and T
i
(E) is the photon attenuation factor. The latter is
given by
T
i
(E) = exp m(E
i
)
sin C
1
sin C
2
Z
E
E
0
dE
S(E)
0
B
@
1
C
A (103)
More details on various x-ray analytical techniques are provided in subsequent chapters.
VIII. IUPACNOTATIONFOR X-RAYSPECTROSCOPY
The nomenclature commonly used in XRS to describe x-ray emission spectra was in-
troduced by M. Sieghahn in the l920s and is based on the relative intensity of lines from
dierent series. A new and more systematic notation for x-ray emission lines and ab-
sorption edges, based on the energy-level designation, was developed by the International
Union of Pure and Applied Chemistry (Jenkins et al. 1991). Because the new notation,
called the IUPAC notation, replaces the Siegbahns notation, some characteristic features
of the new nomenclature must be mentioned. The IUPAC notation prescribes Arabic
numerals for subscripts; the original notation uses Roman numerals (e.g., L
2
and L
3
,
Copyright 2002 Marcel Dekker, Inc.
instead of L
II
and L
III
). In the IUPAC notation, states with double or multiple vacancies
should be denoted by, for example, K
2
; KL
1
, and L
n
2;3
, which correspond to the electron
congurations 1s
2
; 1s
1
2s
1
, and 2p
n
, respectively. X-ray transitions and x-ray emission
diagram lines are denoted by the initial (placed rst) and nal x-ray levels separated by a
hyphen. To conform with the IUPAC notation of x-ray spectra, the hyphen separating the
initial and nal state levels should also be introduced into the notation for Auger electron
emission process. The IUPAC notation is compared with the Sieghahn notation in Ap-
pendix XII.
Copyright 2002 Marcel Dekker, Inc.
APPENDIXI: CRITICAL ABSORPTIONWAVELENGTHS AND CRITICAL ABSORPTIONENERGIES
Atomic
number
K edge L
1
edge L
2
edge L
3
edge M
4
edge M
5
edge
Element A
keV A
keV A
keV A
keV A
keV A
keV
1 H 918 0.014
2 He 504 0.025
3 Li 226.953 0.055
4 Be 106.9 0.116
5 B 64.6 0.192
6 C 43.767 0.283
7 N 31.052 0.399
8 O 23.367 0.531
9 F 18.05 0.687
10 Ne 14.19 0.874 258 0.048 564 0.022 564 0.022
11 Na 11.48 1.08 225 0.055 365 0.034 365 0.034
12 Mg 9.512 1.303 197 0.063 248 0.050 253 0.049
13 Al 7.951 1.559 143 0.087 170 0.073 172 0.072
14 Si 6.745 1.837 105 0.118 125 0.099 127 0.098
15 P 5.787 2.142 81.0 0.153 96.1 0.129 96.9 0.128
16 S 5.018 2.470 64.2 0.193 75.6 0.164 76.1 0.163
17 Cl 4.397 2.819 52.1 0.238 61.1 0.203 61.4 0.202
18 Ar 3.871 3.202 43.2 0.287 50.2 0.247 50.6 0.245
19 K 3.437 3.606 36.4 0.341 41.8 0.297 42.2 0.294
20 Ca 3.070 4.037 30.7 0.399 35.2 0.352 35.5 0.349
21 Sc 2.757 4.495 26.8 0.462 30.2 0.411 30.8 0.402
22 Ti 2.497 4.963 23.4 0.530 27.0 0.460 27.3 0.454
23 V 2.269 5.462 20.5 0.604 23.9 0.519 24.2 0.512
24 Cr 2.070 5.987 18.3 0.679 21.3 0.583 21.6 0.574
25 Mn 1.896 6.535 16.3 0.762 19.1 0.650 19.4 0.639
26 Fe 1.743 7.109 14.6 0.849 17.2 0.721 17.5 0.708
27 Co 1.608 7.707 13.3 0.929 15.6 0.794 15.9 0.779
28 Ni 1.488 8.329 12.22 1.015 14.2 0.871 14.5 0.853
29 Cu 1.380 8.978 11.27 1.100 13.0 0.953 13.3 0.933
Copyright 2002 Marcel Dekker, Inc.
30 Zn 1.283 9.657 10.33 1.200 11.87 1.045 12.13 1.022
31 Ga 1.196 10.365 9.54 1.30 10.93 1.134 11.10 1.117
32 Ge 1.117 11.100 8.73 1.42 9.94 1.248 10.19 1.217
33 As 1.045 11.860 8.107 1.529 9.124 1.358 9.39 1.32
34 Se 0.980 12.649 7.506 1.651 8.416 1.473 8.67 1.43
35 Br 0.920 13.471 6.97 1.78 7.80 1.59 8.00 1.55
36 Kr 0.866 14.319 6.46 1.92 7.21 1.72 7.43 1.67
37 Rb 0.816 15.197 5.998 2.066 6.643 1.865 6.89 1.80
38 Sr 0.770 16.101 5.583 2.220 6.172 2.008 6.387 1.940
39 Y 0.728 17.032 5.232 2.369 5.755 2.153 5.962 2.079
40 Zr 0.689 17.993 4.867 2.546 5.378 2.304 5.583 2.220
41 Nb 0.653 18.981 4.581 2.705 5.026 2.467 5.223 2.373
42 Mo 0.620 19.996 4.298 2.883 4.718 2.627 4.913 2.523
43 Tc 0.589 21.045 4.060 3.054 4.436 2.795 4.632 2.677
44 Ru 0.561 22.112 3.83 3.24 4.180 2.965 4.369 2.837
45 Rh 0.534 23.217 3.626 3.418 3.942 3.144 4.130 3.001
46 Pd 0.509 24.341 3.428 3.616 3.724 3.328 3.908 3.171
47 Ag 0.486 25.509 3.254 3.809 3.514 3.527 3.698 3.351
48 Cd 0.464 26.704 3.085 4.018 3.326 3.726 3.504 3.537
49 In 0.444 27.920 2.926 4.236 3.147 3.938 3.324 3.728
50 Sn 0.425 29.182 2.777 4.463 2.982 4.156 3.156 3.927
51 Sb 0.407 30.477 2.639 4.695 2.830 4.380 3.000 4.131
52 Te 0.390 31.800 2.511 4.937 2.687 4.611 2.855 4.340
53 I 0.374 33.155 2.389 5.188 2.553 4.855 2.719 4.557
54 Xe 0.359 34.570 2.274 5.451 2.429 5.102 2.592 4.780
55 Cs 0.345 35.949 2.167 5.719 2.314 5.356 2.474 5.010
56 Ba 0.331 37.399 2.068 5.994 2.204 5.622 2.363 5.245 15.56 0.7967 15.89 0.7801
57 La 0.318 38.920 1.973 6.282 2.103 5.893 2.258 5.488
58 Ce 0.307 40.438 1.889 6.559 2.011 6.163 2.164 5.727
59 Pr 0.295 41.986 1.811 6.844 1.924 6.441 2.077 5.967 13.122 0.9448 13.394 0.9257
60 Nd 0.285 43.559 1.735 7.142 1.843 6.725 1.995 6.213 12.459 0.9951 23.737 0.9734
61 Pm 0.274 45.207 1.665 7.448 1.767 7.018 1.918 6.466
62 Sm 0.265 46.833 1.599 7.752 1.703 7.279 1.845 6.719 11.288 1.0983 11.552 1.0732
63 Eu 0.256 48.501 1.536 8.066 1.626 7.621 1.775 6.981 10.711 1.1575 11.013 1.1258
Copyright 2002 Marcel Dekker, Inc.
Appendix I Continued
Atomic
number
K edge L
1
edge L
2
edge L
3
edge M
4
edge M
5
edge
Element A
keV A
keV A
keV A
keV A
keV A
keV
64 Gd 0.247 50.215 1.477 8.391 1.561 7.938 1.710 7.250
65 Tb 0.238 51.984 1.421 8.722 1.501 8.256 1.649 7.517
66 Dy 0.231 53.773 1.365 9.081 1.438 8.619 1.579 7.848
67 Ho 0.223 55.599 1.317 9.408 1.390 8.918 1.535 8.072
68 Er 0.216 57.465 1.268 9.773 1.338 9.260 1.482 8.361 8.601 1.4415 8.847 1.4013
69 Tm 0.209 59.319 1.222 10.141 1.288 9.626 1.433 8.650 8.487 1.4609
70 Yb 0.202 61.282 1.182 10.487 1.243 9.972 1.386 8.941
71 Lu 0.196 63.281 1.140 10.870 1.199 10.341 1.341 9.239
72 Hf 0.190 65.292 1.100 11.271 1.155 10.732 1.297 9.554
73 Ta 0.184 67.379 1.061 11.681 1.114 11.128 1.255 9.874 6.87 1.804 7.11 1.743
74 W 0.178 69.479 1.025 12.097 1.075 11.533 1.216 10.196 6.59 1.880 6.83 1.814
75 Re 0.173 71.590 0.990 12.524 1.037 11.953 1.177 10.529 6.33 1.958 6.560 1.890
76 Os 0.168 73.856 0.956 12.968 1.001 12.380 1.140 10.867 6.073 2.042 6.30 1.967
77 Ir 0.163 76.096 0.923 13.427 0.967 12.817 1.106 11.209 5.83 2.126 6.05 2.048
78 Pt 0.158 78.352 0.893 13.875 0.934 13.266 1.072 11.556 5.59 2.217 5.81 2.133
79 Au 0.153 80.768 0.863 14.354 0.903 13.731 1.040 11.917 5.374 2.307 5.584 2.220
80 Hg 0.149 83.046 0.835 14.837 0.872 14.210 1.008 12.3 5.157 2.404 5.36 2.313
81 Tl 0.415 85.646 0.808 15.338 0.843 14.695 0.979 12.655 4.952 2.504 5.153 2.406
82 Pb 0.141 88.037 0.782 15.858 0.815 15.205 0.950 13.041 4.757 2.606 4.955 2.502
83 Bi 0.137 90.420 0.757 16.376 0.789 15.713 0.923 13.422 4.572 2.711 4.764 2.603
84 Po 0.133 93.112 0.732 16.935 0.763 16.244 0.897 13.817
85 At 0.130 95.740 0.709 17.490 0.739 16.784 0.872 14.215
86 Rn 0.126 98.418 0.687 18.058 0.715 17.337 0.848 14.618
87 Fr 0.123 101.147 0.665 18.638 0.693 17.904 0.825 15.028
88 Ra 0.645 19.229 0.671 18.478 0.803 15.439
89 Ac 0.116 106.759 0.625 19.842 0.650 19.078 0.782 15.865
90 Th 0.113 109.741 0.606 20.458 0.630 19.677 0.761 16.293 3.557 3.485 3.729 3.325
91 Pa 0.110 112.581 0.588 21.102 0.611 20.311 0.741 16.731 3.436 3.608 3.618 3.436
92 U 0.108 115.610 0.56 21.764 0.592 20.938 0.722 17.160 3.333 3.720 3.497 3.545
Copyright 2002 Marcel Dekker, Inc.
93 Np 0.105 118.619 0.553 22.417 0.574 21.596 0.704 17.614
94 Pu 0.102 121.720 0.537 23.097 0.557 22.262 0.686 18.066
95 Am 0.099 124.816 0.521 23.793 0.540 22.944 0.669 18.525
96 Cm 0.097 128.088 0.506 24.503 0.525 23.640 0.653 18.990
97 Bk 0.094 131.357 0.491 25.230 0.509 24.352 0.637 19.461
98 Cf 0.092 134.683 0.477 25.971 0.494 25.080 0.622 19.938
99 Es 0.090 138.067 0.464 26.729 0.480 25.824 0.607 20.422
100 Fm 0.088 141.510 0.451 27.503 0.466 26.584 0.593 20.912
Source: From Clark, 1963 and Burr, 1974.
Copyright 2002 Marcel Dekker, Inc.
APPENDIXII: CHARACTERISTICX-RAY WAVELENGTHS (A ) ANDENERGIES (keV)
Table1 K Series Diagram Lines (A
)
a
Line a
1,2
a
1
a
2
b
1
b
3
b
2
b
4
Approximate
intensity 150 100 50 15 5 <1
Li 3 230
B 4 113
B 5 67
C 6 44
N 7 31.603
O 8 23.707
F 9 18.307
Ne 10 14.615 14.460
Na 11 11.909 11.574 11.726
Mg 12 9.889 9.559 9.667
Al 13 8.339 8.338 8.341 7.960 8.059
Si 14 7.126 7.125 7.127 6.778
P 15 6.155 6.154 6.157 5.804
S 16 5.373 5.372 5.375 5.032
Cl 17 4.729 4.728 4.731 4.403
Ar 18 4.192 4.191 4.194 3.886
K 19 3.744 3.742 3.745 3.454
Ca 20 3.360 3.359 3.362 3.089
Sc 21 3.032 3.031 3.034 2.780
Ti 22 2.750 2.749 2.753 2.514
V 23 2.505 2.503 2.507 2.285
Cr 24 2.291 2.290 2.294 2.085
Mn 25 2.103 2.102 2.105 1.910
Fe 26 1.937 1.936 1.940 1.757
Co 27 1.791 1.789 1.793 1.621
Ni 28 1.659 1.658 1.661 1.500 1.489
Cu 29 1.542 1.540 1.544 1.392 1.393 1.381
Zn 30 1.437 1.435 1.439 1.296 1.284
Ga 31 1.341 1.340 1.344 1.207 1.208 1.196
Ge 32 1.256 1.255 1.258 1.129 1.129 1.117
As 33 1.177 1.175 1.179 1.057 1.058 1.045
Se 34 1.106 1.105 1.109 0.992 0.993 0.980
Br 35 1.041 1.040 1.044 0.933 0.933 0.921
Kr 36 0.981 0.980 0.984 0.879 0.879 0.866 0.866
Rb 37 0.927 0.926 0.930 0.829 0.830 0.817 0.815
Sr 38 0.877 0.875 0.880 0.783 0.784 0.771 0.770
Y 39 0.831 0.829 0.833 0.740 0.741 0.728 0.727
Zr 40 0.788 0.786 0.791 0.701 0.702 0.690 0.689
Nb 41 0.748 0.747 0.751 0.665 0.666 0.654 0.653
Mo 42 0.710 0.709 0.713 0.632 0.633 0.621 0.620
Tc 43 0.676 0.675 0.679 0.601 0.590
Ru 44 0.644 0.643 0.647 0.572 0.573 0.562 0.561
Rh 45 0.614 0.613 0.617 0.546 0.546 0.535 0.534
Pd 46 0.587 0.585 0.590 0.521 0.521 0.510
Copyright 2002 Marcel Dekker, Inc.
Table1 Continued
Line a
1,2
a
1
a
2
b
1
b
3
b
2
b
4
Approximate
intensity 150 100 50 15 5 <1
Ag 47 0.561 0.559 0.564 0.497 0.498 0.487 0.486
Cs 48 0.536 0.535 0.539 0.475 0.476 0.465
In 49 0.514 0.512 0.517 0.455 0.455 0.445 0.444
Sn 50 0.492 0.491 0.495 0.435 0.436 0.426 0.425
Sb 51 0.472 0.470 0.475 0.417 0.418 0.408 0.407
Te 52 0.453 0.451 0.456 0.400 0.401 0.391
I 53 0.435 0.433 0.438 0.384 0.385 0.376
Xe 54 0.418 0.416 0.421 0.369 0.360
Cs 55 0.402 0.401 0.405 0.355 0.355 0.346
Ba 56 0.387 0.385 0.390 0.341 0.342 0.333
La 57 0.373 0.371 0.376 0.328 0.329 0.320 0.319
Ce 58 0.359 0.357 0.362 0.316 0.317 0.309 0.307
Pr 59 0.346 0.344 0.349 0.305 0.305 0.297
Nd 60 0.334 0.332 0.337 0.294 0.294 0.287
Pm 61 0.322 0.321 0.325 0.283
Sm 62 0.311 0.309 0.314 0.274 0.274 0.267
Eu 63 0.301 0.299 0.304 0.264 0.265 0.258
Gd 64 0.291 0.289 0.294 0.255 0.256 0.249
Tb 65 0.281 0.279 0.284 0.246 0.246 0.239
Dy 66 0.272 0.270 0.275 0.237 0.238 0.231
Ho 67 0.263 0.261 0.266 0.230 0.231
Er 68 0.255 0.253 0.258 0.222 0.223 0.217
Tm 69 0.246 0.244 0.250 0.215 0.216
Yb 70 0.238 0.236 0.241 0.208 0.208 0.203
Lu 71 0.231 0.229 0.234 0.202 0.203 0.197
Hf 72 0.224 0.222 0.227 0.195 0.196 0.190
Ta 73 0.217 0.215 0.220 0.190 0.191 0.185 0.184
W 74 0.211 0.209 0.213 0.184 0.185 0.179 0.179
Re 75 0.204 0.202 0.207 0.179 0.179 0.174 0.174
Os 76 0.198 0.196 0.201 0.173 0.174 0.169 0.168
Ir 77 0.193 0.191 0.196 0.168 0.169 0.164 0.163
Pt 78 0.187 0.158 0.190 0.163 0.164 0.159 0.159
Au 79 0.182 0.180 0.185 0.159 0.160 0.155 0.154
Hg 80 0.177 0.175 0.180 0.154 0.155 0.150 0.150
Tl 81 0.172 0.170 0.175 0.150 0.151 0.146 0.146
Pb 82 0.167 0.165 0.170 0.146 0.147 0.147 0.141
Bi 83 0.162 0.161 0.165 0.142 0.143 0.138 0.138
Po 84 0.185 0.156 0.161 0.138 0.133
At 85 0.152 0.157 0.134 0.135
Rn 86 0.148 0.153 0.131 0.132
Fr 87 0.144 0.149 0.127 0.128
Ra 88 0.144 0.149 0.127 0.128
Ac 89 0.140 0.145 0.124 0.125
Th 90 0.135 0.133 0.138 0.117 0.118 0.114 0.114
Pa 91 0.131 0.136 0.115 0.116
U 92 0.128 0.126 0.131 0.111 0.112 0.108 0.108
a
Conversion equation: E (keV) =12.4=l (A
).
Source: From Clark, 1963.
Copyright 2002 Marcel Dekker, Inc.
Table 2 L Series Diagram Lines (A
)
Line a a
1
a
2
b
1
b
2
b
3
b
4
b
5
b
6
b
7
b
9
Approximate
intensity 110 100 10 50 20 6 4 1 <1 <1 <1
Na 11
Mg 12
Al 13
Si 14
P 15
S 16
Cl 17
Ar 18
K 19
Ca 20 36.393 36.022
Sc 21 31.393 31.072
Ti 22 27.445 27.074
V 23 24.309 23.898
Cr 24 21.713 21.323 19.429
Mn 25 19.489 19.158 17.757
Fe 26 17.602 17.290 15.742
Co 27 16.000 15.698 14.269
Ni 28 14.595 14.308 13.167
Cu 29 13.357 13.079 12.115
Zn 30 12.282 12.009 11.225
Ga 31 11.313 11.045
Ge 32 10.456 10.194
As 33 9.671 9.414 8.930
Se 34 8.990 8.735
Br 35 8.375 8.126
Kr 36
Rb 37 7.318 7.325 7.075 6.788 6.821 6.984
Sr 38 6.863 6.870 6.623 6.367 6.403 6.519
Y 39 6.449 6.456 6.211 5.983 6.018 6.094
Zr 40 6.070 6.077 5.836 5.586 5.632 5.668 5.710
Nb 41 5.725 5.732 5.492 5.238 5.310 5.346 5.361
Mo 42 5.406 5.414 5.176 4.923 5.013 5.048 5.048
Tc 43
Ru 44 4.846 4.854 4.620 4.372 4.487 4.532 4.487
Rh 45 4.597 4.605 4.374 4.130 4.253 4.289 4.242
Pd 46 4.368 4.376 4.146 3.909 4.034 4.071 4.016 3.792
Ag 47 4.154 4.162 3.935 3.703 3.834 3.870 3.808 3.605
Cd 48 3.956 3.965 3.739 3.514 3.644 3.681 3.614 3.430
In 49 3.752 3.781 3.555 3.339 3.470 3.507 3.436 3.268
Sn 50 3.600 3.609 3.385 3.175 3.306 3.344 3.270 3.155 3.115
Sb 51 3.439 3.448 3.226 3.023 3.152 3.190 3.115 3.005 2.973
Te 52 3.290 3.299 3.077 2.882 3.009 3.046 2.971 2.863 2.839
I 53 3.148 3.157 2.937 2.751 2.874 2.912 2.837 2.730 2.713
Xe 54
Cs 55 2.892 2.902 2.683 2.511 2.628 2.666 2.593 2.485 2.478
Copyright 2002 Marcel Dekker, Inc.
b
10
b
15
b
17
g
1
g
2
g
3
g
4
g
5
g
6
g
8
Z s t
<1 <1 <1 10 1 2 <1 <1 <1 <1 3 1 <1 <1
410
260
180
67.84 67.25
56.212 56.813
47.835 47.325
41.042 40.542
36.671 35.200
31.423 30.942
27.826 27.375
24.840 24.339
22.315 21.864
20.201 19.73
18.358 17.86
16.693 16.304
15.297 14.940
14.081 13.719
12.976 12.620
11.944 11.608
11.069 10.732
10.293 9.959
9.583 9.253
6.045 6.754 8.363 8.042
5.644 6.297 7.836 7.517
5.283 5.875 7.356 7.040
5.384 4.953 5.497 6.918 6.606
5.036 4.654 5.151 6.517 6.210
4.726 4.380 4.387 6.150 5.847
4.182 3.897 4.288 5.503 5.204
3.944 3.685 4.045 5.217 4.922
3.799 3.725 3.489 3.822 4.952 4.660
3.611 3.523 3.307 3.616 4.707 4.418
3.437 3.336 3.137 3.426 4.480 4.193
3.274 3.162 2.980 2.026 3.249 4.269 3.983
3.121 3.001 2.835 2.778 3.085 4.071 3.789
2.979 2.852 2.695 2.639 2.932 3.888 3.607
2.847 2.712 2.567 2.511 2.790 3.716 3.438
2.720 2.582 2.447 2.391 2.657 3.557 3.280
2.492 2.348 2.237 2.233 2.174 2.417 3.267 2.994
Copyright 2002 Marcel Dekker, Inc.
Table 2 Continued
Line a a
1
a
2
b
1
b
2
b
3
b
4
b
5
b
6
b
7
b
9
Approximate
intensity 110 100 10 50 20 6 4 1 <1 <1 <1
Ba 56 2.776 2.785 2.567 2.404 2.516 2.555 2.482 2.382 2.376
La 57 2.665 2.674 2.458 2.303 2.410 2.449 2.379 2.275 2.282
Ce 58 2.561 2.570 2.356 2.208 2.311 2.349 2.282 2.180 2.188
Pr 59 2.463 2.473 2.259 2.119 2.216 2.255 2.190 2.091 2.100
Nd 60 2.370 2.382 2.166 2.035 2.126 2.166 2.103 2.009 2.016
Pm 61 2.283 2.081
Sm 62 2.199 2.210 1.998 1.882 1.962 2.000 1.779 1.946 1.856 1.862
Eu 63 2.120 2.131 1.920 1.812 1.887 1.926 1.875 1.788 1.792
Gd 64 2.046 2.057 1.847 1.746 1.815 1.853 1.807 1.723
Tb 65 1.976 1.986 1.777 1.682 1.747 1.785 1.577 1.742 1.659
Dy 66 1.909 1.920 1.710 1.623 1.681 1.720 1.681 1.599
Ho 67 1.845 1.856 1.647 1.567 1.619 1.658 1.622
Er 68 1.785 1.796 1.587 1.514 1.561 1.601 1.567 1.494 1.485
Tm 69 1.726 1.738 1.530 1.463 1.505 1.544 1.515
Yb 70 1.672 1.682 1.476 1.416 1.452 1.491 1.387 1.466 1.395 1.384
Lu 71 1.619 1.630 1.424 1.370 1.402 1.441 1.342 1.419 1.350 1.336
Hf 72 1.569 1.580 1.374 1.327 1.353 1.392 1.298 1.374 1.306 1.291
Ta 73 1.522 1.533 1.327 1.285 1.307 1.346 1.256 1.331 1.264 1.247
W 74 1.476 1.487 1.282 1.245 1.263 1.302 1.215 1.290 1.224 1.204
Re 75 1.433 1.444 1.238 1.206 1.220 1.260 1.177 1.252 1.186 1.165
Os 76 1.391 1.402 1.197 1.169 1.179 1.218 1.140 1.213 1.149 1.126
Ir 77 1.352 1.363 1.158 1.135 1.141 1.179 1.106 1.179 1.115 1.090
Pt 78 1.313 1.325 1.120 1.102 1.104 1.142 1.072 1.143 1.082 1.054
Au 79 1.277 1.288 1.083 1.070 1.068 1.106 1.040 1.111 1.050 1.021
Hg 80 1.242 1.253 1.049 1.040 1.034 1.072 1.010 1.080 1.019 0.986
Tl 81 1.207 1.218 1.015 1.010 1.001 1.039 0.981 1.050 0.990 0.957
Pb 82 1.175 1.186 0.982 0.983 0.969 1.007 0.953 1.021 0.962 0.927
Bi 83 1.144 1.155 0.952 0.955 0.939 0.977 0.926 0.993 0.935 0.898
Po 84 1.114 1.126 0.921 0.929 0.908 0.948 0.900 0.967
At 85
Rn 86
Fr 87 1.030 0.840 0.858
Ra 88 1.005 1.017 0.814 0.836 0.803 0.841 0.807 0.871 0.817 0.769
Ac 89
Th 90 0.956 0.968 0.766 0.794 0.755 0.793 0.765 0.828 0.775 0.723
Pa 91 0.933 0.945 0.742 0.774 0.732 0.770 0.746 0.803 0.755 0.701
U 92 0.911 0.923 0.720 0.755 0.710 0.748 0.726 0.789 0.736 0.681
Np 93 0.890 0.901 0.698 0.735
Pu 94 0.868 0.880 0.678 0.719 0.669 0.707 0.691
Am 95 0.849 0.860 0.658 0.701
Source: From Clark, 1963.
Copyright 2002 Marcel Dekker, Inc.
b
10
b
15
b
17
g
1
g
2
g
3
g
4
g
5
g
6
g
8
Z s t
<1 <1 <1 10 1 2 <1 <1 <1 <1 3 1 <1 <1
2.387 2.442 2.138 2.134 2.075 2.309 2.222 3.135 2.862
2.290 2.141 2.046 2.041 1.983 2.205 3.006 2.740
2.195 2.048 1.960 1.955 1.899 2.110 2.023 2.892 2.620
2.107 1.961 1.879 1.874 1.819 2.020 1.936 2.784 2.512
2.023 1.878 1.801 1.797 1.745 1.935 1.855 2.675 2.409
1.870 1.726 1.659 1.655 1.606 2.482 2.218
1.800 1.657 1.597 1.591 1.544 1.708 1.632 2.395
1.731 1.592 1.534 1.529 1.485 2.312 2.049
1.667 1.530 1.477 1.471 1.427 2.234
1.473 1.423 1.417 1.374 1.518 2.158 1.898
1.417 1.371 1.364 1.323 1.462 2.086 1.826
1.494 1.364 1.321 1.315 1.276 1.406 2.019 1.757
1.316 1.274 1.268 1.355 1.955 1.695
1.392 1.268 1.228 1.222 1.185 1.307 1.243 1.250 1.894 1.635 1.831
1.343 1.372 1.222 1.185 1.179 1.143 1.260 1.198 1.204 1.836 1.478 1.776
1.299 1.328 1.437 1.179 1.144 1.138 1.103 1.215 1.155 1.161 1.782 1.523 1.663 1.723
1.254 1.287 1.138 1.105 1.099 1.065 1.173 1.114 1.120 1.728 1.471 1.612 1.672
1.212 1.247 1.339 1.098 1.068 1.062 1.028 1.132 1.074 1.081 1.678 1.421
1.172 1.208 1.293 1.061 1.032 1.026 0.993 1.094 1.037 1.044 1.630 1.374
1.133 1.171 1.025 0.998 0.992 0.959 1.057 1.001 1.008 1.585 1.328
1.097 1.137 0.991 0.966 0.959 0.928 1.022 0.967 0.974 1.541 1.285
1.062 1.166 0.958 0.934 0.928 0.897 0.988 0.934 0.941 1.499 1.243
1.028 1.072 1.128 0.927 0.905 0.898 0.867 0.956 0.903 0.010 1.460 1.202 1.352 1.414
0.996 1.041 1.090 0.897 0.876 0.869 0.839 0.925 0.873 0.880 1.422 1.164
0.964 1.012 1.056 0.868 0.848 0.842 0.812 0.895 0.845 0.852 1.385 1.127 1.279 1.342
0.934 0.984 1.022 0.840 0.822 0.815 0.867 0.817 0.824 1.350 1.092 1.244 1.308
0.905 0.957 0.989 0.814 0.796 0.790 0.761 0.840 0.791 0.799 1.317 1.058 1.210
0.931 0.786 0.771 0.764 0.765 1.283
0.716
0.776 0.838 0.844 0.694 0.682 0.675 0.649 0.717 0.673 0.680 1.167 0.908
0.730 0.653 0.642 0.635 0.611 0.675 0.632 0.640 1.115 0.855 1.011 1.080
0.708 0.634 0.624 0.617 0.594 0.655 0.613 1.091 0.830
0.687 0.615 0.605 0.598 0.577 0.635 0.595 0.601 1.067 0.806 0.904 1.035
0.597
0.579
0.562
Copyright 2002 Marcel Dekker, Inc.
Table 3 M Series Diagram Lines (A
)
Line a
1
a
2
b g l
K 19 680
Cu 29 170
Ru 44 26.85
Rh 45 25.00
Pd 46
Ag 47 21.80
Cd 48 20.46
In 49
Sn 50 17.94
Sb 51 16.92
Te 52 15.93
Ba 56 12.700
La 57 14.88 14.51 12.064
Ce 58 14.06 13.78 11.534 18.38
Pr 59 10.997
Nd 60 12.675 Band 10.504
Sm 62 Band Band 9.599
Eu 63 Band 10.744 9.211 14.22
Gd 64 Band 10.253 8.844 13.57
Tb 65 Band 9.792 8.485 12.98
Dy 66 Band 9.364 8.144 12.43
Ho 67 Band 8.965 7.865 11.86
Er 68 Band 8.593 7.545 11.37
Tm 69 8.460 8.246
Yb 70 8.139 8.155 7.909 7.023 10.48
Lu 71 7.840 7.600 6.761 10.07
Hf 72 7.539 7.546 7.304 6.543 9.69
Ta 73 7.251 7.258 7.022 6.312 9.32
W 74 6.983 6.990 6.756 6.088 8.96
Re 75 6.528 6.504 5.887 8.63
Os 76 6.490 6.267 5.681
Ir 77 6.261 6.275 6.037 5.501 8.02
Pt 78 6.046 6.057 5.828 5.320 7.74
Au 79 5.840 5.854 5.623 5.145 7.47
Hg 80 5.666 5.452
Tl 81 5.461 5.472 5.250 4.825 6.97
Pb 82 5.285 5.299 5.075 4.674 6.74
Bi 83 5.118 5.129 4.909 4.531 6.52
Th 90 4.138 4.151 3.942 3.679 5.24
Pa 91 4.022 4.035 3.827 3.577 5.08
U 92 3.910 3.924 3.715 3.480 4.95
Source: From Clark, 1963.
Copyright 2002 Marcel Dekker, Inc.
Table 4 Energies of Principal K and L X-ray Emission Lines (keV)
Atomic
number Element Kb
2
Kb
1
Ka
1
Ka
2
Lg
1
Lb
2
Lb
1
La
1
La
2
3 Li 0.052
4 Be 0.110
5 B 0.185
6 C 0.282
7 N 0.392
8 C 0.523
9 F 0.677
10 Ne 0.851
11 Na 1.067 1.041
12 Mg 1.297 1.254
13 Al 1.553 1.487 1.486
14 Si 1.832 1.40 1.739
15 P 2.136 2.015 2.014
16 S 2.464 2.309 2.306
17 Cl 2.815 2.622 2.621
18 Ar 3.192 2.957 2.955
19 K 3.589 3.313 3.310
20 Ca 4.012 3.691 3.688 0.344 0.341
21 Sc 4.460 4.090 4.085 0.399 0.395
22 Ti 4.931 4.510 4.504 0.458 0.492
23 V 5.427 4.952 4.944 0.519 0.510
24 Cr 5.946 5.414 5.405 0.581 0.571
25 Mn 6.490 5.898 5.887 0.647 0.636
26 Fe 7.057 6.403 6.390 0.717 0.704
27 Co 7.647 6.930 6.915 0.790 0.775
28 Ni 8.328 8.264 7.477 7.460 0.866 0.849
29 Cu 8.976 8.904 8.047 8.027 0.943 0.928
30 Zn 9.657 9.571 8.638 8.615 1.032 1.009
31 Ga 10.365 10.263 9.251 9.234 1.122 1.096
32 Ge 11.100 10.981 9.885 9.854 1.216 1.166
33 As 11.863 11.725 10.543 10.507 1.517 1.282
34 Se 12.651 12.495 11.221 11.181 1.419 1.379
35 Br 13.465 13.290 11.923 11.877 1.526 1.480
36 Kr 14.313 14.112 12.648 12.597 1.638 1.587
37 Rb 15.184 14.960 13.394 13.335 1.752 1.694 1.692
38 Sr 16.083 15.834 14.164 14.097 1.872 1.806 1.805
39 Y 17.011 16.736 14.957 14.882 1.996 1.922 1.920
40 Zr 17.969 17.666 15.774 15.690 2.302 2.219 2.124 2.042 2.040
41 Nb 18.951 18.621 16.614 16.520 2.462 2.367 2.257 2.166 2.163
42 Mo 19.964 19.607 17.478 17.373 2.623 2.518 2.395 2.293 2.290
43 Tc 21.012 20.585 18.410 18.328 2.792 2.674 2.538 2.424 2.420
44 Ru 22.072 21.655 19.278 19.149 2.964 2.836 2.683 2.558 2.554
45 Rh 23.169 22.721 20.214 20.072 3.144 3.001 2.834 2.696 2.692
46 Pd 24.297 23.816 21.175 21.018 3.328 3.172 2.990 2.838 2.833
47 Ag 25.454 24.942 22.162 21.988 3.519 3.348 3.151 2.994 2.978
48 Cd 26.641 26.093 23.172 22.982 3.716 3.528 3.316 3.133 3.127
49 In 27.859 27.274 24.207 24.000 3.920 3.713 3.487 3.287 3.279
50 Sn 29.106 28.483 25.270 25.042 4.131 3.904 3.662 3.444 3.436
Copyright 2002 Marcel Dekker, Inc.
Table 4 Continued
Atomic
number Element Kb
2
Kb
1
Ka
1
Ka
2
Lg
1
Lb
2
Lb
1
La
1
La
2
51 Sb 30.387 29.723 26.357 26.109 4.347 4.100 3.543 3.605 3.595
52 To 31.698 30.993 27.471 27.200 4.570 4.301 4.029 3.769 3.758
53 I 33.016 32.292 28.610 28.315 4.800 4.507 4.220 3.937 3.926
54 Xe 34.446 33.644 29.802 29.485 5.036 4.720 4.422 4.111 4.098
55 Cs 35.819 34.984 30.970 30.623 5.280 4.936 4.620 4.286 4.272
56 Ba 37.255 35.376 32.191 31.815 5.531 5.156 4.828 4.467 4.451
57 La 38.728 37.799 33.440 33.033 5.789 5.384 5.043 4.651 4.635
58 Ce 40.231 39.255 34.717 34.276 6.052 5.613 5.262 4.840 4.823
59 Pr 41.772 40.746 36.023 35.548 6.322 5.850 5.489 5.034 5.014
60 Nd 43.298 42.269 37.359 36.845 6.602 6.090 5.722 5.230 5.208
61 Pm 44.955 43.945 38.649 38.160 6.891 6.336 5.956 5.431 5.408
62 Sm 46.553 45.400 40.124 39.523 7.180 6.587 6.206 5.636 5.609
63 Eu 48.241 47.027 41.529 40.877 7.478 6.842 6.456 5.846 5.816
64 Gd 49.961 48.718 42.983 42.280 7.788 7.102 6.714 6.039 6.027
65 Tb 51.737 50.391 44.470 43.737 8.104 7.368 6.979 6.275 6.241
66 Dy 53.491 52.178 45.985 45.193 8.418 7.638 7.249 6.495 6.457
67 Ho 55.292 53.934 47.528 46.686 8.748 7.912 7.528 6.720 6.680
68 Er 57.088 55.690 49.099 48.205 9.089 8.188 7.810 6.948 6.904
69 Tm 58.969 57.576 50.730 49.762 9.424 8.472 8.103 7.181 7.135
70 Yb 60.959 59.352 52.360 51.326 9.779 8.758 8.401 7.414 7.367
71 Lu 62.946 61.282 54.063 52.959 10.142 9.048 8.709 7.654 7.604
72 Hf 64.936 63.209 55.757 54.579 10.514 9.346 9.021 7.898 7.843
73 Ta 66.999 65.210 57.524 56.270 10.892 9.649 9.341 8.145 8.087
74 W 69.090 67.233 59.310 57.973 11.283 9.959 9.670 8.396 8.333
75 Re 71.220 69.298 61.131 59.707 11.684 10.273 10.008 8.651 8.584
76 Os 73.393 71.404 62.991 61.477 12.094 10.596 10.354 8.910 8.840
77 Ir 75.605 73.549 64.886 63.278 12.509 10.918 10.706 9.173 9.098
78 Pt 77.866 75.736 66.820 65.111 12.939 11.249 11.069 9.441 9.360
79 Au 80.165 77.968 68.794 66.980 13.379 11.582 11.439 9.711 9.625
80 Hg 82.526 80.258 70.821 68.894 13.828 11.923 11.823 9.987 9.896
81 Tl 84.904 82.558 72.860 70.320 14.288 12.268 12.210 10.266 10.170
82 Pb 87.343 94.922 74.957 72.794 14.762 12.620 12.611 10.549 10.448
83 Bi 89.833 87.335 77.097 74.805 15.244 12.977 13.021 10.836 10.729
84 Po 92.386 89.809 79.296 76.868 15.740 13.338 13.441 11.128 11.014
85 At 94.976 92.319 81.525 78.956 16.248 13.705 13.873 11.424 11.304
86 Rn 97.616 94.877 83.800 81.080 16.768 14.077 14.316 11.724 11.597
87 Fr 100.305 97.483 86.119 83.243 17.301 14.459 14.770 12.029 11.894
88 Ra 103.048 100.136 88.485 85.446 17.845 14.839 15.233 12.338 12.194
89 Ac 105.838 102.846 90.894 87.681 l8.405 l5.227 15.712 12.650 12.499
90 Th 108.671 105.592 93.334 89.942 18.977 15.620 16.200 12.966 12.808
91 Pa 111.575 108.408 95.851 92.271 19.559 16.022 16.700 13.291 13.120
92 U 114.549 111.289 98.428 94.648 20.163 16.425 17.218 13.613 13.438
93 Np 117.533 114.181 101.005 97.023 20.774 16.837 17.740 13.945 12.758
94 Pu 120.592 117.146 103.653 99.457 21.401 17.254 18.278 14.279 14.082
95 Am 123.706 120.163 106.351 101.932 22.042 17.667 18.829 14.618 14.411
96 Cm 126.875 123.235 109.098 104.448 22.699 18.106 19.393 14.961 14.743
97 Bk 130.101 126.362 111.896 107.023 23.370 18.540 19.971 15.309 15.079
98 Cf 133.383 129.544 114.745 109.603 24.056 18.980 20.562 15.661 15.420
99 Es 136.724 132.781 118.646 112.244 24.758 19.426 21.l66 16.018 15.764
100 Fm 140.122 136.075 120.598 114.926 25.475 19.879 21.785 16.379 16.113
Copyright 2002 Marcel Dekker, Inc.
APPENDIXIII: RADIATIVE TRANSITIONPROBABILITIES
Table1 Radiative Transition Probabilities for K X-ray Lines
Atomic
number Element Ka
2
=Ka
1
Kb
3
=Kb
1
(Kb
1
Kb
3
)=Ka
1
Kb
e
a
1
=Ka
1
Kb
e
b
2
=Ka
1
Kb
c
=Ka
d
20 Ca 0.505 0.116 0.116 0.069
22 Ti 0.505 0.137 0.137 0.095
24 Cr 0.506 0.155 0.156 0.114
26 Fe 0.506 0.172 0.171 0.128
28 Ni 0.507 0.189 0.187 0.133
30 Zn 0.509 0.202 0.202 0.137
32 Ge 0.511 0.215 0.215 0.142
34 Sc 0.513 0.225 0.225 0.006 0.153
36 Kr 0.515 0.235 0.235 0.013 0.164
38 Sr 0.518 0.244 0.244 0.022 0.175
40 Zr 0.520 0.251 0.252 0.034 0.185
42 Mo 0.523 0.258 0.259 0.043 0.193
44 Ru 0.526 0.264 0.265 0.048 0.201
46 Pd 0.528 0.270 0.271 0.051 0.209
48 Cd 0.531 0.275 0.277 0.054 0.216
50 Sn 0.533 0.516 0.280 0.282 0.056 0.222
52 Te 0.536 0.517 0.285 0.287 0.060 0.226
54 Xe 0.537 0.518 0.290 0.292 0.064 0.232
56 Ba 0.542 0.519 0.294 0.297 0.670 0.240
58 Ce 0.545 0.521 0.298 0.301 0.076 0.244
60 Nd 0.549 0.522 0.303 0.306 0.082 0.247
62 Sm 0.551 0.523 0.307 0.311 0.085 0.250
64 Gd 0.556 0.525 0.310 0.314 0.088 0.253
66 Dy 0.560 0.526 0.314 0.318 0.089 0.256
68 Er 0.565 0.527 0.317 0.322 0.090 0.259
70 Yb 0.568 0.529 0.320 0.325 0.090 0.261
72 Hf 0.572 0.531 0.324 0.329 0.091 0.263
74 W 0.576 0.532 0.326 0.332 0.092 0.267
76 Os 0.580 0.534 0.330 0.336 0.094 0.270
78 Pt 0.585 0.535 0.333 0.339 0.097 0.274
80 Hg 0.590 0.537 0.336 0.343 0.100 0.277
82 Pb 0.595 0.539 0.339 0.346 0.103 0.282
84 Po 0.600 0.541 0.342 0.350 0.106 0.285
86 Rn 0.605 0.542 0.345 0.353 0.110 0.288
88 Ra 0.612 0.544 0.348 0.356 0.113 0.291
90 Th 0.619 0.546 0.351 0.360 0.118 0.295
92 U 0.624 0.548 0.354 0.363 0.123 0.299
94 Pu 0.631 0.550 0.356 0.366 0.125 0.301
96 Cm 0.638 0.552 0.359 0.370 0.130 0.305
98 Cf 0.646 0.554 0.362 0.374 0.134 0.309
100 Em 0.652 0.556 0.364 0.377 0.138 0.312
a
Kb
e
1
= KM
2
KM
3
KM
4;5
:
b
Kb
e
2
= KN
2;3
KO
2;3
:
c
Kb = Kb
e
1
Kb
e
2
:
d
Ka = Ka
1
Ka
2
:
Source: From West, 198283.
Copyright 2002 Marcel Dekker, Inc.
Table 2 Radiative Transition Probabilities for L
1
X-ray Lines Normalized to Lb
3
=100
Atomic
number Element Lb
3
Lb
4
Lg
3
Lg
2
36 Kr 100 10.8
38 Sr 100 14.7
40 Zr 100 18.1
42 Mo 100 71.0 21.0
44 Ru 100 64.9 23.4
46 Pd 100 61.9 25.5
48 Cd 100 61.0 27.2
50 Sn 100 61.4 28.5
52 Te 100 62.6 29.5
54 Xe 100 64.1 30.3
56 Ba 100 65.5 30.8
58 Ce 100 66.7 31.2
60 Nd 100 67.5 31.4
62 Sm 100 69.7 31.5
64 Gd 100 68.0 31.5
66 Dy 100 67.8 31.4
68 Er 100 67.6 31.4
70 Yb 100 67.5 31.4 18.5
72 Hf 100 67.6 31.4 20.0
74 W 100 68.3 31.6 21.5
76 Os 100 69.6 31.9 23.2
78 Pt 100 71.7 32.3 25.0
80 Hg 100 74.8 33.0 26.9
82 Pb 100 78.8 34.0 28.8
84 Po 100 83.9 35.2 30.9
86 Rn 100 89.8 36.8 33.1
88 Ra 100 95.0 38.7 35.3
90 Th 100 102.0 41.0 37.7
92 U 100 110.0 43.8 40.2
94 Pu 100 120.0 47.5 44.0
Copyright 2002 Marcel Dekker, Inc.
Table 3 L
2
X-ray Lines Normalized to Lb
1
=100
Atomic
number Element Lb
1
LZ Lg
1
Lg
6
30 Zn 100 12.3
32 Ge 100 10.4
34 Se 100 8.75
36 Kr 100 7.40
38 Sr 100 6.25
40 Zr 100 5.25 0.91
42 Mo 100 4.35 6.71
44 Ru 100 3.63 10.6
46 Pd 100 3.0 13.1
48 Cd 100 2.6 14.5
50 Sn 100 2.35 15.3
52 Te 100 2.25 15.4
54 Xe 100 2.2 15.6
56 Ba 100 2.16 15.6
58 Ce 100 2.12 15.7
60 Nd 100 2.10 15.8
62 Sn 100 2.10 15.9
64 Gd 100 2.10 16.1
66 Dy 100 2.10 16.5
68 Er 100 2.10 17.0
70 Yb 100 2.12 17.6
72 Hf 100 2.13 18.4
74 W 100 2.16 19.2 0.375
76 Os 100 2.20 20.1 1.73
78 Pt 100 2.23 20.9 2.42
80 Hg 100 2.28 21.7 2.98
82 Pb 100 2.33 22.3 3.45
84 Po 100 2.40 22.8 3.88
86 Rn 100 2.45 23.2 4.29
88 Ra 100 2.50 23.4 4.74
90 Th 100 2.60 23.7 5.25
92 U 100 2.80 24.0 5.88
94 Pu 100 2.30 24.2 6.65
Copyright 2002 Marcel Dekker, Inc.
Table 4 L
3
X-ray Lines Normalized to La
1
=100
Atomic number Element La
1
Lb
2,15
La
2
Lb
5
Lb
6
Ll
22 Ti 100 40.37
24 Cr 100 26.13
26 Fe 100 15.35
28 Ni 100 10.29
30 Zn 100 7.56
32 Ge 100 5.96
34 Se 100 4.98
36 Kr 100 4.36
38 Sr 100 3.98
40 Zr 100 2.43 3.75
42 Mo 100 6.40 12.5 3.65
44 Ru 100 9.55 12.2 3.58
46 Pd 100 12.1 12.1 3.55
48 Cd 100 13.9 11.9 3.56
50 Sn 100 15.4 11.7 5.59
52 Te 100 16.4 11.5 3.62
54 Xe 100 17.2 11.3 3.67
56 Ba 100 17.8 11.2 3.73
58 Ce 100 18.2 11.1 3.79
60 Nd 100 18.5 11.1 3.86
62 Sm 100 18.8 11.0 3.92
64 Gd 100 19.2 11.1 3.99
66 Dy 100 19.6 11.1 4.07
68 Er 100 20.0 11.2 4.15
70 Yb 100 20.5 11.2 4.23
72 Hf 100 21.2 11.3 1.15 4.32
74 W 100 21.9 11.3 0.242 1.28 4.42
76 Os 100 22.7 11.4 0.873 1.38 4.53
78 Pt 100 23.5 11.4 1.74 1.46 4.65
80 Hg 100 24.4 11.5 2.62 1.55 4.78
82 Pb 100 25.3 11.5 3.24 1.59 4.93
84 Po 100 26.2 11.4 3.85 1.65 5.09
86 Rn 100 26.8 11.4 4.28 1.70 5.27
88 Ra 100 27.3 11.3 4.69 1.75 5.46
90 Th 100 27.5 11.1 4.94 1.80 5.69
92 U 100 27.5 11.0 5.20 1.85 5.93
94 Pu 100 27.0 10.5 5.40 1.89 6.18
Source: From West, 198283.
Copyright 2002 Marcel Dekker, Inc.
APPENDIXIV: NATURALWIDTHS OF K ANDL LEVELS ANDK
a
X-RAYLINES
(FWHM), INeV
Level X-ray line
Element K L
1
L
2
L
3
Ka
1
Ka
2
10 Ne 0.24 ( <0.1) (0.000) (0.000) 0.24 0.24
11 Na 0.30 (0.2) (0.000) (0.000) 0.30 0.30
12 Mg 0.36 0.41 0.00l 0.001 0.36 0.36
13 Al 0.42 0.73 0.004 0.004 0.43 0.43
14 Si 0.48 1.03 0.015 0.014 0.49 0.49
15 P 0.53 1.26 0.032 0.033 0.57 0.56
16 S 0.59 1.49 0.054 0.054 0.65 0.64
17 Cl 0.64 1.58 0.083 0.087 0.72 0.72
18 Ar 0.68 1.63 0.126 0.128 0.81 0.80
19 K 0.74 1.92 0.152 0.156 0.89 0.89
20 Ca 0.81 2.07 0.17 0.17 0.98 0.98
21 Sc 0.86 2.21 0.19 0.19 1.05 1.06
22 Ti 0.94 2.34 0.24 0.22 1.16 1.18
23 V 1.01 2.41 0.26 0.24 1.26 1.28
24 Cr 1.08 2.54 0.29 0.27 1.35 1.37
25 Mn 1.16 2.62 0.34 0.32 1.48 1.50
26 Fe 1.25 2.76 0.37 0.36 1.61 1.62
27 Co 1.33 2.79 0.43 0.43 1.76 1.76
28 Ni 1.44 2.89 0.52 0.48 1.94 1.96
29 Cu 1.55 3.06 0.62 0.56 2.11 2.17
30 Zn 1.67 3.28 0.72 0.65 2.32 2.39
31 Ga 1.82 3.38 0.83 0.76 2.59 2.66
32 Ge 1.96 3.53 0.95 0.82 2.78 2.92
33 As 2.14 3.79 1.03 0.94 3.08 3.17
34 Se 2.33 3.94 1.13 1.00 3.33 3.46
35 Br 2.52 4.11 1.21 1.08 3.60 3.73
36 Kr 2.75 4.28 1.31 1.17 3.92 4.06
37 Rb 2.99 4.44 1.43 1.27 4.26 4.92
38 Sr 3.25 4.67 1.54 1.39 4.63 4.79
39 Y 3.52 4.71 1.65 1.50 5.02 5.18
40 Zr 3.84 4.78 1.78 1.57 5.40 5.62
41 Nb 4.14 3.94 1.87 1.66 5.80 6.01
42 Mo 4.52 4.25 1.97 1.78 6.31 6.49
43 Tc 4.91 4.36 2.08 1.91 6.82 6.99
44 Ru 5.33 4.58 2.23 2.00 7.33 7.56
45 Rh 5.77 4.73 2.35 2.13 7.90 8.12
46 Pd 6.24 4.93 2.43 2.25 8.49 8.67
47 Ag 6.75 4.88 2.57 2.40 9.16 9.32
48 Cd 7.28 4.87 2.62 2.50 9.79 9.91
49 In 7.91 5.00 2.72 2.65 10.56 10.63
50 Sn 8.49 2.97 2.84 2.75 11.2 11.3
51 Sb 9.16 3.13 3.00 2.87 12.0 12.2
52 Te 9.89 3.32 3.12 2.95 12.8 13.0
53 I 10.6 3.46 3.25 3.08 13.7 13.8
Copyright 2002 Marcel Dekker, Inc.
Appendix IV Continued
Level X-ray line
Element K L
1
L
2
L
3
Ka
1
Ka
2
54 Xe 11.4 3.64 3.40 3.13 14.6 14.8
55 Cs 12.3 3.78 3.51 3.25 15.6 15.8
56 Ba 13.2 3.92 3.57 3.32 16.5 16.8
57 La 14.1 4.06 3.68 3.41 17.6 17.8
58 Ce 15.1 4.21 3.80 3.48 I8.6 18.9
59 Pr 16.2 4.34 3.89 3.60 19.8 20.1
60 Nd 17.3 4.52 3.97 3.65 20.9 21.3
61 Pm 18.5 4.67 4.06 3.75 22.2 22.5
62 Sm 19.7 4.80 4.15 3.86 23.6 23.8
63 Eu 21.0 4.91 4.23 3.91 24.9 25.2
64 Gd 22.3 5.05 4.32 4.01 26.4 26.7
65 Tb 23.8 5.19 4.43 4.12 27.9 28.2
66 Dy 25.2 5.25 4.55 4.17 29.4 29.8
67 Ho 26.8 5.33 4.66 4.26 31.1 31.5
68 Er 28.4 5.43 4.73 4.35 32.7 33.1
69 Tm 30.1 5.47 4.79 4.48 34.6 34.9
70 Yb 31.9 5.53 4.82 4.60 36.5 36.7
71 Lu 33.7 5.54 4.92 4.68 38.4 38.7
72 Hf 35.7 5.63 5.02 4.80 40.5 40.7
73 Ta 37.7 5.58 5.15 4.88 42.6 42.9
74 W 39.9 5.61 5.33 4.98 44.9 45.2
75 Re 42.1 6.18 5.48 5.04 47.2 47.6
76 Os 44.4 7.25 5.59 5.16 49.6 50.0
77 Ir 46.8 8.30 5.69 5.25 52.1 52.5
78 Pt 49.3 9.39 5.86 5.31 54.6 55.2
79 Au 52.0 10.5 6.00 5.41 57.4 58.0
80 Hg 54.6 11.3 6.17 5.50 60.1 60.8
81 Tl 57.4 12.0 6.32 5.65 63.1 63.8
82 Pb 60.4 12.2 6.48 5.81 66.2 66.8
83 Bi 63.4 12.4 6.67 5.98 69.4 70.1
84 Po 66.6 12.6 6.83 6.13 72.7 73.4
85 At 69.8 12.8 7.01 6.29 76.1 76.8
86 Rn 73.3 13.1 7.20 6.41 79.7 80.5
87 Fr 76.8 13.3 7.47 6.65 83.4 84.2
88 Ra 80.4 13.4 7.68 6.82 87.2 88.1
89 Ac 84.1 13.6 7.95 6.98 91.1 92.0
90 Th 88.0 13.7 8.18 7.13 95.2 96.2
91 Pa 91.9 14.3 8.75 7.33 99.3 100.7
92 U 96.1 14.0 9.32 7.43 103.5 105.4
93 Np 100 14.0 9.91 7.59 108 110
94 Pu 105 13.5 10.5 7.82 113 115
95 Am 109 13.3 10.9 8.04 117 120
96 Cm 114 13.6 11.4 8.26 122 125
97 Bk 119 13.8 11.8 8.55 127 130
98 Cf 124 14.0 12.2 8.75 132 136
99 Es 129 14.3 12.7 9.04 138 141
Copyright 2002 Marcel Dekker, Inc.
APPENDIXIV Continued
Level X-ray line
Element K L
1
L
2
L
3
Ka
1
Ka
2
100 Fm 135 14.4 13.1 9.33 144 148
101 Md 140 14.8 13.6 9.61 150 153
102 No 145 15.1 14.0 9.90 155 159
103 Lr 150 15.9 14.4 10.1 161 165
104 156 16.2 14.9 10.5 167 171
105 162 16.5 15.5 10.8 173 177
106 168 16.8 15.9 11.2 179 184
107 174 17.0 16.4 11.6 185 190
108 181 17.2 17.0 12.0 193 198
109 187 17.6 17.5 12.3 200 205
110 193 18.1 18.1 12.7 206 211
Source: From Krause and Oliver, 1979.
Copyright 2002 Marcel Dekker, Inc.
APPENDIXV: WAVELENGTHS OF KSATELLITE LINES (A )
Atomic
number Element a
II
a
I
a
3
a
3
I
a
3
II
a
4
a
4
I
a
5
a
6
a
7
b
0
b
I
11 Na 11.860 11.830 11.810 11.742 11.711
12 Mg 9.845 9.820 9.804 9.749 9.724
13 Al 8.302 8.283 8.267 8.226 8.206
14 Si 7.100 7.095 7.082 7.073 7.069 7.035 7.017 7.026 6.816
15 P 6.140 6.131 6.117 6.109 6.075 6.063 6.838
16 S 5.370 5.354 5.341 5.339 5.344 5.334
17 Cl 4.725 4.415
18 Ar
19 K 3.739 3.728 3.721 3.718 3.724 3.717 3.496
20 Ca 3.356 3.347 3.340 3.338 3.343 3.337 3.133 3.101
21 Sc 3.029 3.021 3.015 3.013 3.018 3.013 2.819 2.789
22 Ti 2.747 2.740 2.733 2.731 2.737 2.731 2.551 2.522
23 V 2.502 2.496 2.491 2.489 2.493 2.490 2.320 2.291
24 Cr 2.282 2.279 2.277 2.280 2.278 2.118 2.090
25 Mn 2.095 2.093 2.091 2.094 2.091 1.914
26 Fe 1.930 1.928 1.926 1.929 1.927 1.783 1.760
27 Co 1.784 1.782 1.780 1.783 1.781 1.645 1.623
28 Ni 1.653 1.651 1.650 1.652 1.650 1.522 1.502
29 Cu 1.536 1.535 1.533 1.534 1.394
30 Zn 1.431 1.430 1.428 1.429 1.296
31 Ga 1.138 1.335 1.334 1.335 1.209
32 Ge 1.250 1.250 1.248 1.249
33 As 1.173 1.172 1.170 1.171 1.061
34 Se 1.101 1.100 1.101 0.9958
35 Br 1.036 1.035 1.036 0.9367
36 Kr 0.8832
37 Rb 0.8835
38 Sr 0.8727 0.8646 0.8721 0.7881
39 Y 0.7457
40 Zr 0.7836 0.7828 0.7832 0.7056
41 Nb 0.7432 0.7436 0.6706
42 Mo 0.7065 0.7070 0.6369
43 Tc
44 Ru 0.5766
45 Rh 0.5493
46 Pd 0.5833 0.5837
47 Ag 0.5007
Source: From Clark, 1963.
Copyright 2002 Marcel Dekker, Inc.
b
1x
b
x
b
V
b
II
b
II
III
b
III
b
IV
b
5
b
6
b
7
b
8
b
9
b
10
6.753
5.800 5.792 5.712 5.691
5.028 5.023
4.400 4.395
3.882
3.449 3.441 3.412 3.404
3.087 3.082 3.054 3.048
2.772 2.749 2.744
2.506 2.489
2.277 2.262
2.078 2.066
1.895
1.749 1.742
1.614 1.607
1.494 1.487
1.391 1.380
1.295 1.282
1.207
1.128 1.116
1.042 1.044 1.064 1.050 1.054 1.047
0.9770 0.9786 0.9996 0.9854 0.9889
0.9177 0.9194 0.9403 0.9629 0.9297 0.9228
1.8745
0.8135 0.8155 0.8365 0.8234 0.8259 0.8193
0.7681 0.7699 0.7911 0.7778 0.7802 0.7736
0.7261 0.7218 0.7492 0.7362 0.7382 0.7315
0.6870 0.6890 0.7105 0.6973 0.6993 0.6924
0.6510 0.6531 0.6741 0.6619 0.6636 0.6568
0.6288 0.6303 0.6235 0.6256 0.6519
0.5589 0.5808 0.5693 0.5708 0.5639 0.5607
0.5427 0.5439 0.5374 0.5399 0.5452
0.5189
0.4943 0.4954 0.4902 0.4920
Copyright 2002 Marcel Dekker, Inc.
APPENDIXVI: FLUORESCENCE YIELDS AND COSTER^KRONIGTRANSITION
PROBABILITIES
Table1 K Shell Fluorescence Yield o
K
Atomic number Element o
K
Atomic number Element o
K
6 C 0.0009 45 Rh 0.81
7 N 0.0015 46 Pd 0.82
8 O 0.0022 47 Ag 0.83
10 Ne 0.0100 48 Cd 0.84
11 Na 0.020 49 In 0.85
12 Mg 0.030 50 Sn 0.86
13 Al 0.040 51 Sb 0.87
14 Si 0.055 52 Te 0.875
15 P 0.070 53 I 0.88
16 S 0.090 54 Xe 0.89
17 Cl 0.105 55 Cs 0.895
18 Ar 0.125 56 Ba 0.90
19 K 0.140 57 La 0.905
20 Sc 0.165 58 Ce 0.91
21 Ca 0.190 59 Pr 0.915
22 Ti 0.220 60 Nd 0.92
23 V 0.240 61 Pm 0.925
24 Cr 0.26 62 Sm 0.93
25 Mn 0.285 63 Eu 0.93
26 Fe 0.32 64 Gd 0.935
27 Co 0.345 65 Tb 0.94
28 Ni 0.375 66 Dy 0.94
29 Cu 0.41 67 Ho 0.945
30 Zn 0.435 68 Er 0.945
31 Ga 0.47 69 Tm 0.95
32 Ge 0.50 70 Yb 0.95
33 As 0.53 71 In 0.95
34 Se 0.565 72 Hf 0.955
35 Br 0.60 73 Ta 0.955
36 Kr 0.635 74 W 0.96
37 Rb 0.665 75 Re 0.96
38 Sr 0.685 76 Os 0.96
39 Y 0.71 77 Ir 0.96
40 Zr 0.72 78 Pd 0.965
41 Nb 0.755 79 Au 0.965
42 Mo 0.77 80 Hg 0.965
43 Tc 0.785 82 Pb 0.97
44 Ru 0.80 92 U 0.97
Source: From Birks, 1971b and Fink, 1974.
Copyright 2002 Marcel Dekker, Inc.
Table 2 Experimental L Subshell Fluorescence Yields o
i
Atomic number Element o
1
o
2
o
3
54 Xe 0.10 0.01
56 Ba 0.06 0.05 0.01
65 Tb 0.18 0.165 0.018 0.188 0.016
67 Ho 0.22 0.03
0.170 0.055 0.169 0.030
68 Er 0.21 0.03
0.185 0.060 0.172 0.032
70 Yb 0.20 0.02
0.188 0.011 0.183 0.011
71 Lu 0.22 0.03
0.251 0.035
72 Hf 0.22 0.03
0.228 0.025
73 Ta 0.25 0.02 0.27 0.01
0.257 0.013 0.25 0.03
0.191
0.228 0.013
0.254 0.025
74 W 0.207
0.272 0.037
75 Re 0.284 0.043
76 Os 0.290 0.030
77 Ir 0.244
0.262 0.036
78 Pt 0.331 0.021 0.262
0.31 0.04
0.317 0.029
0.291 0.018
79 Au 0.276
0.31 0.04
0.317 0.025
80 Hg 0.39 0.03 0.40 0.02
0.319 0.010 0.32 0.05
0.367 0.050
0.300 0.010
81 Tl 0.07 0.02 0.319 0.010 0.37 0.07
0.373 0.025 0.386 0.053
0.306 0.010
0.330 0.021
82 Pb 0.07 0.02 0.363 0.015 0.337
0.09 0.02 0.315 0.013
0.32
0.35 0.05
0.354 0.028
83 Bi 0.12 0.01 0.32 0.04 0.367
0.095 0.005 0.38 0.02 0.36
0.37 0.05
0.362 0.029
0.40 0.05
Copyright 2002 Marcel Dekker, Inc.
Table 2 Continued
Atomic number Element o
1
o
2
o
3
0.340 0.018
90 Th 0.42
0.517 0.042
91 Pa 0.46 0.05
92 U 0.44
0.500 0.040
96 Cm 0.28 0.06 0.552 0.032 0.515 0.034
0.55 0.02 0.63 0.02
Source: From Bambynek et al., 1972.
Table 3 Measured L Shell CosterKronig Yields
Atomic number Element f
12
f
13
f
23
56 Ba 0.66 0.07
65 Tb 0.41 0.36 0.43 0.28 0.066 0.014
67 Ho 0.205 0.034
68 Er 0.255 0.025
70 Yb 0.142 0.009
73 Ta <0.14 0.19 0.148 0.010
<0.36 0.20 0.04
74 W 0.27 0.03
75 Re 0.30 0.04
77 Ir 0.46 0.06
78 Pt 0.50 0.05
79 Au 0.25 0.13 0.51 0.13 0.22
0.61 0.07
80 Hg 0.74 0.04 0.22 0.04
0.08 0.02
0.188 0.010
81 Tl 0.17 0.05 0.76 0.10 0.25 0.13
0.14 0.03 0.57 0.10 0.169 0.010
0.56 0.07 0.159 0.013
0.56 0.05
82 Pb 0.15 0.04 0.57 0.03 0.164 0.016
0.17 0.05 0.61 0.08 0.156 0.010
83 Bi 0.19 0.05 0.58 0.05 0.14
0.06
70.06
0.18 0.02 0.58 0.02 0.164
92 U 0.23 0.12
93 Np 0.10 0.04 0.55 0.09 0.05
0.02
70.02
94 Pu 0.22 0.08
0.24 0.08
96 Cm 0.038 0.022 0.68 0.04 0.188 0.019
Source: From Bambynek et al., 1972.
Copyright 2002 Marcel Dekker, Inc.
Table 4 Theoretical L Subshell Fluorescence Yields o
i
and CosterKronig Yields f
ij
a
Atomic number Element o
1
o
2
o
3
f
12
f
13
f
12
f
13
f
23
13 Al 3.05-6 2.40-3 0.982
14 Si 9.77-6 1.08-3 0.975
15 P 2.12-5 4.1-4 0.971
16 S 3.63-5 2.9-4 0.968
17 Cl 5.60-5 2.3-4 0.964
18 Ar 8.58-5 1.9-4 0.965
19 K 1.15-4 2.1-4 0.962
20 Ca 1.56-4 2.1-4 0.955
22 Ti 2.80-4 1.18-3 0.313 0.629
24 Cr 2.97-4 3.29-3 0.317 0.636
26 Fe 3.84-4 1.43-3 5.59-3 0.302 0.652 7.24-2
1.49-3
28 Ni 4.63-4 2.69-3 8.02-3 0.325 0.622 9.97-2
29 Cu 3.57-3 3.83-3 0.109
30 Zn 5.23-4 1.08-2 0.322 0.624
32 Ge 7.70-4 7.72-3 1.44-2 0.266 0.671 2.49-2
33 As 1.40-3 8.85-3 9.74-3 0.282 0.547 4.13-2
34 Se 1.30-3 9.94-3 1.78-2 0.302 0.616 5.95-2
35 Br 1.09-2 7.64-2
36 Kr 1.85-3 2.20-2 2.36-2 0.230 0.686 8.97-2
2.19-3 1.19-2 1.23-2 0.225 0.585 9.22-2
37 Rb 1.32-2 0.107
38 Sr 3.00-3 2.24-2 2.43-2 0.249 0.646 0.115
40 Zr 3.97-3 2.94-2 2.95-2 0.236 0.648 0.118
3.96-3 1.89-2 2.01-2 0.271 0.522 0.123
42 Mo 5.75-3 3.50-2 3.73-2 0.166 0.689 0.124
6.34-3 2.45-2 2.59-2 0.048 0.692 0.126
44 Ru 7.74-3 4.18-2 4.50-2 0.057 0.779 0.136
47 Ag 1.02-2 5.47-2 6.02-2 0.052 0.786 0.152
1.01-2 4.30-2 4.49-2 0.064 0.695 0.130
50 Sn 1.30-2 6.56-2 7.37-2 0.052 0.784 0.162
1.30-2 5.67-2 0.072 0.693 0.136
51 Sb 3.11-2 6.16-2 6.33-2 0.164 0.316 0.138
54 Xe 5.84-2 9.12-2 9.70-2 0.179 0.274 0.173
56 Ba 4.46-2 9.07-2 8.99-2 0.168 0.336 0.151
60 Nd 7.46-2 0.133 0.135 0.207 0.303 0.141
6.00-2 0.120 0.120 0.165 0.332 0.142
65 Tb 0.166 0.160 0.131
67 Ho 0.112 0.203 0.201 0.202 0.309 0.138
0.094 0.178 0.317
70 Yb 0.112 0.180 0.316
74 W 0.115 0.287 0.268 0.195 0.332 0.123
0.138 0.271 0.253 0.160 0.324 0.117
79 Au 0.105 0.357 0.327 0.083 0.644 0.132
80 Hg 0.098 0.352 0.321 0.101 0.618 0.108
83 Bi 0.120 0.417 0.389 0.069 0.656 0.101
85 At 0.129 0.422 0.380 0.082 0.612 0.100
90 Th 0.197 0.529 0.461 0.069 0.575 0.102
93 Np 0.460 0.472 0.209
a
Figures following a sign indicate powers of 10. For example, 3.05-6 means 3.05610
76
.
Source: From Bambynek et al., 1972. Copyright 2002 Marcel Dekker, Inc.
Table 5 Measured M Shell Fluorescence Yields and CosterKronig Probabilities
Atomic
number Element " o
M
o
LM
a
o
LM
b
o
1
f
12
o
2
n
i
o
i
76 Os 0.013 0.0024 0.016 0.003
79 Au 0.023 0.001
79 Au 0.024 0.005 0.030 0.006
82 Pb 0.029 0.002
82 Pb 0.026 0.005 0.032 0.006
83 Bi 0.037 0.007
83 Bi 0.035 0.002
83 Bi 0.030 0.006 0.037 0.005
92 U 0.06
93 Np 0.003 n
1
=0.065 0.014
0.002
70.002 n
2
=0.0080 0.029
n
3
=0.062 0.005
n
4
=0.065 0.012 o
5
=0.06 0.012
n
4
=0.065 0.012
96 Cm 0.0089 n
4,5
=0.081 0.016
0.0075 0.0051
70.0075 n
1
=0.068 0.023 o
2
=0.0046
n
2
=0.062 0.019 70.0046
n
3
=0.080 0.006
n
4,5
=0.075 0.012
o
5
=0.075 0.012
a
Corrected for a 20% contribution from double M shell vacancies.
b
Uncorrected values.
Source: From Bambynek et al., 1972.
Copyright 2002 Marcel Dekker, Inc.
Table 6 Adopted Values of Fluorescence and Coster-Kronig Yields
a
Z o
k
o
1
o
2
o
3
f
1
f
12
f
13
f
23
5 B 1.7E-03
6 C 2.8E-03
7 N 5.2E-03
8 O 8.3E-03
9 F 0.013
10 Ne 0.018
11 Na 0.023
12 Mg 0.030 2.9E-05 1.2E-03 1.2E-03 0.962 0.32 0.64
13 Al 0.039 2.6E-05 7.5E-04 7.5E-04 0.965 0.32 0.64
14 Si 0.050 3.0E-05 3.7E-04 3.8E-04 0.959 0.32 0.64
15 P 0.063 3.9E-05 3.1E-04 3.1E-04 0.951 0.32 0.63
16 S 0.078 7.4E-05 2.6E-04 2.6E-04 0.944 0.32 0.62
17 Cl 0.097 1.2E-04 2.4E-04 2.4E-04 0.939 0.32 0.62
18 Ar 0.118 1.8E-04 2.2E-04 2.2E-04 0.934 0.31 0.62
19 K 0.140 2.4E-04 2.7E-04 2.7E-04 0.929 0.31 0.62
20 Ca 0.163 3.1E-04 3.3E-04 3.3E-04 0.920 0.31 0.61
21 Sc 0.188 3.9E-04 8.4E-04 8.4E-04 0.912 0.31 0.60
22 Ti 0.214 4.7E-04 1.5E-03 1.5E-03 0.902 0.31 0.59
23 V 0.243 5.8E-04 2.6E-03 2.6E-03 0.894 0.31 0.58
24 Cr 0.275 7.1E-04 3.7E-03 3.7E-03 0.885 0.31 0.57
25 Mn 0.308 8.4E-04 5.0E-03 5.0E-03 0.877 0.30 0.58
26 Fe 0.340 1.0E-03 6.3E-03 6.3E-03 0.868 0.30 0.57
27 Co 0.373 1.2E-03 7.7E-03 7.7E-03 0.856 0.30 0.56
28 Ni 0.406 1.4E-03 8.6E-03 9.3E-03 0.847 0.30 0.55 0.028
29 Cu 0.440 1.6E-03 0.010 0.011 0.839 0.30 0.54 0.028
30 Zn 0.474 1.8E-03 0.011 0.012 0.831 0.29 0.54 0.026
31 Ga 0.507 2.1E-03 0.012 0.013 0.822 0.29 0.53 0.032
32 Ge 0.535 2.4E-03 0.013 0.015 0.815 0.28 0.53 0.050
33 As 0.562 2.8E-03 0.014 0.016 0.809 0.28 0.53 0.063
34 Se 0.589 3.2E-03 0.016 0.018 0.804 0.28 0.52 0.076
35 Br 0.618 3.6E-03 0.018 0.020 0.800 0.28 0.52 0.088
36 Kr 0.643 4.1E-03 0.020 0.022 0.797 0.27 0.52 0.100
37 Rb 0.667 4.6E-03 0.022 0.024 0.794 0.27 0.52 0.109
38 Sb 0.690 5.1E-03 0.024 0.026 0.790 0.27 0.52 0.117
39 Y 0.710 5.9E-03 0.026 0.028 0.785 0.26 0.52 0.126
40 Zr 0.730 6.8E-03 0.028 0.031 0.779 0.26 0.52 0.132
41 Nb 0.747 9.4E-03 0.031 0.034 0.713 0.10 0.61 0.137
42 Mo 0.765 0.010 0.034 0.037 0.712 0.10 0.61 0.141
43 Tc 0.780 0.011 0.037 0.040 0.711 0.10 0.61 0.144
44 Ru 0.794 0.012 0.040 0.043 0.709 0.10 0.61 0.148
45 Rh 0.808 0.013 0.043 0.046 0.705 0.10 0.60 0.150
46 Pd 0.820 0.014 0.047 0.049 0.700 0.10 0.60 0.151
47 Ag 0.831 0.016 0.051 0.052 0.694 0.10 0.59 0.153
48 Cd 0.843 0.018 0.056 0.056 0.688 0.10 0.59 0.155
49 In 0.853 0.020 0.061 0.060 0.681 0.10 0.59 0.157
Copyright 2002 Marcel Dekker, Inc.
Table 6 Continued
Z o
k
o
1
o
2
o
3
f
1
f
12
f
13
f
23
50 Sn 0.862 0.037 0.065 0.064 0.439 0.17 0.27 0.157
51 Sb 0.870 0.039 0.069 0.069 0.448 0.17 0.28 0.156
52 Te 0.877 0.041 0.074 0.074 0.495 0.18 0.28 0.155
53 I 0.884 0.044 0.079 0.079 0.461 0.18 0.28 0.154
54 Xe 0.891 0.046 0.083 0.085 0.466 0.19 0.28 0.154
55 Cs 0.897 0.049 0.090 0.091 0.470 0.19 0.28 0.154
56 Ba 0.902 0.052 0.096 0.097 0.474 0.19 0.28 0.153
57 La 0.907 0.055 0.103 0.104 0.478 0.19 0.29 0.153
58 Ce 0.912 0.058 0.110 0.111 0.482 0.19 0.29 0.153
59 Pr 0.917 0.061 0.117 0.118 0.485 0.19 0.29 0.153
60 Nd 0.921 0.064 0.124 0.125 0.488 0.19 0.30 0.152
61 Pm 0.925 0.066 0.132 0.132 0.490 0.19 0.30 0.151
62 Sm 0.929 0.071 0.140 0.139 0.492 0.19 0.30 0.150
63 Eu 0.932 0.075 0.149 0.147 0.493 0.19 0.30 0.149
64 Gd 0.935 0.079 0.158 0.155 0.493 0.19 0.30 0.147
65 Tb 0.938 0.083 0.167 0.164 0.493 0.19 0.30 0.145
66 Dy 0.941 0.089 0.178 0.174 0.492 0.19 0.30 0.143
67 Ho 0.944 0.044 0.189 0.182 0.490 0.19 0.30 0.142
68 Er 0.947 0.100 0.200 0.192 0.487 0.19 0.30 0.140
69 Tm 0.949 0.106 0.211 0.201 0.483 0.19 0.29 0.139
70 Yb 0.951 0.112 0.222 0.210 0.478 0.19 0.29 0.138
71 Lu 0.953 0.120 0.234 0.220 0.472 0.19 0.28 0.136
72 Hf 0.955 0.128 0.246 0.231 0.465 0.18 0.28 0.135
73 Ta 0.957 0.137 0.258 0.243 0.457 0.18 0.28 0.134
74 W 0.958 0.147 0.270 0.255 0.447 0.17 0.28 0.133
75 Re 0.959 0.144 0.283 0.268 0.485 0.16 0.33 0.130
76 Os 0.961 0.130 0.295 0.281 0.552 0.16 0.39 0.128
77 Ir 0.962 0.120 0.308 0.294 0.603 0.15 0.45 0.126
78 Pt 0.963 0.114 0.321 0.306 0.640 0.14 0.50 0.124
79 Au 0.964 0.107 0.334 0.320 0.672 0.14 0.53 0.122
80 Hg 0.965 0.107 0.347 0.333 0.690 0.13 0.56 0.120
81 Tl 0.966 0.107 0.360 0.347 0.696 0.13 0.57 0.118
82 Pb 0.967 0.112 0.373 0.360 0.696 0.12 0.58 0.116
83 Bi 0.968 0.117 0.387 0.373 0.694 0.11 0.58 0.113
84 Po 0.968 0.122 0.401 0.386 0.689 0.11 0.58 0.111
85 At 0.969 0.128 0.415 0.399 0.685 0.10 0.59 0.111
86 Rn 0.969 0.134 0.429 0.411 0.682 0.10 0.58 0.110
87 Fr 0.970 0.139 0.443 0.424 0.677 0.10 0.58 0.109
88 Ra 0.970 0.146 0.456 0.437 0.672 0.09 0.58 0.108
89 Ac 0.971 0.153 0.468 0.450 0.664 0.09 0.58 0.108
90 Th 0.971 0.161 0.479 0.463 0.660 0.09 0.57 0.108
91 Pa 0.972 0.162 0.472 0.476 0.664 0.08 0.58 0.139
92 U 0.972 0.176 0.467 0.489 0.652 0.08 0.57 0.167
93 Np 0.973 0.187 0.466 0.502 0.642 0.07 0.57 0.192
94 Pu 0.973 0.205 0.464 0.514 0.605 0.05 0.56 0.198
Copyright 2002 Marcel Dekker, Inc.
Table 6 Continued
Z o
k
o
1
o
2
o
3
f
1
f
12
f
13
f
23
95 Am 0.974 0.218 0.471 0.526 0.595 0.05 0.55 0.203
96 Cm 0.974 0.228 0.479 0.539 0.587 0.04 0.55 0.200
97 Bk 0.975 0.236 0.485 0.550 0.580 0.04 0.54 0.198
98 Cf 0.975 0.244 0.490 0.560 0.573 0.03 0.54 0.197
99 Es 0.975 0.253 0.497 0.570 0.565 0.03 0.54 0.196
100 Fm 0.976 0.263 0.506 0.579 0.556 0.03 0.53 0.194
101 Md 0.976 0.272 0.515 0.588 0.548 0.02 0.53 0.191
102 No 0.976 0.280 0.524 0.596 0.540 0.02 0.52 0.189
103 Lr 0.977 0.282 0.533 0.604 0.538 0.01 0.53 0.185
104 0.977 0.291 0.544 0.611 0.531 0.01 0.52 0.181
105 0.977 0.300 0.553 0.618 0.522 0.01 0.51 0.178
106 0.978 0.310 0.562 0.624 0.513 0.51 0.174
107 0.978 0.320 0.573 0.630 0.505 0.50 0.171
108 0.978 0.331 0.584 0.635 0.497 0.50 0.165
109 0.978 0.343 0.590 0.640 0.488 0.49 0.163
110 0.979 0.354 0.598 0.644 0.477 0.48 0.158
a
Note: 1.7E-03 =1.7610
73
; designation for L shell is omitted, read, for example, o
1
= o
L1
; f
1
= f
12
f
13
.
Source: From Krause, 1979.
Copyright 2002 Marcel Dekker, Inc.
5 B 0.998
6 C 0.997
7 N 0.995
8 O 0.992
9 F 0.987
10 Ne 0.982
11 Na 0.977
12 Mg 0.970 0.038 0.999 0.999
13 Al 0.961 0.035 0.999 0.999
14 Si 0.950 0.041 1.000 1.000
15 P 0.937 0.049 1.000 1.000
16 S 0.922 0.056 1.000 1.000
17 Cl 0.903 0.061 1.000 1.000
18 Ag 0.882 0.066 1.000 1.000
19 K 0.860 0.071 1.000 1.000
20 Ca 0.837 0.080 1.000 1.000
21 Sc 0.812 0.088 0.999 0.999
22 Ti 0.786 0.098 0.999 0.999
23 V 0.757 0.105 0.997 0.997
24 Cr 0.725 0.114 0.996 0.996
25 Mn 0.692 0.122 0.995 0.995
26 Fe 0.660 0.131 0.994 0.994
27 Co 0.627 0.143 0.992 0.992
28 Ni 0.394 0.152 0.963 0.991
29 Cu 0.560 0.159 0.962 0.989
30 Zn 0.526 0.167 0.963 0.988
31 Ga 0.493 0.176 0.956 0.987
32 Ge 0.465 0.183 0.937 0.985
33 As 0.438 0.188 0.923 0.984
34 Se 0.411 0.193 0.908 0.982
35 Br 0.382 0.196 0.894 0.980
36 Kr 0.357 0.199 0.880 0.978
37 Rb 0.333 0.201 0.869 0.976
38 Sr 0.310 0.205 0.859 0.974
39 Y 0.290 0.209 0.848 0.972
40 Zr 0.270 0.214 0.840 0.969
41 Nb 0.253 0.278 0.832 0.966
42 Mo 0.235 0.278 0.825 0.963
43 Tc 0.220 0.278 0.819 0.960
44 Ru 0.206 0.279 0.812 0.957
45 Rh 0.192 0.282 0.807 0.954
46 Pd 0.180 0.286 0.802 0.951
47 Ag 0.169 0.290 0.796 0.948
48 Cd 0.157 0.294 0.789 0.944
49 In 0.147 0.299 0.782 0.940
50 Sn 0.138 0.524 0.778 0.936
51 Sb 0.130 0.513 0.775 0.931
52 Te 0.123 0.504 0.771 0.926
53 I 0.116 0.495 0.767 0.921
54 Xe 0.109 0.488 0.763 0.915
55 Cs 0.103 0.481 0.756 0.909
56 Ba 0.098 0.474 0.751 0.903
57 La 0.093 0.467 0.744 0.896
58 Ce 0.088 0.460 0.737 0.889
59 Pr 0.083 0.454 0.730 0.882
60 Nd 0.079 0.448 0.724 0.875
61 Pm 0.075 0.444 0.717 0.868
62 Sm 0.071 0.437 0.710 0.861
63 Eu 0.068 0.432 0.702 0.853
64 Gd 0.065 0.428 0.695 0.845
65 Tb 0.062 0.424 0.688 0.836
66 Dy 0.059 0.419 0.679 0.826
67 Ho 0.056 0.416 0.669 0.818
68 Er 0.053 0.413 0.660 0.808
69 Tm 0.051 0.411 0.650 0.799
70 Yb 0.049 0.410 0.640 0.790
71 Lu 0.047 0.408 0.630 0.780
72 Hf 0.045 0.407 0.619 0.769
73 Ta 0.043 0.406 0.608 0.757
74 W 0.042 0.406 0.597 0.745
75 Re 0.041 0.371 0.587 0.732
76 Os 0.039 0.318 0.577 0.719
77 Ir 0.038 0.277 0.566 0.706
78 Pt 0.037 0.246 0.555 0.694
79 Au 0.036 0.221 0.344 0.680
80 Hg 0.035 0.203 0.533 0.667
81 Tl 0.034 0.197 0.522 0.653
82 Pb 0.033 0.192 0.511 0.640
83 Bi 0.032 0.189 0.500 0.627
84 Po 0.032 0.189 0.488 0.614
85 At 0.031 0.187 0.474 0.601
86 Rn 0.031 0.184 0.461 0.589
87 Fr 0.030 0.184 0.448 0.576
88 Ra 0.030 0.182 0.436 0.563
89 Ac 0.029 0.183 0.424 0.590
90 Th 0.029 0.179 0.413 0.537
91 Pa 0.028 0.174 0.389 0.524
92 U 0.028 0.172 0.366 0.511
93 Np 0.027 0.171 0.342 0.498
94 Pu 0.027 0.190 0.338 0.486
Table 7 Auger Yields for K, L
1
, L
2
and L
3
Levels
a
Z a
K
a
1
a
2
a
3
Z a
K
a
1
a
2
a
3
Copyright 2002 Marcel Dekker, Inc.
Table 7 Continued
Z a
K
a
1
a
2
a
3
Z a
K
a
1
a
2
a
3
95 Am 0.026 0.167 0.326 0.474
96 Cm 0.026 0.185 0.321 0.461
97 Bk 0.025 0.184 0.317 0.450
98 Cf 0.025 0.183 0.313 0.440
99 Es 0.025 0.182 0.307 0.430
100 Fm 0.024 0.181 0.300 0.421
101 Md 0.024 0.180 0.294 0.412
102 No 0.024 0.180 0.287 0.404
103 Lr 0.023 0.180 0.282 0.396
104 0.023 0.178 0.275 0.389
105 0.023 0.178 0.269 0.382
106 0.022 0.177 0.264 0.376
107 0.022 0.175 0.256 0.370
108 0.022 0.172 0.251 0.365
109 0.022 0.169 0.247 0.360
110 0.021 0.169 0.244 0.356
a
Designation for L shell is omitted; read, for example, a
1
=a
L1
.
Source: From Krause, 1979.
Copyright 2002 Marcel Dekker, Inc.
APPENDIXVII: COEFFICIENTS FOR CALCULATINGTHE PHOTOELECTRIC
ABSORPTIONCROSS SECTIONS (BARNS=ATOM) VIA ln^ln REPRESENTATION
Table1 E>E
K
a
Atomic
number Element
Atomic
weight A
0
A
1
A
2
A
3
1 H 1.008 2.44964 73.34953 74.71370-2 7.09962-3
2 He 4.003 6.06488 73.29055 71.07256-1 1.44465-2
3 Li 6.940 7.75370 72.81801 72.41738-1 2.62542-2
4 Be 9.012 9.04511 72.83487 72.10021-1 2.29526-2
5 B 10.811 9.95057 72.74173 72.15138-1 2.27845-2
6 C 12.010 1.06879 1 72.71400 72.00530-1 2.07248-2
7 N 14.008 1.12765 1 72.65400 72.00445-1 2.00765-2
8 O 16.000 1.17130 1 72.57229 72.05893-1 1.99244-2
9 F 19.000 1.20963 1 72.44148 72.34461-1 2.19537-2
10 Ne 20.183 1.24485 1 72.45819 72.12591-1 1.96489-2
11 Na 22.997 1.26777 1 72.24521 72.74873-1 2.50270-2
12 Mg 24.320 1.28793 1 72.12574 72.99392-1 2.67643-2
13 Al 26.970 1.31738 1 72.18203 72.58960-1 2.22840-2
14 Si 28.086 1.32682 1 71.98174 73.16950-1 2.73928-2
15 P 30.975 1.33735 1 71.86342 73.39440-1 2.88858-2
16 S 32.066 1.37394 1 72.04786 72.73259-1 2.29976-2
17 Cl 35.457 1.36188 1 71.71937 73.54154-1 2.90841-2
18 Ar 39.944 1.39491 1 71.82276 73.28827-1 2.74382-2
19 K 39.102 1.37976 1 71.54015 73.94528-1 3.23561-2
20 Ca 40.080 1.42950 1 71.88644 72.83647-1 2.26263-2
21 Sc 44.960 1.39664 1 71.40872 74.14365-1 3.34355-2
22 Ti 47.900 1.43506 1 71.66322 73.31539-1 2.62065-2
23 V 50.942 1.47601 1 71.88867 72.71861-1 2.15792-2
24 Cr 51.996 1.48019 1 71.82430 72.79116-1 2.17324-2
25 Mn 54.940 1.48965 1 71.79872 72.83664-1 2.22095-2
26 Fe 55.850 1.43456 1 71.23491 74.18785-1 3.21662-2
27 Co 58.933 1.47047 1 71.38933 73.86631-1 3.03286-2
28 Ni 58.690 1.42388 1 79.67736-1 74.78070-1 3.66138-2
29 Cu 63.540 1.45808 1 71.18375 74.13850-1 3.12088-2
30 Zn 65.380 1.44118 1 79.33083-1 74.77357-1 3.62829-2
31 Ga 69.720 1.36182 1 73.18459-1 76.11348-1 4.58138-2
32 Ge 72.590 1.39288 1 74.79613-1 75.72897-1 4.31277-2
33 As 74.920 1.34722 1 77.73513-2 76.60456-1 4.92177-2
34 Se 78.960 1.30756 1 1.83235-1 76.94264-1 5.02280-2
35 Br 79.920 1.32273 1 1.37130-1 76.83203-1 4.95424-2
36 Kr 83.800 1.35927 1 73.05214-2 76.51340-1 4.77616-2
37 Rb 85.480 1.30204 1 3.82736-1 77.32427-1 5.29874-2
38 Sr 87.620 1.35888 1 2.20194-3 76.38940-1 4.60070-2
39 Y 88.905 1.34674 1 1.91023-1 76.86616-1 4.97356-2
40 Zr 91.220 1.27538 1 6.97409-1 77.89307-1 5.64531-2
41 Nb 92.906 1.33843 1 2.81028-1 76.86607-1 4.86607-2
42 Mo 95.950 1.39853 1 71.17426-1 75.91094-1 4.17843-2
43 Tc 99.000 1.28214 1 7.51993-1 77.87006-1 5.58668-2
44 Ru 101.070 1.26658 1 8.85020-1 78.11144-1 5.73759-2
45 Rh 102.910 1.21760 1 1.19682 78.66697-1 6.06931-2
Copyright 2002 Marcel Dekker, Inc.
Table1 Continued
Atomic
number Element
Atomic
weight A
0
A
1
A
2
A
3
46 Pd 106.400 1.39389 1 1.64528-1 76.62170-1 4.76289-2
47 Ag 107.880 1.33926 1 4.41380-1 76.93711-1 4.82085-2
48 Cd 112.410 1.15254 1 1.07714 78.31424-1 5.79120-2
49 In 114.820 1.18198 1 1.45768 78.88529-1 6.05982-2
50 Sn 118.690 1.30323 1 7.90788-1 77.62349-1 5.27872-2
51 Sb 121.760 9.06999 3.28791 71.26203 8.53470-2
52 Te 127.600 1.16656 1 1.71052 79.48281-1 6.53213-2
53 I 126.910 1.21075 1 1.43635 78.82038-1 6.03575-2
54 Xe 131.300 1.10857 1 2.08356 71.01209 6.90310-2
55 Cs 132.910 1.13757 1 1.94161 79.83232-1 6.7l986-2
56 Ba 137.360 1.02250 1 2.67835 71.12648 7.62669-2
57 La 138.920 1.09780 1 2.23814 71.03549 7.02339-2
58 Ce 140.130 1.02725 1 2.74562 71.14174 7.74162-2
59 Pr 140.920 1.10156 1 2.22056 71.02216 6.90465-2
60 Nd 144.270 1.17632 1 1.79481 79.36661-1 6.35332-2
61 Pm 147.000 1.13864 1 2.05593 79.88180-1 6.69106-2
62 Sm 150.350 1.19223 1 1.79546 79.42902-1 6.44202-2
63 Eu 152.000 1.16168 1 1.97533 79.70901-1 6.58459-2
64 Gd 157.260 9.91968 3.03111 71.17520 7.86751-2
65 Tb 158.930 1.13818 1 2.14447 79.99222-1 6.75569-2
66 Dy 162.510 1.14845 1 2.10451 79.89870-1 6.69382-2
67 Ho 164.940 8.75203 3.71822 71.29273 8.55026-2
68 Er 167.270 1.20195 1 1.84815 79.39582-1 6.38106-2
69 Tm 168.940 1.25613 1 1.57523 78.90467-1 6.09779-2
70 Yb 173.040 7.42791 4.28955 71.35167 8.66136-2
71 Lu 174.990 1.26387 1 1.55476 78.81094-1 6.02036-2
72 Hf 178.500 7.58160 4.47037 71.42808 9.39044-2
73 Ta 180.950 8.65271 3.73117 71.26359 8.23539-2
74 W 183.920 7.57541 4.28874 71.34998 8.65200-2
75 Re 186.200 1.36944 7.79444 71.99822 1.26225-1
76 Os 190.200 1.37534 1 1.02122 77.77126-1 5.38811-2
77 Ir 192.200 1.25506 1 1.63090 78.75676-1 5.92011-2
78 Pt 195.090 1.27882 1 1.63605 78.98523-1 6.18550-2
79 Au 197.200 4.96352 5.79212 71.61842 1.02911-1
80 Hg 200.610 1.97594 1 71.97990 72.76981-1 2.68856-2
81 Tl 204.390 1.52879 1 2.73664-1 76.38890-1 4.57495-2
82 Pb 207.210 8.63374 3.69400 71.21312 7.74601-2
83 Bi 209.000 9.44293 3.44965 71.19886 7.83484-2
86 Rn 222.000 1.51782 1 3.49021-1 76.37638-1 4.51377-2
90 Th 232.000 1.34336 1 1.34805 78.13282-1 5.55664-2
92 U 238.070 1.37951 1 1.23983 78.01545-1 5.53596-2
94 Pu 239.100 1.82787 1 71.17371 73.68344-1 2.98738-2
a
Notation abbreviated as in Appendix VI, Table 4.
Copyright 2002 Marcel Dekker, Inc.
Table 2 E
L1
<E<E
K
a
Atomic
number Element
Atomic
weight A
0
A
1
A
2
11 Na 22.997 1.02355 1 72.55905 71.19524-1
12 Mg 24.310 1.05973 1 72.89818 2.34506-1
13 Al 26.970 1.08711 1 72.77860 1.75853-1
14 Si 28.086 1.12237 1 72.73694 1.27557-1
15 P 30.975 1.15508 1 72.92200 2.54262-1
16 S 32.066 1.18181 1 72.64618 79.68049-2
17 Cl 35.457 1.20031 1 72.41694 72.40897-1
18 Ar 39.944 1.22960 1 72.63279 77.36600-2
19 K 39.102 1.24878 1 72.53656 71.04892-1
20 Ca 40.080 1.27044 1 72.55011 79.43195-2
21 Sc 44.960 1.28949 1 72.40609 71.77791-1
22 Ti 47.900 1.31075 1 72.53576 79.57177-2
23 V 50.942 1.32514 1 72.49765 7 1.06383-1
24 Cr 51.996 1.34236 1 72.51532 71.01999-1
25 Mn 54.040 1.35761 1 72.49761 71.05943-1
26 Fe 55.850 1.36696 1 72.39195 71.37648-1
27 Co 58.933 1.38699 1 72.50669 78.69945-2
28 Ni 58.690 1.39848 1 72.48080 78.88115-2
29 Cu 63.540 1.42439 1 72.58677 76.67398-2
30 Zn 65.380 1.43221 1 72.62384 72.64926-2
31 Ga 69.720 1.44795 1 72.54469 77.57204-2
32 Ge 72.590 1.46813 1 72.69285 72.08355-2
33 As 74.920 1.46431 1 72.48397 77.96180-2
34 Se 78.960 1.47048 1 72.38853 71.05877-1
35 Br 79.920 1.48136 1 72.42347 79.14590-2
36 Kr 83.800 1.49190 1 72.42418 78.76447-2
37 Rb 85.480 1.49985 1 72.39108 79.59473-2
38 Sr 87.620 1.50114 1 72.28169 71.26485-1
39 Y 88.905 1.51822 1 72.38946 78.81174-2
40 Zr 91.220 1.52906 1 72.38703 79.12292-2
41 Nb 92.906 1.52088 1 72.20278 71.36759-1
42 Mo 95.950 1.53494 1 72.26646 71.16881-1
43 Tc 99.000 1.55086 1 72.33733 79.87857-1
44 Ru 101.070 1.54734 1 72.23080 71.19454-1
45 Rh 102.910 1.55757 1 72.24976 71.13377-1
46 Pd 106.400 1.55649 1 72.17229 71.27652-1
47 Ag 107.880 1.56869 1 72.22636 71.12223-1
48 Cd 112.410 1.59668 1 72.38363 78.01104-2
49 In 114.820 1.62101 1 72.51838 75.49951-2
50 Sn 118.690 1.58638 1 72.19010 71.13539-1
51 Sb 121.760 1.57557 1 72.04460 71.40745-1
52 Te 127.600 1.61087 1 72.27876 79.29405-2
53 I 126.910 1.64086 1 72.48214 75.07179-2
54 Xe 131.300 1.63098 1 72.31679 78.54498-2
55 Cs 132.910 1.65418 1 72.46363 75.42849-2
56 Ba 137.360 1.66217 1 72.48972 74.49623-2
57 La 138.920 1.63134 1 72.20156 79.80569-2
58 Ce 140.130 1.65862 1 72.36288 76.54708-2
Copyright 2002 Marcel Dekker, Inc.
Table 2 Continued
Atomic
number Element
Atomic
weight A
0
A
1
A
2
59 Pr 140.920 1.67179 1 72.40326 76.12619-2
60 Nd 144.270 1.65964 1 72.26073 78.72426-2
61 Pm 147.000 1.68368 1 72.38881 76.45041-2
62 Sm 150.350 1.68725 1 72.39051 76.01080-2
63 Eu 152.000 1.70692 1 72.48046 74.47055-2
64 Cd 157.260 1.71159 1 72.47838 74.37107-2
65 Tb 158.930 1.71499 1 72.45507 74.71370-2
66 Dy 162.510 1.73446 1 72.54821 73.17606-2
67 Ho 164.940 1.76583 1 72.72523 78.19409-4
68 Er 167.270 1.77988 1 72.74671 72.87580-3
69 Tm 168.940 1.74250 1 72.51103 73.29454-2
70 Yb 173.040 1.69795 1 72.22577 77.32557-2
71 Lu 174.990 1.72638 1 72.37189 74.95994-2
72 Hf 178.500 1.64329 1 71.82851 71.32268-1
73 Ta 180.950 1.72410 1 72.30313 75.91006-2
74 W 183.920 1.72533 1 72.23874 77.27338-2
75 Re 186.200 1.78750 1 72.61051 71.36093-2
76 Os 190.200 1.73525 1 72.28550 75.88047-2
77 Ir 192.200 1.65270 1 71.76315 71.35232-1
78 Pt 195.090 1.73636 1 72.21112 77.30934-2
79 Au 197.200 1.74240 1 72.23911 76.63720-2
80 Hg 200.610 1.71857 1 72.08470 78.53294-2
81 Tl 204.390 1.77379 1 72.37745 74.33223-2
82 Pb 207.210 1.77963 1 72.37691 74.55883-2
83 Bi 209.000 1.75348 1 72.23353 75.96161-2
86 Rn 222.000 1.75028 1 72.13876 77.24638-2
90 Th 232.000 1.85481 1 72.61281 77.90574-3
92 U 238.070 1.75258 1 72.07237 77.23932-2
94 Pu 239.100 1.75519 1 72.02162 78.22940-2
a
Notation abbreviated as in Appendix VI, Table 4.
Copyright 2002 Marcel Dekker, Inc.
Table 3 E
M1
< E < E
L3
a
Atomic number Element Atomic weight A
0
A
1
31 Ga 69.720 1.22646 1 72.68965
32 Ge 72.590 1.24133 1 72.53085
33 As 74.920 1.25392 1 72.41380
34 Se 78.960 1.26773 1 72.39750
35 Br 79.920 1.27612 1 72.37730
36 Kr 83.800 1.28898 1 72.26021
37 Rb 85.480 1.30286 1 72.38693
38 Sr 87.620 1.31565 1 72.36655
39 Y 88.905 1.32775 1 72.43174
40 Zr 91.220 1.34508 1 72.50201
41 Nb 92.906 1.35434 1 72.50135
42 Mo 95.950 1.36568 1 72.48982
43 Tc 99.000 1.37498 1 72.44737
44 Ru 101.070 1.38782 1 72.48066
45 Rh 102.910 1.40312 1 72.61303
46 Pd 106.400 1.41392 1 72.57206
47 Ag 107.880 1.41673 1 72.48078
48 Cd 112.410 1.43497 1 72.52756
49 In 114.820 1.44115 1 72.49401
50 Sn 118.690 1.45572 1 72.56792
51 Sb 121.760 1.46268 1 72.55562
52 Te 127.600 1.47125 1 72.54324
53 I 126.910 1.47496 1 72.48179
54 Xe 131.300 1.47603 1 72.45068
55 Cs 132.910 1.49713 1 72.53145
56 Ba 137.360 1.50844 1 72.56341
57 La 138.920 1.51863 1 72.58287
58 Ce 140.130 1.52693 1 72.58174
59 Pr 140.920 1.53379 1 72.57086
60 Nd 144.270 1.54353 1 72.59006
61 Pm 147.000 1.55131 1 72.59623
62 Sm 150.350 1.56006 1 72.61328
63 Eu 152.000 1.57063 1 72.63481
64 Gd 157.260 1.57159 1 72.60843
65 Tb 158.930 1.58415 1 72.64040
66 Dy 162.510 1.59225 1 72.65289
67 Ho 164.940 1.60140 1 72.67903
68 Er 167.270 1.60672 1 72.67587
69 Tm 168.940 1.61269 1 72.67886
70 Yb 173.040 1.61794 1 72.67715
71 Lu 174.990 1.62289 1 72.67128
72 Hf 178.500 1.62758 1 72.66622
73 Ta 180.950 1.63068 1 72.66148
74 W 183.920 1.62613 1 72.60672
75 Re 186.200 1.63564 1 72.62453
76 Os 190.200 1.64233 1 72.63163
77 Ir 192.200 1.65144 1 72.64832
78 Pt 195.090 1.67024 1 72.71631
79 Au 197.200 1.64734 1 72.57834
Copyright 2002 Marcel Dekker, Inc.
Table 3 Continued
Atomic number Element Atomic weight A
0
A
1
80 Hg 200.610 1.65903 1 72.60670
81 Tl 204.390 1.66564 1 72.61593
82 Pb 207.210 1.67131 1 72.61538
83 Bi 209.000 1.67078 1 72.58648
86 Rn 222.000 1.69000 1 72.60945
90 Th 232.000 1.70483 1 72.58569
92 U 238.070 1.70353 1 72.56903
94 Pu 239.100 1.72953 1 72.62164
a
Notation abbreviated in Appendix VI, Table 4.
Table 4 E < E
M5
a
Atomic number Element Atomic weight A
0
A
1
61 Pm 147.000 1.55131 1 72.59623
62 Sm 150.350 1.56006 1 72.61328
63 Eu 152.000 1.57063 1 72.63481
64 Gd 157.260 1.57159 1 72.60843
65 Tb 158.930 1.58415 1 72.64040
66 Dy 162.510 1.59225 1 72.65289
67 Ho 164.940 1.60140 1 72.67903
68 Er 167.270 1.60672 1 72.67587
69 Tm 168.940 1.61269 1 72.67886
70 Yb 173.040 1.39111 1 72.40380
71 Lu 174.990 1.39813 1 72.40841
72 Hf 178.500 1.40548 1 72.42829
73 Ta 180.950 1.41313 1 72.47214
74 W 183.920 1.42536 1 72.32582
75 Re 186.200 1.42392 1 72.35326
76 Os 190.200 1.42795 1 72.21971
77 Ir 192.200 1.43422 1 72.40183
78 Pt 195.090 1.43785 1 72.34834
79 Au 197.200 1.44398 1 72.32838
80 Hg 200.610 1.45195 1 72.33016
81 Tl 204.390 1.45473 1 72.26773
82 Pb 207.210 1.45771 1 72.25279
83 Bi 209.000 1.46832 1 72.30940
86 Rn 222.000 1.47243 1 72.12905
90 Th 232.000 1.47730 1 71.91192
92 U 238.070 1.49036 1 72.12148
94 Pu 239.100 1.48535 1 71.87733
a
Notation abbreviated as in Appendix VI, Table 4.
Source: From McMaster et al., 1969.
Copyright 2002 Marcel Dekker, Inc.
APPENDIXVIII: COEFFICIENTS FOR CALCULATINGTHE INCOHERENT
COLLISIONCROSS SECTIONS
C
(BARNS=ATOM) VIATHE ln^ln
REPRESENTATION
Atomic Number Element A
0
A
1
A
2
A
3
1 H 72.l5772 1.32685 73.05620-1
a
1.85025-2
2 He 72.56357 2.02536 74.48710-1 2.79691-2
3 Li 71.08740 1.03368 71.90377-1 7.79955-3
4 Be 76.90079-1 9.46448-1 71.71142-1 6.51413-3
5 B 77.91177-1 1.21611 72.39087-1 1.17686-2
6 C 79.87878-1 1.46693 72.93743-1 1.56005-2
7 N 71.23693 1.74510 73.54660-1 1.98705-2
8 O 71.73679 2.17686 74.49050-1 2.64733-2
9 F 71.87570 2.32016 74.75412-1 2.80680-2
10 Ne 71.75510 2.24226 74.47640-1 2.55801-2
11 Na 79.67717-1 1.61794 72.87191-1 1.31526-2
12 Mg 75.71611-1 1.35498 72.22491-1 8.30141-3
13 Al 74.39322-1 1.30867 72.11648-1 7.54210-3
14 Si 74.14971-1 1.34868 72.22315-1 8.41959-3
15 P 74.76903-1 1.46032 72.51331-1 1.07202-2
16 S 76.56419-1 1.65408 72.98623-1 1.42979-2
17 Cl 77.18627-1 1.74294 73.19429-1 1.58429-2
18 Ar 76.82105-1 1.74279 73.17646-1 1.56467-2
19 K 73.44007-1 1.49236 72.54135-1 1.07684-2
20 Ca 79.82420-2 1.32829 72.13747-1 7.73065-3
21 Sc 71.59831-1 1.39055 72.25849-1 8.51954-3
22 Ti 72.30573-1 1.45848 72.39160-1 9.38528-3
23 V 73.08103-1 1.52879 72.52768-1 1.02571-2
24 Cr 73.87641-1 1.59727 72.66240-1 1.11523-2
25 Mn 72.47059-1 1.49722 72.38781-1 8.93208-3
26 Fe 73.42379-1 1.57245 72.53198-1 9.85822-3
27 Co 74.28804-1 1.64129 72.66013-1 1.06512-2
28 Ni 75.04360-1 1.70040 72.76443-1 1.12628-2
29 Cu 75.70210-1 1.75042 72.84555-1 1.16930-2
30 Zn 74.20535-1 1.63400 72.53646-1 9.27233-3
31 Ga 73.58218-1 1.60050 72.44908-1 8.61898-3
32 Ge 73.34383-1 1.60237 72.45555-1 8.71239-3
33 As 73.39189-1 1.62535 72.50783-1 9.09103-3
34 Se 74.32927-1 1.72833 72.77138-1 1.11735-2
35 Br 74.48001-1 1.76082 72.85099-1 1.17865-2
36 Kr 73.91810-1 1.73010 72.76824-1 1.11280-2
37 Rb 71.28039-1 1.53044 2.27403-1 7.39033-3
38 Sr 7.99161-2 1.38397 71.92225-1 4.78611-3
39 Y 6.29057-2 1.41577 71.99713-1 5.33312-3
40 Zr 3.66697-2 1.45207 72.08122-1 5.95139-3
41 Nb 2.02289-4 1.49347 72.17419-1 6.62245-3
42 Mo 75.62860-2 1.55778 72.33341-1 7.85506-3
43 Tc 7.57616-2 1.44950 72.04890-1 5.64745-3
44 Ru 74.24981-2 1.54639 72.26470-1 7.18375-3
45 Rh 71.60399-1 1.64861 72.50238-1 8.93818-3
46 Pd 72.67564-1 1.73740 72.69883-1 1.03248-2
Copyright 2002 Marcel Dekker, Inc.
Appendix VIII Continued
Atomic Number Element A
0
A
1
A
2
A
3
47 Ag 71.66475-1 1.65794 72.48740-1 8.66218-3
48 Ce 75.16701-2 1.57426 72.27646-1 7.05650-3
49 In 78.17283-2 1.55865 72.24492-1 6.85776-3
50 Sn 1.42151-2 1.55754 72.24736-1 6.91395-3
51 Sb 1.56362-2 1.57175 72.28753-1 7.26386-3
52 Te 74.07579-2 1.64267 72.47897-1 8.80567-3
53 I 74.04420-2 1.65596 72.51067-1 9.04874-3
54 Xe 72.82407-3 1.64039 72.47642-1 8.82144-3
55 Cs 1.84861-1 1.50030 72.13333-1 6.24264-3
56 Ba 3.44376-1 1.38742 71.86356-1 4.24917-3
57 La 4.09104-1 1.33075 71.70883-1 3.04111-3
58 Ce 4.39881-1 1.30925 71.64548-1 2.52641-3
59 Pr 4.49124-1 1.30351 71.61841-1 2.27394-3
60 Nd 4.37283-1 1.31370 71.62866-1 2.29377-3
61 Pm 4.05823-1 1.33837 71.67229-1 2.55570-3
62 Sm 3.55383-1 1.37733 71.74941-1 3.06213-3
63 Eu 2.80316-1 1.44016 71.88641-1 4.01226-3
64 Gd 2.73133-1 1.43842 71.86137-1 3.75240-3
65 Tb 2.57539-1 1.45064 71.87591-1 3.79932-3
66 Dy 2.42685-1 1.46266 71.89102-1 3.85628-3
67 Ho 2.28493-1 1.47438 71.90559-1 3.90903-3
68 Er 2.15233-1 1.48545 71.91908-1 3.95645-3
69 Tm 2.02656-1 1.49625 71.93234-1 4.00233-3
70 Yb 2.02248-1 1.48804 71.89143-1 3.62264-3
71 Lu 1.97176-1 1.50264 71.92474-1 3.85751-3
72 Hf 1.99469-1 1.50233 71.91385-1 3.74011-1
73 Ta 1.96871-1 1.50623 71.91396-1 3.70889-3
74 W 1.91015-1 1.51240 71.91922-1 3.71450-3
75 Re 1.89644-1 1.50867 71.89570-1 3.49584-3
76 Os 1.16448-1 1.57615 72.05532-1 4.66731-3
77 Ir 7.19908-2 1.61204 72.13186-1 5.20497-3
78 Pt 4.20186-2 1.63611 72.17964-1 5.52670-3
79 Au 1.56916-2 1.65406 72.20982-1 5.70751-3
80 Hg 1.14587-1 1.58076 72.02968-1 4.35692-3
81 Tl 1.47052-1 1.56695 72.00347-1 4.20901-3
82 Pb 1.82167-1 1.54661 71.95793-1 3.90772-3
83 Bi 1.89860-1 1.56125 72.00932-1 4.36768-3
86 Rn 1.96619-1 1.60080 72.13800-1 5.51717-3
90 Th 1.70890-1 1.65561 72.29702-1 6.92516-3
92 U 1.08277-1 1.74158 72.54104-1 8.95056-3
94 Pu 3.88791-2 1.82229 72.76009-1 1.07392-2
a
Notation as in Appendix VI, Table 4.
Source: From McMaster et al., 1969.
Copyright 2002 Marcel Dekker, Inc.
APPENDIXIX: COEFFICIENTS FOR CALCULATINGTHE COHERENT SCATTERING
CROSS SECTIONS
R
(BARNS/ATOM) VIATHE ln^ln REPRESENTATION
Atomic number Element A
0
A
1
A
2
A
3
1 H 71.19075-1
a
79.37086-1 72.00538-1 1.06587-2
2 He 1.04768 78.51805-2 74.03527-1 2.69398-2
3 Li 1.34366 1.81557-1 74.23981-1 2.66190-2
4 Be 2.00860 74.61920-2 73.37018-1 1.86939-2
5 B 2.61862 72.07916-1 72.86283-1 1.44966-2
6 C 3.10861 72.60580-1 72.71974-1 1.35181-2
7 N 3.47760 72.15762-1 72.88874-1 1.51312-2
8 O 3.77239 71.48539-1 73.07124-1 1.67303-2
9 F 4.00716 75.60908-2 73.32017-1 1.87934-2
10 Ne 4.20151 4.16247-2 73.56754-1 2.07585-2
11 Na 4.26374 1.34662-1 73.70080-1 2.14467-2
12 Mg 4.39404 1.37858-1 73.59540-1 2.02380-2
13 Al 4.51995 1.40549-1 73.52441-1 1.93692-2
14 Si 4.64678 1.62780-1 73.58563-1 1.96926-2
15 P 4.76525 1.68708-1 73.60383-1 1.97155-2
16 S 4.92707 1.65746-1 73.59424-1 1.95505-2
17 Cl 5.07222 1.49127-1 73.52858-1 1.89439-2
18 Ar 5.21079 1.35618-1 73.47214-1 1.84333-2
19 K 5.25587 1.88040-1 73.59623-1 1.93085-2
20 Ca 5.32375 2.06685-1 73.61664-1 1.93328-2
21 Sc 5.43942 2.00174-1 73.59064-1 1.91027-2
22 Ti 5.55039 1.97697-1 73.57694-1 1.89866-2
23 V 5.65514 1.99533-1 73.57487-1 1.89691-2
24 Cr 5.77399 2.03858-1 73.59699-1 1.92225-2
25 Mn 5.84604 2.13814-1 73.59718-1 1.91459-2
26 Fe 5.93292 2.25048-1 73.61748-1 1.93024-2
27 Co 6.01478 2.37959-1 73.64056-1 1.94754-2
28 Ni 6.09204 2.52277-1 73.66568-1 1.96586-2
29 Cu 6.17739 2.73123-1 73.72360-1 2.01638-2
30 Zn 6.23402 2.84312-1 73.72143-1 2.00525-2
31 Ga 6.28298 2.91334-1 73.69391-1 1.97029-2
32 Ge 6.33896 2.91512-1 73.65643-1 1.92895-2
33 As 6.39750 2.88866-1 73.61747-1 1.88788-2
34 Se 6.45637 2.86737-1 73.58794-1 1.85618-2
35 Br 6.51444 2.86324-1 73.57027-1 1.83557-2
36 Kr 6.57129 2.87711-1 73.56311-1 1.82470-2
37 Rb 6.59750 3.02389-1 73.56755-1 1.81706-2
38 Sr 6.62203 3.24559-1 73.61651-1 1.84800-2
39 Y 6.67096 3.25075-1 73.60613-1 1.83325-2
40 Zr 6.72275 3.23964-1 73.59463-1 1.81890-2
41 Nb 6.79013 3.11282-1 73.55233-1 1.78231-2
42 Mo 6.84600 3.02797-1 73.51131-1 1.74403-2
43 Tc 6.87599 3.26165-1 73.58969-1 1.80482-2
44 Ru 6.93136 3.34794-1 73.63497-1 1.84429-2
45 Rh 6.97547 3.46394-1 73.67794-1 1.87885-2
46 Pd 7.03216 3.49838-1 73.70099-1 1.89983-2
47 Ag 7.06446 3.63456-1 73.73597-1 1.92478-2
Copyright 2002 Marcel Dekker, Inc.
Appendix IX Continued
Atomic number Element A
0
A
1
A
2
A
3
48 Cd 7.09856 3.72199-1 73.75345-1 1.93481-2
49 In 7.12708 3.82082-1 73.76855-1 1.94151-2
50 Sn 7.16085 3.85512-1 73.76481-1 1.93305-2
51 Sb 7.19665 3.85543-1 73.75054-1 1.91608-2
52 Te 7.23464 3.82493-1 73.72715-1 1.89194-2
53 I 7.27415 3.77223-1 73.69728-1 1.86280-2
54 Xe 7.31469 3.70315-1 73.66280-1 1.83025-2
55 Cs 7.33490 3.76825-1 73.65713-1 1.81843-2
56 Ba 7.35812 3.79361-1 73.64099-1 1.79817-2
57 La 7.39532 3.69895-1 73.59376-1 1.75406-2
58 Ce 7.44255 3.71328-1 73.59642-1 1.75852-2
59 Pr 7.48347 3.68431-1 73.57689-1 1.74099-2
60 Nd 7.52334 3.66462-1 73.56048-1 1.72620-2
61 Pm 7.56222 3.65055-1 73.54511-1 1.71214-2
62 Sm 7.60020 3.64134-1 73.53086-1 1.69894-2
63 Eu 7.63711 3.63957-1 73.51909-1 1.68783-2
64 Gd 7.66938 3.59752-1 73.48899-1 1.65890-2
65 Tb 7.70798 3.65345-1 73.50031-1 1.66927-2
66 Dy 7.74188 3.67107-1 73.49433-1 1.66273-2
67 Ho 7.77470 3.69722-1 73.49132-1 1.65862-2
68 Er 7.80643 3.73226-1 73.49147-1 1.65710-2
69 Tm 7.83711 3.77547-1 73.49441-1 1.65780-2
70 Yb 7.86662 3.82933-1 73.50126-1 1.66173-2
71 Lu 7.89137 3.86034-1 73.49756-1 1.65480-2
72 Hf 7.91803 3.87021-1 73.48881-1 1.64406-2
73 Ta 7.94534 3.87299-1 73.47926-1 1.63299-2
74 W 7.97266 3.87704-1 73.47155-1 1.62372-2
75 Re 7.99940 3.88739-1 73.46726-1 1.61751-2
76 Os 8.02574 3.90458-1 73.46658-1 1.61455-2
77 Ir 8.05150 3.93143-1 73.47052-1 1.61573-2
78 Pt 8.08084 3.95790-1 73.48032-1 1.62345-2
79 Au 8.10524 4.00576-1 73.49340-1 1.63264-2
80 Hg 8.12542 4.05858-1 73.50329-1 1.63772-2
81 Tl 8.14399 4.08692-1 73.49802-1 1.62888-2
82 Pb 8.15996 4.18031-1 73.52330-1 1.64660-2
83 Bi 8.17489 4.27916-1 73.55068-1 1.66601-1
86 Rn 8.22553 4.51478-1 73.62056-1 1.71556-2
90 Th 8.27843 4.79056-1 73.67657-1 1.74621-2
92 U 8.33010 4.78314-1 73.67250-1 1.74129-2
94 Pu 8.38174 4.77085-1 73.66556-1 1.73422-2
a
Notation as in Appendix VI, Table 4.
Source: From McMaster et al., 1969.
Copyright 2002 Marcel Dekker, Inc.
APPENDIXX: PARAMETERSFORCALCULATINGTHE TOTALMASSATTENUATIONCOEFFICIENTSINTHEENERGYRANGE
0.1^1000 keV [VIAEQ. (78)]
Table1 Parameters for Calculating the Total Mass Attenuation Coefcients for Photon Energies Above the K Absorption Edge (k
1
, k
2
, k
3
, k
4
) and for
Energies Between L
1
and E
K
Absorption Edge (l
1
, l
2
, l
3
, l
4
); Z=Atomic Number, Sy =Symbol of Corresponding Element
E
L1
<E<E
K
E>E
K
Z Sy E
L1
l
1
l
2
l
3
l
4
E
K
k
1
k
2
k
3
k
4
1 H 76.48071 3.27538 0.04385 70.00866
2 He 73.51566 2.60589 0.26361 70.03787
3 Li 72.36523 2.88866 0.15389 70.02935
4 Be 71.38914 3.05349 0.07919 70.02484
5 B 70.58244 3.07947 0.05871 70.02647
6 C 77.00266 5.19086 70.34413 0.2838 0.11476 3.09245 0.03213 70.02494
7 N 72.38938 3.24818 70.11121 0.4016 0.63927 3.08731 0.01587 70.02538
8 O 72.50341 3.43696 70.12806 0.5320 1.08730 3.07389 0.00035 70.02438
9 F 72.03564 3.43789 70.13520 0.6854 1.42132 3.06182 70.00946 70.02708
10 Ne 72.91417 4.13834 70.22931 0.8669 1.81194 3.04034 70.02320 70.02568
11 Na 0.15632 2.00149 0.32080 70.04820 1.0721 2.07992 3.04059 70.02978 70.03575
12 Mg 2.10634 0.28244 0.92539 70.11707 1.3050 2.39196 3.00737 70.04492 70.02734
13 Al 3.34768 70.83777 1.34549 70.16880 1.5596 2.62043 2.99387 70.05681 70.03063
14 Si 3.81364 71.04042 1.42378 70.17980 1.8389 2.88077 2.96887 70.06174 70.02724
15 P 3.05305 70.24003 1.25060 70.17486 2.1455 3.06672 2.95078 70.07943 70.03227
16 S 0.2292 2.76125 0.29160 1.09998 70.16451 2.4720 3.28801 2.92821 70.08514 70.03094
17 Cl 0.2702 3.00797 0.19834 1.11670 70.16546 2.8224 3.42667 2.90289 70.09061 70.02881
18 Ar 0.3200 2.62786 0.68939 1.01331 70.16625 3.2029 3.53219 2.88277 70.10191 70.03142
19 K 0.3771 0.68453 3.86200 70.43637 0.04226 3.6074 3.75863 2.85955 70.09992 70.02689
20 Ca 0.4378 2.44853 1.46227 0.69759 70.12878 4.0381 3.93329 2.83775 70.11226 70.02972
21 Sc 0.5004 2.03061 2.27680 0.31534 70.07434 4.4928 4.00179 2.81618 70.11678 70.02898
22 Ti 0.5637 2.33976 1.88017 0.55084 70.11829 4.9664 4.11127 2.79692 70.11885 70.02784
23 V 0.6282 2.38673 1.99445 0.50415 70.11481 5.4651 4.21463 2.77190 70.13140 70.03049
24 Cr 0.6946 2.53782 1.85115 0.62868 70.14231 5.9892 4.35405 2.75878 70.13335 70.03142
25 Mn 0.7690 2.65431 1.83552 0.65972 70.15525 6.5390 4.44148 2.72589 70.14641 70.03205
Copyright 2002 Marcel Dekker, Inc.
26 Fe 0.8461 2.41689 2.80062 70.00696 70.01904 7.1120 4.56600 2.70395 70.15515 70.03352
27 Co 0.9256 2.60591 2.55718 0.18669 70.06696 7.7089 4.64390 2.67587 70.16255 70.03403
28 Ni 1.0081 2.69756 2.72893 0.04244 70.03220 8.3328 4.77308 2.65503 70.16936 70.03476
29 Cu 1.0961 2.75840 2.72647 0.04391 70.03423 8.9789 4.81074 2.62267 70.18383 70.03729
30 Zn 1.1936 2.87225 2.72029 0.04441 70.03527 9.6586 4.89850 2.60534 70.18393 70.03594
31 Ga 1.2977 2.93846 2.73406 0.03264 70.03438 10.3671 4.93990 2.57045 71.19961 70.03856
32 Ge 1.4143 3.02673 2.74409 0.01747 70.03117 11.1031 5.00743 2.55814 70.19972 70.03827
33 As 1.5265 3.12353 2.74414 0.01165 70.03096 11.8667 5.07635 2.53376 70.20666 70.03872
34 Se 1.6539 3.19106 2.76121 70.01683 70.02249 12.6578 5.11759 2.50569 70.21581 70.03969
35 Br 1.7820 3.30117 2.73599 0.01398 70.03428 13.4737 5.19474 2.48145 70.22156 70.04024
36 Kr 1.9210 3.36926 2.74007 0.00224 70.03125 14.3256 5.23284 2.44844 70.23505 70.04200
37 Rb 2.0651 3.46256 2.74320 70.10161 70.02723 15.1997 5.29354 2.41378 70.24899 70.04399
38 Sr 2.2163 3.54696 2.73914 70.00741 70.03104 16.1046 5.34695 2.38309 70.25992 70.04534
39 Y 2.3725 3.63965 2.73646 70.01290 70.03036 17.0384 5.41166 2.37364 70.25521 70.04393
40 Zr 2.5316 3.71677 2.73020 70.01273 70.03038 17.9976 5.45290 2.33638 70.26879 70.04558
41 Nb 2.6977 3.79880 2.72885 70.01869 70.02896 18.9856 5.50180 2.30251 70.28165 70.04737
42 Mo 2.8655 3.86387 2.72657 70.02278 70.02881 19.9995 6.47503 3.03067 70.04965 70.03050
43 Te 3.0425 3.93865 2.72262 70.02487 70.03004 21.0440 5.57355 2.24202 70.29909 70.04910
44 Ru 3.2240 3.99914 2.70967 70.02706 70.03l74 22.1172 5.60095 2.21314 70.30745 70.04983
45 Rh 3.4119 4.07189 2.71531 70.02854 70.03173 23.2199 5.63478 2.17349 70.32262 70.05195
46 Pd 3.6043 4.12607 2.71194 70.03442 70.02941 24.3503 5.65532 2.14514 70.32915 70.05230
47 Ag 3.8058 4.19974 2.70350 70.03661 70.02694 25.5140 5.68392 2.05134 70.34706 70.05473
48 Cd 3.0180 4.24309 2.70139 70.04152 70.02776 26.7112 5.68137 2.09840 70.36341 70.05665
49 In 4.2375 4.30447 2.69926 70.04319 70.03320 27.9399 5.72030 2.04363 70.36014 70.05573
50 Sn 4.4647 4.35252 2.69160 70.04781 70.02689 29.2001 5.76598 2.05890 70.34949 70.05406
51 Sb 4.6983 4.40598 2.68835 70.04963 70.03001 30.4912 5.74957 1.98639 70.37446 70.05693
52 Te 4.9392 4.43690 2.68323 70.05424 70.03161 31.8138 3.74700 1.96282 70.37979 70.05738
53 I 5.1881 4.51763 2.67963 70.05663 70.03296 33.1694 5.79401 1.93222 70.38924 70.05845
54 Xe 5.4528 4.53737 2.66911 70.05780 70.02658 34.5614 5.80690 1.91476 70.39242 70.05865
55 Cs 5.7143 4.61780 2.66299 70.06277 70.02884 35.9846 5.82191 1.87715 70.040213 70.05948
56 Ba 5.9888 4.65770 2.65946 70.07172 70.03577 37.4406 5.84551 1.87466 70.39966 70.05904
57 La 6.2663 4.71490 2.65505 70.07121 70.03442 38.9246 5.85321 1.82364 70.41682 70.06098
58 Ce 6.5488 4.77277 2.64710 70.07113 70.03077 40.4430 5.87476 1.79700 70.42193 70.06126
59 Pr 6.8348 4.83442 2.64088 70.07717 70.03325 41.9906 5.90293 1.77311 70.42727 70.06168
60 Nd 7.1260 4.87359 2.63355 70.07273 70.02815 43.5689 5.90905 1.74652 70.43341 70.06218
Copyright 2002 Marcel Dekker, Inc.
Table1 Continued
E
L1
<E<E
K
E>E
K
Z Sy E
L1
l
1
l
2
l
3
l
4
E
K
k
1
k
2
k
3
k
4
61 Pm 7.4279 4.93318 2.62474 70.08002 70.02998 45.1840 5.92214 1.70715 70.44407 70.06326
62 Sm 7.7368 4.95712 2.61647 70.07625 70.02487 46.8342 5.92934 1.69904 70.44407 70.06312
63 Eu 8.0520 5.00931 2.61226 70.08680 70.03205 48.5190 5.94975 1.67827 70.44795 70.06335
64 Gd 8.3756 5.03443 2.60481 70.08912 70.03162 50.2391 5.92827 1.64215 70.45629 70.06402
65 Tb 8.7080 5.08055 2.60115 70.07597 70.02286 51.9957 5.95498 1.63471 70.45467 70.06359
66 Dy 9.0458 5.11495 2.59238 70.08230 70.02541 53.7885 5.95991 1.61100 70.46029 70.06408
67 Ho 9.3942 5.15537 2.58071 70.08885 70.02644 55.6177 5.94804 1.55826 70.47717 70.06595
68 Er 9.7513 5.19648 2.57148 70.09459 70.02802 57.4855 5.96408 1.54243 70.47949 70.06606
69 Tm 10.1157 5.23938 2.56466 70.09047 70.02454 59.3896 5.88779 1.43854 70.50808 70.06870
70 Yb 10.4864 5.26855 2.55541 70.10174 70.03012 61.3323 6.04546 1.57974 70.45878 70.06334
71 Lu 10.8704 5.31163 2.55490 70.09208 70.02484 63.3138 6.07395 1.57891 70.45631 70.06301
72 Hf 11.2707 5.34305 2.54268 70.10077 70.02734 65.3508 6.00771 1.48367 70.48598 70.06608
73 Ta 11.6815 5.37806 2.53595 70.09548 70.02359 67.4164 6.05159 1.50856 70.47273 70.06439
74 W 12.0998 5.41320 2.53128 70.10036 70.02631 69.5250 6.00250 1.43225 70.49552 70.06672
75 Re 12.5267 5.44913 2.52042 70.10577 70.02724 71.6764 6.03030 1.43738 70.48952 70.06580
76 Os 12.9680 5.47715 2.52666 70.09007 70.02154 73.8708 6.11400 1.50446 70.46496 70.06309
77 Ir 13.4185 5.50865 2.50428 70.10492 70.02495 76.1110 6.09889 1.45990 70.47717 70.06429
78 Pt 13.8799 5.54158 2.50551 70.09847 70.02303 78.3948 6.15378 1.53187 70.44176 70.05974
79 Au 14.3528 5.57564 2.48980 70.10793 70.02561 80.7249 6.15303 1.47494 70.46441 70.06253
80 Hg 14.8393 5.60041 2.47759 70.11241 70.02570 83.1023 6.19862 1.50676 70.45079 70.06097
81 Tl 15.3467 5.62415 2.46173 70.12251 70.02850 85.5304 6.17641 1.46912 70.46009 70.06183
82 Pb 15.8608 5.65349 2.45575 70.12044 70.02698 88.0045 6.17648 1.45115 70.46277 70.06196
83 Bi 16.3875 5.69400 2.47086 70.10349 70.02277 90.5259 6.30427 1.55736 70.42520 70.05792
84 Po 16.9393 5.73466 2.46840 70.09608 70.01973 93.1050 6.21512 1.43087 70.46426 70.06195
85 At 17.4930 5.76863 2.45600 70.10153 70.02099 95.7299 6.18945 1.38290 70.47598 70.06300
86 Rn 18.0490 5.74682 2.43686 70.11031 70.02272 98.4040 6.24492 1.47699 70.43939 70.05879
87 Fr 18.6390 5.78319 2.43220 70.10993 70.02259 101.1370 6.31471 1.51865 70.42408 70.05721
88 Ra 19.2367 5.80588 2.41932 70.11330 70.02255 103.9219 6.26760 1.45136 70.44373 70.05917
Copyright 2002 Marcel Dekker, Inc.
89 Ac 19.8400 5.83228 2.39643 70.12322 70.02407 106.7553 6.32123 1.48583 70.42883 70.05744
90 Th 20.4721 5.85156 2.40038 70.11549 70.02204 109.6509 6.39823 1.55418 70.40447 70.05479
91 Pa 21.1046 5.87741 2.35466 70.14418 70.02817 112.6014 6.28773 1.41526 70.44597 70.05907
92 U 21.7574 5.88224 2.34307 70.15054 70.03027 115.6061 6.30701 1.44776 70.43028 70.05716
Source: From Orlic et al., 1993. Reprinted with kind permission from Elsevier Science NL, Sara Burgehartstraat 25, 1055 KV Amsterdam, The Netherlands.
Copyright 2002 Marcel Dekker, Inc.
Table 2 Parameters for Calculating the Total Mass Attentuation Coefcients for Photon Energy Between L
1
, L
2
, and L
3
Absorption Edges and for
Energies Between L
3
and Lower Energy Limit E
0
or M Absorption Edge
E
0
<E<E
L3
E
L3
< E < E
L2
E
L2
< E < E
L1
Z Sy E
0
=E
edge
m
1
m
2
m
3
m
4
E
L3
l
31
l
32
E
L2
l
21
l
22
E
L1
16 S 0.1 3.98381 71.96207 1.29140 70.13371 0.2292
17 Cl 0.1 72.71337 2.78307 0.23822 70.05767 0.2702
18 Ar 0.1 2.07772 70.84611 1.21097 70.14523 0.3200
19 K 0.1 8.53825 74.62277 1.97295 70.19719 0.3771
20 Ca 0.1 5.10006 71.85097 1.29907 70.14560 0.4378
21 Sc 0.1 7.28382 73.26298 1.60507 70.16673 0.5004
22 Ti 0.1 6.39592 72.45282 1.38399 70.14698 0.5637
23 V 0.1 2.83998 0.44991 0.61932 70.08033 0.6282
24 Cr 0.1 77.38556 8.17416 71.29292 0.07794 0.6946
25 Mn 0.1 2.13747 1.11467 0.46032 70.06687 0.7690
26 Fe 0.1 2.21508 1.10224 0.47160 70.06697 0.8461
27 Co 0.1 4.76476 71.05687 1.09423 70.12565 0.9256
28 Ni 0.1 4.28916 70.66344 1.02071 70.12207 1.0081
29 Cu 0.1 2.32551 1.22642 0.45413 70.06650 1.0961
30 Zn 0.1 4.23694 70.64439 1.08840 70.13570 1.0197 77.206 727.639 1.0428 9.4520 70.2320 1.1936
31 Ga 0.1 2.60598 0.84383 0.68376 70.10262 1.1154 47.827 716.433 1.1423 6.2330 1.1180 1.2977
32 Ge 0.1 4.27624 70.65284 1.15806 70.15244 1.2167 20.461 75.2040 1.2478 5.4560 1.4580 1.4143
33 As 0.1 5.14580 71.51272 1.47391 70.19017 1.3231 6.9330 0.6640 1.3586 5.1580 1.6080 1.5265
34 Se 0.1 6.74588 73.03410 1.98313 70.24674 1.4358 0.0065 3.8833 1.4762 3.9491 2.1968 1.6539
35 Br 0.1 7.26941 73.69837 2.28437 70.28824 1.5499 74.6062 6.2403 1.5960 3.8764 2.2734 1.7820
36 Kr 0.1 11.95287 78.69443 4.04104 70.48832 1.6749 75.6370 6.9490 1.7272 3.4100 2.5440 1.9210
37 Rb 0.1 15.89070 712.8788 5.51561 70.65735 1.8044 3.2959 2.4609 1.8639 4.0979 2.2097 2.0651
38 Sr 0.1085 11.09104 78.15432 4.08776 70.52451 1.9396 2.0366 3.1926 2.0068 3.9355 2.3328 2.2163
39 Y 0.1487 5.18978 71.99566 2.09278 70.32184 2.0800 1.9840 3.2996 2.1555 3.9621 2.3596 2.3725
40 Zr 0.2122 6.15438 72.90381 2.42622 70.36813 2.2223 2.2290 3.2290 2.3067 3.9270 2.4170 2.5316
41 Nb 0.2118 5.83085 72.56434 2.36756 70.37176 2.3705 2.724 2.999 2.4647 3.981 2.425 2.6977
42 Mo 0.2370 6.35098 73.21302 2.67445 70.41914 2.5202 2.813 2.999 2.6251 3.996 2.447 2.8655
43 Tc 0.2631 5.57003 72.03397 2.13536 70.33831 2.6769 3.044 2.907 2.7932 4.051 2.452 3.0425
Copyright 2002 Marcel Dekker, Inc.
44 Ru 0.2901 9.99349 78.27969 5.01647 70.76636 2.8379 3.320 2.770 2.9669 4.303 2.302 3.2240
45 Rh 0.3180 6.55219 73.56318 2.98931 70.48554 3.0038 3.653 2.586 3.1469 4.091 2.502 3.4119
46 Pd 0.3425 6.32821 73.29827 2.95809 70.49645 3.1733 4.146 2.250 3.3303 4.344 2.337 3.6043
47 Ag 0.3766 5.65378 72.26740 2.49736 70.43066 3.3511 4.111 2.321 3.5237 4.217 2.487 3.8058
48 Cd 0.4129 3.12801 1.54731 0.70511 70.16236 3.5375 3.971 2.454 3.7270 4.328 2.420 3.0180
49 In 0.4530 2.78289 2.29146 0.30064 70.09544 3.7301 4.293 2.217 3.9380 4.620 2.196 4.2375
50 Sn 0.4946 2.57173 2.70219 0.11208 70.07092 3.9288 4.834 1.753 4.1561 4.592 2.238 4.4647
51 Sb 0.5379 2.52338 2.80299 0.07373 70.06562 4.1322 4.637 1.940 4.3804 4.584 2.279 4.6983
52 Te 0.5827 4.98562 71.29951 2.27938 70.44589 4.3414 4.618 1.955 4.6120 4.554 2.323 4.9392
53 I 0.6293 3.00387 2.36374 0.26067 70.09483 4.5571 4.577 2.048 4.8521 4.625 2.318 5.1881
54 Xe 0.6775 3.00203 2.56894 0.10775 70.06179 4.7822 4.572 2.073 5.1037 4.599 2.375 5.4528
55 Cs 1.2171 3.16785 2.42799 0.12755 70.05218 5.0119 4.215 2.528 5.3594 4.563 2.482 5.7143
56 Ba 1.2928 3.19080 2.48392 0.08776 70.04429 5.2470 4.363 2.384 5.6236 4.601 2.470 5.9888
57 La 1.3613 3.23265 2.53710 0.05145 70.03744 5.4827 4.401 2.397 5.8906 4.698 2.404 6.2663
58 Ce 1.4346 3.29411 2.53524 0.05465 70.03926 5.7234 4.449 2.402 6.1642 4.702 2.470 6.5488
59 Pr 1.5110 3.35891 2.54117 0.04839 70.03842 5.9642 4.469 2.438 6.4404 4.757 2.467 6.8348
60 Nd 1.5753 3.41765 2.51947 0.05777 70.04029 6.2079 4.463 2.500 6.7215 4.778 2.490 7.1260
61 Pm 1.6540 3.45084 2.59000 0.00563 0.02894 6.4593 4.500 2.528 7.0128 4.854 2.449 7.4279
62 Sm 1.7228 3.49847 2.55440 0.02671 70.03367 6.7162 4.536 2.502 7.3118 4.861 2.469 7.7368
63 Eu 1.8000 3.54762 2.56798 0.01585 70.03181 6.9769 4.553 2.556 7.6171 4.924 2.428 8.0520
64 Gd 1.8808 3.58135 2.57591 0.00022 70.02718 7.2428 4.583 2.543 7.9303 4.927 2.467 8.3756
65 Tb 1.9675 3.61719 2.61430 70.03278 70.01918 7.5140 4.633 2.537 5.2516 4.985 2.423 8.7080
66 Dy 2.0468 3.67219 2.57359 70.00002 70.02988 7.7901 4.655 2.572 8.5806 5.014 2.420 9.0458
67 Ho 2.1283 3.72181 2.55757 0.010760 70.03537 8.0711 4.695 2.558 8.9178 5.047 2.426 9.3942
68 Er 2.2065 3.77845 2.52004 0.05588 70.04817 8.3579 4.727 2.581 9.2643 5.086 2.414 9.7513
69 Tm 2.3068 3.82800 2.53261 0.03019 70.03886 8.6480 4.767 2.592 9.6169 5.128 2.398 10.1157
70 Yb 2.3981 3.86458 2.50726 0.06374 70.05192 8.9436 4.802 2.569 9.9782 5.082 2.815 10.4864
71 Lu 2.4912 3.91793 2.48402 0.08777 70.06146 9.2441 4.847 2.560 10.3386 5.181 2.431 10.8704
72 Hf 2.6009 3.94527 2.52598 0.03404 70.04239 9.5607 4.866 2.629 10.7394 5.214 2.377 11.2707
73 Ta 2.7080 3.99165 2.50855 0.05734 70.05289 9.8811 4.911 2.584 11.1361 5.237 2.468 11.6815
74 W 2.8196 4.02988 2.51689 0.04574 70.05071 10.2068 4.940 2.624 11.5440 5.276 2.318 12.0998
75 Re 2.9317 4.07043 2.52499 0.03041 70.04512 10.5353 4.979 2.629 11.9687 5.302 2.373 12.5267
76 Os 3.0485 4.10244 2.53312 0.01448 70.03948 10.8709 5.006 2.661 12.3850 5.323 2.314 12.9680
77 Ir 3.1737 4.14636 2.52893 0.01458 70.04009 11.2152 5.050 2.673 12.8241 5.355 2.361 13.4185
78 Pt 3.2960 4.18356 2.53390 0.00520 70.03783 11.5637 5.084 2.666 13.2726 5.375 2.278 13.8799
Copyright 2002 Marcel Dekker, Inc.
Table 2 Continued
E
0
<E<E
L3
E
L3
< E < E
L2
E
L2
< E < E
L1
Z Sy E
0
=E
edge
m
1
m
2
m
3
m
4
E
L3
l
31
l
32
E
L2
l
21
l
22
E
L1
79 Au 3.4249 4.22770 2.52409 0.01426 70.04262 11.9187 5.125 2.690 13.7336 5.406 2.303 14.3528
80 Hg 3.5616 4.26093 2.51954 0.02092 70.04665 12.2839 5.156 2.643 14.2087 5.430 2.337 14.8393
81 Tl 3.7041 4.29302 2.51974 0.01369 70.04376 12.6575 5.184 2.630 14.6979 5.451 2.323 15.3467
82 Pb 3.8507 4.32934 2.51778 0.00460 70.03614 13.0352 5.220 2.644 15.2000 5.512 2.502 15.8608
83 Bi 3.9991 4.36991 2.52184 70.01213 70.02654 13.4186 5.257 2.642 15.7111 5.536 2.453 16.3875
84 Po 4.1494 4.41866 2.52158 70.00971 70.03298 13.8138 5.301 2.625 16.2443 5.555 2.387 16.9393
85 At 4.3170 4.46126 2.51394 0.00640 70.04593 14.2135 5.347 2.642 16.7847 5.574 2.335 17.4930
86 Rn 4.4820 4.45433 2.51411 70.00666 70.03871 14.6194 5.334 2.637 17.3371 5.535 2.281 18.0490
87 Fr 4.6520 4.49562 2.51108 70.01358 70.02823 15.0312 5.376 2.647 17.9065 5.650 2.494 18.6390
88 Ra 4.8220 4.52666 2.50733 0.00692 70.05016 15.4444 5.410 2.653 18.4843 5.644 2.411 19.2367
89 Ac 5.0020 4.56932 2.50607 70.01644 70.03124 15.8710 5.454 2.662 19.0832 5.761 2.594 19.8400
90 Th 5.1823 4.59153 2.50190 70.00815 70.03948 16.3003 5.490 2.696 19.6932 5.883 2.800 20.4721
91 Pa 5.3669 4.63772 2.49776 70.00936 70.03490 16.7331 5.525 2.666 20.3137 5.798 2.534 21.1046
92 U 5.5480 4.65216 2.49170 70.02703 70.01144 17.1663 5.541 2.672 20.9476 5.829 2.571 21.7574
Source: From Orlic et al., 1993. Reprinted with kind permission from Elsevier Science NL, Sara Burgehartstraat 25, 1055 KV Amsterdam, The Netherlands.
Copyright 2002 Marcel Dekker, Inc.
Table 3 Parameters for Calculating the Total Mass Attenuation Coefcients for Photon Energies Between M
1
, M
2
, . . ., M
5
Absorption Edges and for Energies
Between M
5
and Lower Energy Limit E
0
E
0
<E<E
edge
E
M5
<E<E
M4
E
M4
<E<E
M3
E
M3
<E<E
M2
E
M2
<E<E
M1
Z E
0
n
1
n
2
n
3
n
4
E
M5
m
42
m
41
E
M4
m
32
m
31
E
M3
m
22
m
21
E
M2
m
12
m
11
E
M1
41 0.1085 115.013 769.4208 14.9364 71.0543
42 0.1085 106.662 767.3126 15.2602 71.1408
43 0.1085 3.5107 2.3889 70.4178 0.0324
44 0.1085 89.5750 757.9740 13.8424 71.1025
45 0.1085 80.5898 757.4238 15.0813 71.3052
46 0.1328 76.3873 1.0408 1.9749 70.3214
47 0.1328 29.0389 718.5282 5.5664 70.5470
48 0.1487 11.0289 76.3556 2.6658 70.3046
49 0.1717 174.946 7138.963 38.3880 73.5077
50 0.1717 42.9485 731.1679 9.1280 70.8669
51 0.1511 39.8331 730.6663 9.5340 70.9502
52 0.1511 37.0196 728.3109 8.8955 70.8939
53 0.1511 14.2756 79.7352 3.9179 70.4554
54 0.1833 23.1086 718.4405 6.7995 70.7732
55 0.2122 32.5527 727.5567 9.7373 71.0874 0.7330 2.364 2.621 1.0650 3.983 2.072 1.2171
56 0.2770 31.6390 728.1233 10.4351 71.2218 0.7905 0.7961 3.768 2.088 1.0622 3.897 2.098 1.1367 3.957 2.097 1.2928
57 0.1926 28.6047 724.1290 8.8027 71.0079 0.8473 0.8485 3.554 2.190 1.1234 3.870 2.132 1.2044 3.970 2.114 1.3613
58 0.2122 3.9181 70.9835 1.6398 70.2735 0.8861 0.9013 3.194 2.376 1.1854 3.852 2.162 1.2728 4.007 2.120 1.4346
59 0.3117 4.6993 0.0130 0.6969 70.0975 0.9349 0.9511 3.317 2.355 1.2422 5.314 1.551 1.3774 2.489 2.863 1.5110
60 0.3117 20.5246 716.5997 6.5042 70.7721 0.9843 0.9999 5.484 1.419 1.2974 3.879 2.193 1.4028 3.993 2.167 1.5753
61 0.2770 16.9684 713.1825 5.4255 70.6587 1.0269 96.373 735.542 1.0515 6.609 0.921 1.3569 3.845 2.234 1.4714 4.039 2.170 1.6540
62 0.2770 11.8798 77.7898 3.5371 70.4381 1.0802 89.021 733.267 1.1060 6.352 1.023 1.4198 3.825 2.255 1.5407 3.986 2.206 1.7228
63 0.2770 5.6437 71.4828 1.4474 70.2096 1.1309 83.564 731.660 1.1606 6.316 1.039 1.4806 3.825 2.278 1.6139 4.005 2.220 1.8000
64 0.2770 6.6183 72.4260 1.7586 70.2443 1.1852 76.154 729.235 1.2172 8.597 0.048 1.5440 3.792 2.308 1.6883 4.032 2.217 1.8808
65 0.1717 75.6112 10.4330 72.6765 0.2602 1.2412 73.753 728.683 1.2750 6.289 1.062 1.6113 3.863 2.296 1.7677 4.027 2.242 1.9675
66 0.2122 1.4535 3.1211 70.1907 70.0138 1.2949 69.682 727.456 1.3325 6.096 1.157 1.6756 3.809 2.342 1.8418 4.060 2.243 2.0468
67 0.2770 4.077 0.1743 0.9232 70.1528 1.3514 65.046 725.898 1.3915 5.446 1.471 1.7412 3.811 2.363 1.9228 4.092 2.245 2.1283
68 0.2122 1.6315 2.7196 0.0767 70.0607 1.4093 60.848 724.477 1.4533 5.353 1.523 1.8118 3.847 2.368 2.0058 4.087 2.269 2.2065
Copyright 2002 Marcel Dekker, Inc.
Table 3 Continued
E
0
<E<E
edge
E
M5
<E<E
M4
E
M4
<E<E
M3
E
M3
<E<E
M2
E
M2
<E<E
M1
Z E
0
n
1
n
2
n
3
n
4
E
M5
m
42
m
41
E
M4
m
32
m
31
E
M3
m
22
m
21
E
M2
m
12
m
11
E
M1
69 0.1926 75.0355 9.4382 72.1146 0.1726 1.4677 63.572 726.240 1.5146 5.841 1.257 1.8845 3.889 2.370 2.0998 4.114 2.278 2.3068
70 0.1085 71.8380 6.2849 71.0736 0.0593 1.5278 54.494 722.411 1.5763 5.900 1.222 1.9498 3.916 2.373 2.1730 4.125 2.287 2.3981
71 0.1085 70.3276 4.8522 70.6175 0.0113 1.5885 57.769 724.531 1.6394 6.893 0.681 2.0236 3.876 2.422 2.2635 4.129 2.309 2.4912
72 0.1085 70.1868 4.8402 70.6304 0.0127 1.6517 55.928 724.196 1.7164 7.170 0.527 2.1076 3.916 2.420 2.3654 4.158 2.311 2.6009
73 0.1085 71.9405 6.5977 71.1795 0.0672 1.7351 53.589 723.567 1.7932 6.974 0.627 2.1940 3.957 2.420 2.4687 4.146 2.343 2.7080
74 1.1085 71.1407 5.9151 70.9671 0.0439 1.8092 50.662 722.588 1.8716 6.894 0.671 2.2810 4.019 2.408 2.5749 4.178 2.344 2.8196
75 0.1085 70.3842 5.1929 70.7225 0.0155 1.8829 30.913 712.446 1.9489 6.854 0.666 2.3673 4.063 2.405 2.6816 4.214 2.346 2.9317
76 0.1085 71.8806 6.8129 71.2684 0.0733 1.9601 54.605 725.802 2.0308 6.900 0.581 2.4572 4.032 2.445 2.7922 4.228 2.356 3.0485
77 0.1085 72.9368 7.9106 71.6078 0.1054 2.0404 48.393 722.958 2.1161 7.249 0.343 2.5507 4.126 2.410 2.9087 4.307 2.327 3.1737
78 0.1328 1.5113 3.4177 70.1106 70.0581 2.1216 45.233 721.695 2.2019 7.209 0.344 2.6454 4.166 2.408 3.0265 4.323 2.338 3.2960
79 0.1511 3.5698 1.1260 0.7499 70.1652 2.2057 43.502 721.199 2.2911 7.102 0.393 2.7430 4.199 2.414 3.1478 4.317 2.370 3.4249
80 0.2122 3.6427 1.0962 0.7581 70.1655 2.2949 40.458 719.895 2.3849 6.847 0.540 2.8471 4.187 2.445 3.2785 4.382 2.344 3.5616
81 0.1926 3.6296 1.1014 0.7912 70.1770 2.3893 31.751 715.013 2.4851 6.055 1.064 2.9566 4.191 2.467 3.4157 4.314 2.419 3.7041
82 0.1717 4.6393 70.1055 1.2868 70.2451 2.4840 20.254 78.092 2.5856 5.339 1.574 3.0664 4.259 2.443 3.5542 4.453 2.332 3.8507
83 0.1926 5.7697 71.3933 1.7810 70.3078 2.5796 9.171 71.075 2.6876 4.475 2.212 3.1769 4.277 2.459 3.6963 4.383 2.420 3.9991
84 0.2122 5.6131 71.3279 1.8314 70.3245 2.6839 70.200 5.161 2.7980 3.586 2.895 3.3019 4.327 2.459 3.8541 4.460 2.395 4.1494
85 0.3117 5.1757 70.6452 1.5552 70.2925 2.7867 74.893 8.501 2.9087 3.141 3.276 3.4260 4.372 2.457 4.0080 4.494 2.402 4.3170
86 0.3924 4.2011 0.6074 1.0490 70.2281 2.8924 77.059 10.172 3.0215 3.222 3.251 3.5280 4.382 2.442 4.1590 4.500 2.386 4.4820
87 0.3924 4.5198 0.1763 1.2693 70.2654 2.9999 72.615 7.198 3.1362 3.649 2.961 3.6638 4.405 2.458 4.3270 4.523 2.398 4.6520
88 0.3924 5.5170 71.1906 1.9034 70.3623 3.1049 0.627 4.924 3.2484 3.982 2.716 3.7918 4.391 2.501 4.4895 4.511 2.442 4.8220
89 0.3924 6.3317 72.2638 2.3875 70.4354 3.2190 2.581 3.529 3.3702 4.255 2.525 3.9098 4.460 2.475 4.6560 4.599 2.387 5.0020
90 0.2770 4.1761 0.8164 1.0031 70.2383 3.3320 3.793 2.621 3.4908 4.355 2.455 4.0461 4.495 2.462 4.8304 4.576 2.437 5.1823
91 0.2770 4.3146 0.8409 0.9488 70.2273 3.4418 3.639 2.790 3.6112 4.369 2.484 4.1738 4.526 2.477 5.0009 4.603 2.457 5.3669
92 0.2770 3.9954 1.2801 0.7662 70.2038 3.5517 3.662 2.788 3.7276 4.409 2.460 4.3034 4.527 2.492 5.1822 4.614 2.458 5.5480
Source: From Orlic et al., 1993. Reprinted with kind permission from Elsevier Science NL, Sara Burgehartstraat 25, 1055 KV Amsterdam, The Netherlands.
Copyright 2002 Marcel Dekker, Inc.
APPENDIX XI: TOTAL MASS ATTENUATIONCOEFFICIENTS FOR LOW-ENERGYKLINES
Absorber Emitter wavelength (A
) energy (keV)
B C N O F Ne Na Mg Al
Atomic 6.67 1
a
4.48 1 3.16 1 2.36 1 1.83 1 1.46 1 1.19 1 9.89 0 8.34 0
number Element 1.83 71 2.77 71 3.92 71 5.25 71 6.77 71 8.49 71 1.04 0 1.25 0 1.49 0
1 H 1.81 3 4.89 2 1.51 2 5.85 1 2.62 1 1.22 1 6.32 0 3.76 0 2.23 0
2 He 1.20 4 4.06 3 1.25 3 5.07 2 2.23 2 1.06 2 5.58 1 3.15 1 1.77 1
3 Li 2.98 4 1.08 4 3.67 3 1.60 3 7.57 2 3.84 2 2.09 2 1.20 2 6.82 1
4 Be 5.82 4 2.29 4 8.40 3 3.78 3 1.86 3 9.59 2 5.44 2 3.26 2 1.89 2
5 B 3.93 3 3.75 4 1.53 4 7.22 3 3.63 3 1.93 3 1.10 3 6.73 2 3.94 2
6 C 7.32 3 2.75 3 2.44 4 1.22 4 6.29 3 3.43 3 1.98 3 1.21 3 7.22 2
7 N 1.19 4 4.71 3 1.83 3 1.73 4 9.23 3 5.14 3 3.02 3 1.90 3 1.14 3
8 O 1.74 4 6.90 3 2.68 3 1.27 3 1.21 4 6.94 3 4.17 3 2.67 3 1.63 3
9 F 2.45 4 9.73 3 3.73 3 1.79 3 9.12 2 8.54 3 5.23 3 3.44 3 2.11 3
10 Ne 3.72 4 1.36 4 5.70 3 2.71 3 1.39 3 7.54 2 7.11 3 4.55 3 2.84 3
11 Na 4.25 4 1.80 4 7.26 3 3.52 3 1.83 3 1.01 3 5.93 2 5.65 3 3.53 3
12 Mg 5.62 4 2.44 4 1.00 4 4.89 3 2.55 3 1.42 3 8.37 2 5.06 2 4.39 3
13 Al 6.44 4 2.92 4 1.23 4 6.10 3 3.22 3 1.80 3 1.07 3 6.63 2 4.07 2
14 Si 7.48 4 3.47 4 1.58 4 7.93 3 4.22 3 2.37 3 1.41 3 8.93 2 5.52 2
15 P 7.49 4 4.19 4 1.93 4 9.74 3 5.19 3 2.92 3 1.74 3 1.09 3 6.78 2
16 S 8.18 4 4.95 4 2.38 4 1.21 4 6.52 3 3.69 3 2.21 3 1.39 3 8.64 2
17 Cl 7.44 3 5.17 4 2.67 4 1.39 4 7.50 3 4.27 3 2.57 3 1.63 3 1.01 3
18 Ar 9.51 3 5.24 4 3.00 4 1.59 4 8.59 3 4.89 3 2.93 3 1.85 3 1.15 3
19 K 1.14 4 5.92 3 3.51 4 1.94 4 1.07 4 6.13 3 3.71 3 2.36 3 1.47 3
20 Ca 1.31 4 7.15 3 3.49 4 2.21 4 1.24 4 7.19 3 4.38 3 2.75 3 1.74 3
21 Sc 1.43 4 7.80 3 3.97 3 2.34 4 1.34 4 7.79 3 4.74 3 2.99 3 1.90 3
22 Ti 1.54 4 8.45 3 4.35 3 2.20 4 1.45 4 8.59 3 5.27 3 3.34 3 2.13 3
23 V 1.68 4 9.22 3 4.77 3 1.66 4 1.57 4 9.40 3 5.81 3 3.70 3 2.36 3
24 Cr 2.09 4 1.11 4 5.61 3 3.13 3 1.58 4 1.09 4 6.75 3 4.30 3 2.74 3
25 Mn 2.23 4 1.20 4 6.13 3 3.46 3 1.68 4 1.16 4 7.27 3 4.67 3 3.00 3
26 Fe 2.61 4 1.40 4 7.10 3 4.00 3 2.31 3 1.31 4 8.22 3 5.31 3 3.42 3
27 Co 2.86 4 1.54 4 7.85 3 4.41 3 2.58 3 1.22 4 8.84 3 5.73 3 3.71 3
Copyright 2002 Marcel Dekker, Inc.
Appendix XI Continued
Absorber Emitter wavelength (A
) energy (keV)
B C N O F Ne Na Mg Al
Atomic 6.67 1
a
4.48 1 3.16 1 2.36 1 1.83 1 1.46 1 1.19 1 9.89 0 8.34 0
number Element 1.83 71 2.77 71 3.92 71 5.25 71 6.77 71 8.49 71 1.04 0 1.25 0 1.49 0
28 Ni 3.34 4 1.81 4 9.19 3 5.25 3 3.01 3 1.80 3 1.02 4 6.58 3 4.26 3
29 Cu 3.65 4 1.96 4 9.96 3 5.94 3 3.20 3 2.01 3 9.47 3 7.04 3 4.52 3
30 Zn 4.03 4 2.26 4 1.17 4 6.56 3 3.73 3 2.25 3 5.26 3 7.51 3 4.96 3
31 Ga 4.18 4 2.42 4 1.26 4 7.10 3 4.08 3 2.47 3 1.52 3 6.92 3 5.23 3
32 Ge 4.41 4 2.64 4 1.40 4 7.89 3 4.56 3 2.75 3 1.70 3 4.72 3 5.63 3
33 As 4.41 4 2.86 4 1.55 4 8.80 3 5.10 3 3.08 3 1.95 3 1.30 3 5.32 3
34 Se 4.49 4 3.03 4 1.69 4 9.64 3 5.61 3 3.38 3 2.17 3 1.53 3 4.86 3
35 Br 4.19 4 3.22 4 1.86 4 1.07 4 6.28 3 3.97 3 2.44 3 1.68 3 1.01 3
36 Kr 3.94 4 3.28 4 2.05 4 1.19 4 6.98 3 4.23 3 2.66 3 1.72 3 1.14 3
37 Rb 3.52 4 3.56 4 2.17 4 1.28 4 7.57 3 4.63 3 2.92 3 1.89 3 1.25 3
38 Sr 2.83 4 3.47 4 2.29 4 1.37 4 8.21 3 5.04 3 3.20 3 2.08 3 1.38 3
39 Y 2.04 4 3.10 4 2.34 4 1.51 4 9.06 3 5.58 3 3.55 3 2.32 3 1.52 3
40 Zr 6.41 3 3.06 4 2.46 4 1.61 4 9.77 3 6.03 3 3.86 3 2.53 3 1.66 3
41 Nb 4.28 3 2.96 4 2.67 4 1.78 4 1.08 4 6.69 3 4.27 3 2.80 3 1.84 3
42 Mo 4.66 3 2.23 4 2.29 4 1.85 4 1.14 4 7.09 3 4.55 3 2.99 3 1.97 3
43 Tc 4.99 3 1.41 4 2.29 4 1.78 4 1.19 4 7.45 3 4.90 3 3.57 3 2.12 3
44 Ru 5.49 3 4.18 3 2.47 4 1.95 4 1.30 4 8.16 3 5.37 3 3.92 3 2.32 3
45 Rh 5.87 3 4.64 3 2.48 4 1.94 4 1.38 4 8.79 3 5.80 3 4.23 3 2.51 3
46 Pd 6.09 3 5.22 3 2.21 4 1.80 4 1.46 4 9.42 3 6.23 3 4.54 3 2.69 3
47 Ag 6.34 3 5.48 3 1.10 4 1.91 4 1.47 4 1.00 4 6.66 3 4.86 3 2.90 3
48 Cd 6.45 3 5.75 3 4.10 3 2.01 4 1.47 4 1.04 4 7.09 3 5.84 3 4.48 3
49 In 6.61 3 6.06 3 4.42 3 2.23 4 1.36 4 1.10 4 7.51 3 6.20 3 4.75 3
50 Sn 6.64 3 6.27 3 4.70 3 9.12 3 1.38 4 1.08 4 7.88 3 6.50 3 4.99 3
51 Sb 6.71 3 6.50 3 4.94 3 3.36 3 1.48 4 1.13 4 8.26 3 6.82 3 5.25 3
52 Te 6.65 3 6.58 3 5.09 3 3.53 3 1.75 4 1.13 4 8.35 3 6.90 3 5.32 3
53 I 5.77 3 6.99 3 5.52 3 3.88 3 1.18 4 1.02 4 8.24 3 6.93 3 5.39 3
54 Xe 5.79 3 7.27 3 5.88 3 4.17 3 2.84 3 1.08 4 8.72 3 6.75 3 5.30 3
55 Cs 5.91 3 7.52 3 6.13 3 4.40 3 3.02 3 1.15 4 9.79 3 6.67 3 5.29 3
Copyright 2002 Marcel Dekker, Inc.
56 Ba 4.09 3 7.64 3 6.21 3 4.52 3 3.13 3 1.16 4 7.51 3 6.36 3 5.21 3
57 La 3.76 3 7.74 3 6.32 3 4.66 3 3.26 3 4.21 3 7.89 3 6.56 3 5.22 3
58 Ce 6.79 3 7.98 3 7.14 3 5.16 3 3.57 3 2.45 3 8.81 3 7.12 3 5.43 3
59 Pr 8.74 3 8.77 3 7.60 3 5.45 3 3.77 3 2.59 3 9.49 3 6.10 3 5.26 3
60 Nd 1.05 4 9.43 3 7.92 3 5.64 3 3.89 3 2.68 3 8.29 3 6.19 3 5.47 3
61 Pm 1.26 4 1.04 4 8.40 3 5.92 3 4.07 3 2.79 3 2.51 3 6.59 3 5.73 3
62 Sm 1.48 4 1.13 4 8.95 3 6.26 3 4.27 3 2.93 3 2.01 3 6.94 3 5.84 3
63 Eu 1.58 4 1.20 4 9.42 3 6.58 3 4.49 3 3.08 3 2.12 3 7.32 3 4.88 3
64 Gd 1.49 4 1.12 4 9.39 3 6.60 3 4.53 3 3.12 3 2.17 3 4.57 3 5.01 3
65 Tb 1.87 4 1.24 4 9.93 3 7.38 3 4.99 3 3.39 3 2.34 3 1.61 3 5.49 3
66 Dy 2.04 4 1.37 4 1.06 4 7.83 3 5.26 3 3.56 3 2.47 3 1.73 3 5.68 3
67 Ho 2.16 4 1.47 4 1.13 4 8.29 3 5.56 3 3.76 3 2.61 3 1.87 3 5.70 3
68 Er 2.32 4 1.60 4 1.21 4 8.86 3 5.94 3 4.00 3 2.77 3 2.01 3 2.84 3
69 Tm 2.43 4 1.72 4 1.29 4 9.49 3 6.36 3 4.27 3 2.96 3 2.14 3 1.56 3
70 Yb 2.28 4 1.79 4 1.35 4 9.94 3 6.67 3 4.47 3 3.10 3 2.32 3 1.47 3
71 Lu 2.15 4 1.75 4 1.36 4 1.01 4 6.86 3 4.62 3 3.22 3 2.42 3 1.54 3
72 Hf 2.13 4 1.80 4 1.27 4 9.96 3 7.20 3 4.86 3 3.38 3 2.54 3 1.62 3
73 Ta 2.04 4 1.82 4 1.32 4 1.04 4 7.54 3 5.11 3 3.53 3 2.45 3 1.77 3
74 W 1.93 4 1.85 4 1.37 4 1.09 4 7.90 3 5.36 3 3.70 3 2.57 3 1.86 3
75 Re 1.76 4 1.84 4 1.39 4 1.12 4 8.18 3 5.57 3 3.87 3 2.69 3 1.95 3
76 Os 1.58 4 1.67 4 1.39 4 1.12 4 8.36 3 5.74 3 4.02 3 2.92 3 1.99 3
77 Ir 1.38 4 1.61 4 1.44 4 1.15 4 8.36 3 6.09 3 4.27 3 3.11 3 2.12 3
78 Pt 1.13 4 1.57 4 1.50 4 1.12 4 8.83 3 6.43 3 4.51 3 3.28 3 2.24 3
79 Au 9.09 3 1.49 4 1.51 4 1.16 4 9.23 3 6.74 3 4.68 3 3.28 3 2.29 3
80 Hg 6.62 3 1.47 4 1.50 4 1.17 4 9.30 3 6.95 3 4.77 3 3.31 3 2.36 3
81 Tl 5.42 3 1.29 4 1.40 4 1.21 4 9.55 3 7.24 3 5.13 3 3.88 3 2.50 3
82 Pb 4.19 3 1.14 4 1.38 4 1.23 4 8.93 3 7.10 3 5.34 3 4.04 3 2.61 3
83 Bi 3.17 3 1.01 4 1.35 4 1.26 4 9.23 3 7.35 3 5.55 3 4.20 3 2.72 3
84 Po 2.54 3 8.62 3 1.34 4 1.33 4 9.68 3 7.72 3 5.82 3 4.42 3 2.88 3
85 At 2.28 3 6.67 3 1.31 4 1.23 4 1.02 4 8.14 3 5.70 3 4.57 3 3.04 3
86 Rn 2.52 3 4.13 3 1.21 4 1.16 4 1.02 4 8.26 3 5.72 3 4.46 3 3.04 3
87 Fr 2.69 3 3.02 3 1.10 4 1.17 4 1.05 4 7.52 3 5.99 3 4.57 3 3.18 3
88 Ra 2.66 3 2.42 3 9.31 3 1.17 4 1.04 4 7.75 3 6.22 3 4.68 3 3.30 3
89 Ac 2.49 3 2.59 3 8.17 3 1.15 4 1.01 4 7.88 3 6.35 3 4.49 3 3.40 3
90 Th 1.99 3 2.12 3 6.44 3 1.11 4 8.83 3 7.85 3 6.22 3 4.64 3 3.48 3
Copyright 2002 Marcel Dekker, Inc.
Appendix XI Continued
Absorber Emitter wavelength (A
) energy (keV)
B C N O F Ne Na Mg Al
Atomic 6.67 1
a
4.48 1 3.16 1 2.36 1 1.83 1 1.46 1 1.19 1 9.89 0 8.34 0
number Element 1.83 71 2.77 71 3.92 71 5.25 71 6.77 71 8.49 71 1.04 0 1.25 0 1.49 0
91 Pa 2.34 3 2.38 3 3.44 3 1.13 4 9.45 3 8.35 3 5.96 3 4.97 3 3.71 3
92 U 2.57 3 2.30 3 2.70 3 1.09 4 9.39 3 8.36 3 6.03 3 4.96 3 3.76 3
93 Np 2.93 3 2.41 3 2.62 3 9.65 3 9.74 3 8.59 3 6.34 3 5.21 3 3.74 3
94 Pu 3.73 3 2.39 3 2.72 3 6.59 3 1.01 4 8.24 3 6.66 3 5.40 3 3.89 3
a
Notation as in Appendix VI, Table 4.
Source: From Veigele, 1974.
Copyright 2002 Marcel Dekker, Inc.
APPENDIX XII: CORRESPONDENCE BETWEENOLD SIEGBAHNANDNEWIUPAC
NOTATIONX-RAYDIAGRAMLINES
Siegbahn IUPAC Siegbahn IUPAC Siegbahn IUPAC
Ka
1
K-L
3
La
1
L
3
-M
5
Lg
1
L
2
-N
4
Ka
2
K-L
2
La
2
L
3
-M
4
Lg
2
L
1
-N
2
Kb
1
K-M
3
Lb
1
L
2
-M
4
Lg
3
L
1
-N
3
Kb
I
2
K-N
3
Lb
2
L
3
-N
5
Lg
4
L
1
-O
3
Kb
II
2
K-N
2
Lb
3
L
1
-M
3
Lg
/
4
L
1
-O
2
Kb
3
K-M
2
Lb
4
L
1
-M
2
Lg
5
L
2
-N
1
Kb
I
4
K-N
5
Lb
5
L
3
-O
4,5
Lg
6
L
2
-O
4
Kb
II
4
K-N
4
Lb
6
L
3
-N
1
Lg
8
L
2
-O
1
Kb
4x
K-N
4
Lb
7
L
3
-O
1
Lg
/
8
L
2
-N
6(7)
Kb
I
5
K-M
5
Lb
/
7
L
3
-N
6,7
LZ L
2
-M
1
Kb
II
5
K-M
4
Lb
9
L
1
-M
5
Ll L
3
-M
1
Lb
10
L
1
-M
4
Ls L
3
-M
3
Lb
15
L
3
-N
4
Lt L
3
-M
2
Lb
17
L
2
-M
3
Lu L
3
-N
6,7
Lv L
2
-N
6(7)
Ma
1
M
5
-N
7
Ma
2
M
5
-N
6
Mb M
4
-N
6
Mg M
3
-N
5
Mz M
4,5
-N
2,3
Source: From Jenkins et al., 1991.
Copyright 2002 Marcel Dekker, Inc.
REFERENCES
Arndt UW, Brown PJ, Colliex C, Cowley JM, Creagh DC, Dolling G, Fink M, Freund AK,
Hilderbrandt R, Howie A, Hubbell JH, Gjnnes, Lander G, Lynch DF, McAuley WJ, Pynn R,
Ross AW, Rowe JM, Sears VF, Spence JCH, Steeds JW, Valvoda V, Willis BTM, Wilson
AJC, Zvyagin BB. In: Wilson AJC, ed. International Tables for Crystallography. Vol. C.
Mathematical, Physical and Chemical Tables. Dordrecht: Kluwer Academic Publishers, 1992,
p 162.
Auger P. J Phys 6:205, 1925.
Bambynek W, Crasemann B, Fink RW, Freund HU, Mark H, Swift CD, Price RE, Venugopala
Rao P. Rev Mod Phys 44:716, 1972; erratum: 46:853, 1974.
Behrens P. Trends Anal Chem 11:218, 1992a.
Behrens P. Trends Anal Chem 11:237, 1992b.
Bethe HA. In: Annalen der Physik, Band 5. Leipzig: Verlag von Johann Ambrosius Barth, 1930,
p 325.
Bethe HA, Ashkin J. Experimental Nuclear Physics, Vol. 1. New York: Wiley, 1953, p 252.
Birks LS. Electron Probe Microanalysis. New York: WileyInterscience, 1971a, p 51.
Birks LS. Electron Probe Microanalysis. New York: WileyInterscience, 1971b, p 172.
Burhop EHS. The Auger Eect and Other Radiationless Transitions, New York: Cambridge
University Press, 1952.
Burr A. In: Robinson JW, ed. Handbook of Spectroscopy, Vol. I. Cleveland, OH: CRC Press, 1974,
p 24.
Campbell JI, Cookson JA. Nucl Instrum Methods Phys Res B3:185, 1984.
Clark GL. Applied X-Rays, New York: McGraw-Hill, 1955, p 144.
Clark GL. In: Clark GL, ed. Encyclopedia of X-Rays and Gamma Rays. New York: Reinhold
Publishing, 1963, p 1124.
Compton AH. Phys Rev 21:15, 1923a.
Compton AH. Phys Rev 22:409, 1923b.
Compton AH, Allison SK. X-Rays in Theory and Experiment. New York: Van Nostrand, 1935,
p 537.
Condon EU. In: Condon EU, Odishaw H, eds. Handbook of Physics, Part 7. New York: McGraw-
Hill, 1958, p 7.
Davis B, Mitchell DP. Phys Rev 32:331, 1928.
Deslattes RD. Acta Crystallogr A25:89, 1969.
Dirac PAM. The Principles of Quantum Mechanics. 4th ed. Oxford: Clarendon Press=Oxford
University Press, 1958, p 136.
Dyson NA. X-rays in Atomic and Nuclear Physics. London: Longman Group, 1973, p 42.
Evans RD. In: Flu gge S, ed. Encyclopedia of Physics, Vol. XXXIV. Berlin: Springer-Verlag, 1958,
p 218.
Evans RD. In: Gray DE, ed. American Institute of Physics Handbook. New York: McGraw-Hill,
1963, p 8.
Fink RW. In: Robinson, JW, ed. Handbook of Spectroscopy, Vol. I. Cleveland, OH: CRC Press,
1974, p 219.
Goldstein JI, Newbury DE, Echlin P, Joy DC, Fiori C, Lifshin E. Scanning Electron Microscopy
and X-ray Microanalysis. New York: Plenum Press, 1981, p 105.
Green M, Cosslett VE. Proc Phys Soc 78:1206, 1961.
Hanke W, Wernisch J, Po hn C. X-ray Spectrom 14(1):43, 1985.
Hasnain SS, Ed. X-ray Absorption Fine Structure, Proceedings of the Conference Held in York,
August 1990. Chichester: Ellis Horwood, 1991.
Heinrich KFJ. Electron Beam X-Ray Microanalysis. New York: Van NostrandReinhold, l98l,
p 234.
Heitler W. The Quantum Theory of Radiation. 3rd ed. London: Oxford University Press, 1954,
p 207.
Copyright 2002 Marcel Dekker, Inc.
Hirsch PB. In: Thewlis J, ed. Encyclopaedic Dictionary of Physics. Oxford: Pergamon Press, 1962,
p 800.
Hubbell JH. Photon Cross Sections, Attenuation Coecients, and Energy Absorption Coecients
from 10 keV to 100 GeV. Washington, DC: National Bureau of Standards, l969.
Hubbell JH. Int J Appl Rad Isot 33:1269, 1982.
Hubbell JH. Proceedings of the Workshop on New Directions in Soft X-Ray Photoabsorption,
Asilomar Conference Center, Pacic Grove, CA, 1984.
Hubbell JH. Report NISTIR-89-4144, U.S. Department of Commerce, National Institute of
Standards and Technology, August 1989.
Hubbell JH, McMaster WH, Kerr Del Grande N, Mallett JH. In: Ibers JA, Hamilton WC, eds.
International Tables for X-Ray Crystallography, Vol. 4. Birmingham, UK: Kynoch Press,
1974, p 47.
Hubbell JH, Veigele VJ, Briggs EA, Brown RT, Cromer DT, Howerton RJ. J Phys Chem Ref Data
4(3):471, 1975.
Hubbell JH, Trehan PN, Singh N, Chand B, Mehta D, Garg ML, Garg RR, Singh S, Puri S.
J Phys Chem Ref Data 23(2):339, 1994.
Jenkins R, Gould RW, Gedcke D. Quantitative X-Ray Spectrometry. New York: Marcel Dekker,
1981.
Jenkins R, Manne R, Robin J, Senemaud C. Pure Appl Chem 63(5):735, 1991.
Kallithrakas-Kontos N. Spectrochim Acta Part B51:1655, 1996.
Kawai J, Gohshi Y. Spectrochim Acta 41B:265, 1986.
Kawai J, Nakajima K, Gohshi Y. Spectrochim Acta 48B:1281, 1993.
Kenney RW. In: Besanc on RM, ed. Encyclopedia of Physics. New York: Reinhold Publishing, 1966,
p 86.
Klein O, Nishina Y. Z Phys 52:853. 1929.
Koningsberger DC, Prins R, eds. X-ray Absorption. Chemical Analysis Vol. 92, New York: Wiley,
1988.
Kramers HA. Philos Mag 46:836, 1923.
Krause MO. J Phys Chem Ref Data 8(2):307, 1979.
Krause MO, Oliver JH. J Phys Chem Ref Data 8(2):329, 1979.
Krishnan KS. Nature 122:961, 1928.
Kulenkamp H. Ann Phys 69:548, 1922.
Kulenkamp H. Ann Phys 87:597, 1928.
Kulenkamp H. Handbuch der Physik, Bd. 23. Berlin: Geiger-Scheel 1933, p 433.
Lagarde P. Nucl Instrum Methods 208:621, 1983.
Lee PA, Pendry JB. Phys Rev B11:2795, 1975.
Lifshin L. Adv X-ray Anal 19:113, 1976.
Love G, Scott VD, Scanning 4:111, 1981.
Love G, Cox MG, Scott VD. J Phys D 11:7, 1978.
Markowicz AA, Van Grieken RE. Anal Chem 56:2049, 1984.
Markowicz AA, Storms HM, Van Grieken RE. Anal Chem 57:2885, 1985.
Markowicz AA, Storms H, Van Grieken R. X-ray Spectrom 15:131, 1986.
McGuire EJ. Phys Rev A9: 1840, 1974.
McMaster WH, Kerr Del Grande N, Mallett JH, Hubbell JH. Compilation of X-Ray Cross
Sections. Lawrence Radiation Laboratory (Livermore) Report UCRL-50174, Sec. II, Rev. 1,
University of California, 1969.
Merzbacher E, Lewis HW. In: Flu gge S, ed. Encyclopedia of Physics, Vol. 34. Berlin: Springer-
Verlag, 1958, p 166.
Mizuno Y, Ohmura Y. J Phys Soc Japan 22:445, 1967.
Moseley HGJ. Philos Mag 27:703, 1914.
Norrish K, Tao GY. X-ray Spectrom 22:410, 1993.
Orlic I, Loh KK, Sow CH, Tang SM, Thong P. Nucl Instrum Methods Phys Res B74:352, 1993.
Copyright 2002 Marcel Dekker, Inc.
Pirenne MN. The Diraction of X-Rays and Electrons by Free Molecules. New York: Cambridge
University Press, 1946.
Poehn C, Wernisch J, Hanke W. X-ray Spectrom 14(3):120, 1985.
Rao-Sahib TS, Wittry DB. Proceedings of the 6th International Conference on X-Ray Optics and
Microanalysis. Tokyo: University of Tokyo Press, 1972, p 131.
Rebohle L, Lehnert U, Zschornack G. X-ray Spectrom 25:295, 1996.
Sandstro m AE. In: Flu gge S, ed. Encyclopedia of Physics, Vol. 30. Berlin: Springer-Verlag, 1957,
p 78.
Sauter F. In: Annalen der Physik, Band 9. Leipzig: Verlag von Johann Ambrosius Barth, 1931, p 217.
Sayers DE, Lytle FW, Stern EA. Adv X-ray Anal 13:248, 1970.
Scherzer O. Ann Phys 13:137. 1932.
Schreiber TP, Wims AM. X-ray Spectrom 11:42, 1982.
Sherry WM, Vander Sande JB. X-ray Spectrom 6:154, 1977.
Smith DGW, Reed SJB. X-ray Spectrom 10:198, 1981.
Sparks CJ Jr. Phys Rev Lett 33:262, 1974.
Sparks CJ Jr. In: Ramaseshan S, Abrahams SC, Hodgkin D, eds. Proc. Inter-Congress Conference
on Anomalous Scattering held in Madrid, Spain, 2226 April l974. Copenhagen: Munksgaard,
1975, p 175.
Statham PJ. X-ray Spectrom 5:154, 1976.
Statham PJ. Proc Annu Conf Microbeam Anal Soc 14:247, 1979.
Stephenson ST. In: Flu gge S, ed. Encyclopedia of Physics, Vol. 30. Berlin, Springer-Verlag, 1957,
p 337.
Stobbe M. In: Annalen der Physik, Band 7. Leipzig: Verlag von Johann Ambrosius Barth, 1930,
p 661.
Stoev KN, Dlouhy JF. X-ray Anal 37:697, 1994.
Storm E, Israel HI. Nucl Data Tables A7:565, 1970.
Suzuki T. J Phys Soc Japan 21:2087, 1966.
Suzuki T, Nagasawa H. J Phys Soc Japan 39:1579, 1975.
Suzuki T, Kishimoto T, Kaji T, Suzuki T. J Phys Soc Japan 29:730, 1970.
Szalo ki I. X-ray Spectrom 25:21, 1996.
Tertian R, Claisse F. Principles of Quantitative X-Ray Fluorescence Analysis. London: Heyden,
1982.
Tertian R, Broll N. X-ray Spectrom 13:134, 1984.
Veigele WJ. In: Robinson JW, ed. Handbook of Spectroscopy, Vol. I. Cleveland, OH: CRC Press,
1974, p 155.
Ware NG, Reed SJB. J Phys E 6:286, 1973.
Wentzel G. Z Phys 27:257, 1924.
Wernisch J. X-ray Spectrom 14:109, 1985.
Wernisch J, Po hn C, Hanke W, Ebel H. X-ray Spectrom 13:180, 1984.
West RC, ed. Handbook of Chemistry and Physics. 63rd ed. Cleveland, OH: Chemical Rubber Co,
198283, p E-l88.
Copyright 2002 Marcel Dekker, Inc.
2
Wavelength-Dispersive X-ray Fluorescence
Jozef A. Helsen
Catholic University of Leuven, Leuven, Belgium
Andrzej Kuczumow
Lublin Catholic University, Lublin, Poland
I. INTRODUCTION
Moseleys law, formulated soon after the discovery of x-rays by Ro ntgen (Ro ntgen, 1896),
represented the direct and denite onset of the use of x-ray spectrometry in chemistry. The
rst successful period was completed in the beginning of the 1920s by creating order in the
periodic table of the elements and by the discovery of the missing elements (Weebs and
Leicester, 1967). X-ray uorescence (XRF) spectrometry has some unique features so that
it became an irreplaceable tool for the analyst. Not may techniques had such a brilliant
start to their analytical careers!
gThe rst commercial instrument became available around 1940 and was derived
from the x-ray goniometer of a diraction instrument. A reasonable estimation of the
number of wavelength-dispersive (WDXRF) instruments working today is 15000, of
which some 20% are multichannel spectrometers. In addition, about 3000 instruments
working in this mode are attached to the electron microprobes, mainly in Japan. Some
3000 energy-dispersive (EDXRF) instruments are working independently, and some
15002000 others are attached to electron and other [e.g., proton (Uda et al., 1987,
Folkmann and Frederiksen, 1990)] microprobes; a few hundred instruments are con-
nected to synchrotrons as excitation sources. About 300 instruments are working now in
total reection (TRXRF) and several in grazing emission modes (Klockenka mper, 1997).
The average price of a new WDXPF instrument in 1998 is of the order of magnitude of
$250,000, $100,000 for EDXRF, and $200,000 for TRXRF (Claes et al., 1997) and this is
not expected to rise much in the coming years. The calculation of the total money in-
volved in the x-ray instrument market made hereafter was based on the 50% amortization
estimate per instrument. We consider in our account only these parts of Particle-induced
x-ray emission (PIXE), electron microscope, and synchrotron arrangements which can be
treated as x-ray attachments. An auxiliary market was created also around the main
equipment business, covering the spare parts, recent supplements, and service connected
with the maintenance. Then, the commercial value of the instruments running today may
be estimated at some US$4 billion (Markowicz and Van Grieken, 1986; Van Grieken
et al., 1986). An important body of scientic literature relates to x-ray spectrometry
Copyright 2002 Marcel Dekker, Inc.
(Markowicz and Van Grieken, 1986; van Grieken et al., 1986; Jenkins, 1984; de Vries,
1986; Klockenka mper, 1987; Kikkert, 1983; Potts and Webb, 1992; To ro k et al., 1996,
To ro k et al., 1998; Szalo ki et al., 2000; Brill, 1996; Ellis et al., 1997 and Quiset, 1996).
Literature about essentially diraction and scattering problems can also be very useful for
the researchers working in the eld of XRF (Fewster, 1996).
About 10% of the WDXRF instruments are applied in geological research and
prospecting (sequential and simultaneous spectrometers). Most of the simultaneous in-
struments are installed in industry, and the sequential instruments are distributed over
industry, analytical laboratories, and research institutes at universities and public services.
X-ray uorescence spectrometry is unique in element analysis of inanimate matter:
In qualitative analysis, it has unrivaled selectivity for all elements between beryllium and
uranium and also for transuranium elements and an extremely wide dynamic range in
quantitative analysis (ppm to 100%). Although the plasma emission spectrometry [Direct
Current Plasma (DCP) or Inductively Coupled Plasma (ICP)], atomic absorption spec-
trometry (AAS), neutron activation analysis (NAA), and dierent kinds of mass spec-
trometry (MS) have lower or much lower detection limits, the wide dynamic range remains
a unique feature of XRF (over ve decades). Moreover, it possesses the scarce virtue
among spectral methodsthe upper useful limit reaches the concentration of several tens
of percent of the total concentration for the main component. The detection limits at the
ppm level without preconcentration and the limits of precision and accuracy that can be
obtained (connected, in turn, to sample homogeneity and counting statistics) determine
the boundaries of its application.
Earlier books describing the techniques are mentioned in the reference list (Compton
and Allison, 1935; Jenkins and De Vries, 1967; Birks, 1969; Dyson, 1973; Jenkins, 1974;
Bertin, 1975; Tertian and Claisse, 1982; Lachance and Claisse, 1995, Jenkins et al., 1995;
Buhrke et al., 1998the latter is a kind of a guide for sample preparation in XRF and
XRD, but is cited here, because the problem is very important and nontrivial). The
purpose here is to describe the basic principles that allow wavelength dispersion of x-rays
and to describe all the components of a spectrometer.
The phenomenon of x-ray uorescence is only a small part of a much wider problem
of the interaction of charged particles or photons with matter (Feldman and Mayer,
1986). Dierent kinds of secondary particles are then emitted (Fig. 1a). Various combi-
nations of exciting=emitted particles are possible, giving birth to dierent analytical
techniques, with the combination photons as exciting particles=emitted photons as a
base for the x-ray uorescence. The angles the trajectories of the exciting or emitted
particles create with the surface of the sample dene normal XRF and PIXE methods and
total reection versions: TXRF (Chapter 9) and TRPIXE (Vis and van Langevelde, 1991;
van Kan and Vis, 1996a, 1996b, 1997). In addition, for the x-ray uorescence, a grazing-
emission method can be mentioned (GEXRF) (Sasaki, 1990; Becker et al., 1983; de Bokx
and Urbach, 1995; Urbach and de Bokx, 1996; Tsuji et al., 1995; Pe rez and Sa nchez,
1997) (Fig. 1b).
Insofar as the fundamental principles are concerned, we may recall the law of
Moseley (Moseley, 1913, 1914), which is the basis of the qualitative application of XRF:
n
p
c
l
_
kZ s 1
where n is the frequency (in Hz), c is the speed of light (in m=s), l is the wavelength (in m),
k is the constant for a given series, and s is the screening constant. The relationship
Copyright 2002 Marcel Dekker, Inc.
Figure 1 Phenomena accompanying the interaction of x-rays with matter: (a) different
applications for analytical purposes (abbreviations of methods coupled with the use of particular
phenomena are mentioned; (b) different versions according to the incidence and emergence angles
(Itotal reflection XRF or PIXE; IIgrazing emission XRF; IIIgrazing incidencegrazing
emission XRF).
Copyright 2002 Marcel Dekker, Inc.
between the wavelength of the lines belonging to the same spectral series and the atomic
number is illustrated by Figure 2.
For high-Z elements, the XRF spectra become more complex and full exploitation
for qualitative analysis of mixtures is not always simple. Practical unraveling of spectra is
dealt with in Chapter 4.
The intensity of any spectral line is proportional to the number of atoms emitting
photons of energies attributed to this line. However, a simple linear proportionality is not
the rule. From the concepts of mass attenuation coecients introduced in Chapter 1,
it should be clear that the intensity of an analyte line of one species is attenuated by atoms
of the same species and by any other atoms present in the matrix. Matrix refers to all
elements present in a sample except the analyte element. Attenuation by absorption re-
presents a rst complication compromising the proportionality.
If the matrix contains elements with absorption edges of slightly lower energy than
the energy of the characteristic line of the analyte, strong absorption can occur. On the one
hand, this results in further attenuation of the analyte line and, on the other hand, in
enhancement of the spectral line related to the said absorption edge of the other matrix
element. This phenomenon is known as secondary uorescence. The process may be re-
peated with respect to all matrix elements with an absorption edge with a lower energy
than the energy of a uorescence line emitted by another matrix element. Secondary and
higher-order uorescence is the major complication deteriorating the simple proportion-
ality between intensity and concentration.
In Figure 3, three typical kinds of relationship between the relative intensity (mea-
sured intensity divided by the intensity obtained for the pure element) and concentration
(expressed as weight fraction) are represented. Curve I is obtained when the analyte line
undergoes absorption only. If enhancement occurs, the analyte line intensity is higher than
expected from primary excitation only and the curve is situated above the diagonal.
A similar curve, however, can be obtained when the mass attenuation coecient of the
matrix is lower than the related coecient of the element for its own radiation, as shown in
the case for Fe in FeO (curve II). In a special case, the mutual relation between attenuation
Figure 2 Representation of Moseleys law of K and L spectral lines.
Copyright 2002 Marcel Dekker, Inc.
and enhancement coecients may be balanced such that both eects cancel precisely and
then the diagonal is approximated (curve III).
The eects just discussed are all energy (wavelength) dependent. This means that any
calculations converting intensities into concentration should include the attenuation and
higher-order uorescence eects and also integrate over all energies present in the exciting
primary beam above the value of the absorption edge and over all uorescence lines. This
is an ab initio approach for conversion of intensities into concentration implemented by
several authors. We mention here the articles by Gillam and Heal (1952), Sherman (1954,
1955), Shiraiwa and Fujino (1966, 1967), who in the 1950s and the 1960s proposed suitable
equations. Sparks (1976) and Li-Xing (1984) implemented small corrections and rene-
ments concerning the details of these expressions, and de Boer (1990) and de Boer and
Brouwer (1990) gave solutions to some dicult exponential integrals related to the en-
hancement. Important contributions to the numerical side of quantitative XRF are due to
Fernandez (1989), Fernandez and Molinari (1990), Rousseau et al. (1996), and Mantler
(1984). The rst two authors prepared commercial, customer-oriented versions of the
numerical programs XRFPc and CiROU, CiLAC, CiLT, and CiREG, respectively, and
Mantler paved the way for the analyses of multilayers. It should be emphasized that
Fernandez solved the transport equations for x-ray beams, the very general and powerful
but dicult tool for the description of particle beams. Likewise, Gardner and Hawthorne
(1975) obtained similar results by Monte Carlo simulation of x-ray excitation. A specic
approach was presented by Dane et al. (1996). These authors solved the problem of
quantication in situations where either an incomplete data set is available or where a
reasonable assumption about composition and=or depth prole cannot be done. Then, the
idea of processing whole sets of solutions (populations or strings or generations) has been
introduced, instead of processing the particular concentration values in consecutive
iterations.
In the fundamental parameter methods, the incidence and takeo angles are the
important parameters. As a rule, they are kept constant during the traditional run of
the analysis. However, there have been signicant eorts to use the angle-resolved version
Figure 3 Absorption phenomena in a heavy matrix (curve I), in a light matrix (curve II), and in a
neural matrix (curve III).
Copyright 2002 Marcel Dekker, Inc.
of x-ray XRF spectrometers for depth proling of the samples (Gries, 1992; Schmitt
et al., 1994).
For a complete treatment of equations deriving the line intensities in x-ray uor-
escence spectrum from the elemental sample composition, the reader referred is to
Chapter 5.
II. FUNDAMENTALS OF WAVELENGTHDISPERSION
The crystal monochromator is the heart of a wavelength-dispersive spectrometer. Wave-
length dispersion of electromagnetic radiation in the x-ray region cannot be performed, as
a rule, by normal gratings but only by diraction on crystals or, for the long-wavelength
regions, on multilayers. We briey explain the principles because the construction features
of the monochromator are directly derived from them.
Consider a monochromatic beam of x-rays of wavelength l with their electrical
vectors of equal amplitude in phase along any point of the direction of propagation.
Assume further that the beam is parallel and is incident on a crystal at an angle W between
a given crystal plane (and all the planes parallel to this rst) and the incident beam di-
rection. The beam is scattered and diracted rays of equal l result but interfering con-
structively only in those directions for which the phase relationship is conserved. This
happens at an angle W for scattered rays 1 and 2 (Fig. 4), for which the path dierence
ABC of ray 2 with ray 1 is equal to an integral number n of wavelengths. From Figure 4, it
is clear that ABBCd sin Wd sin Wnl or, according to Bragg and Bragg (1913),
who rst formulated this relation, written as
nl 2d sin W 2
where n denotes the number of wavelength dierences between the rays scattered by the
adjacent planes. Following the recommendation of the IUPAC (Freiser and Nancollas,
1987), n is the order of diraction. If n 1, the dierence is one wavelength and the
diraction is said to be of rst order. If n 2, the dierence is two wavelengths and the
diraction is second order, and so on.
All x-rays emitted at angles dierent than W cancel because they are out of phase and
destructive interference occurs. From Figure 4, one can note that in the Bragg scattering,
Figure 4 Derivation of the Braggs law. ABC is the path difference.
Copyright 2002 Marcel Dekker, Inc.
the incident and exit angles are equal and, in this sense, there is some analogy to mirror
reection in the classical optics. However, diraction is by no means equal to reection in
classical optics, because the diraction is a volume not a surface process. It is less ecient
(with a great loss of intensity) and is performed only under particular angular conditions;
that is, according to Braggs law. Full analogy between the reection of x-rays and the
reection of optical rays happens only at grazing incident angles (i.e., smaller than the
critical Snells angle). This is intrinsically a consequence of the refraction index, which is
slightly smaller than 1 for x-rays. There are continuous eorts to construct a real mirror
on the basis of multilayers, which would eciently reect x-rays under large incident
angles, at least for soft x-rays (Kearney et al., 1991).
Thus, the static condition for obtaining diraction of a monochromatic x-ray beam
in some direction in the volume surrounding the analyzing crystal is given. What happens
in the case of a polychromatic beam of x-rays? For a crystal, one set of planes is selected
(for dierent reasons) and d is constant. If only rst-order diraction is considered and
constructive interference must be realized for all l present in the incident beam, then W is
the only variable:
l constantsin W 2a
The signal arriving from the diraction angle W is detected by a detector placed on a
goniometer arm. The detector rotates around an axis through the macroscopic plane of the
analyzing crystal. For a source at a xed position, the detector rotates over an angle 2W.
The wavelength is calculated from constant 2d and sin W. Note that this holds only for
the rst-order diraction. If the second-order diraction is used, the wavelength is equal to
half of that value. The maximum wavelength l
max
that can be diracted in the rst-order
diraction by a crystal is equal to 2d because sin W1.
Because wavelength and energy are related, one more equations must be given, al-
lowing us to convert wavelengths into energy units:
E
12398:5
l
3
where the energy E is in electron volts (eV) and l is in angstroms (A
). It is a common
practice that the units used by x-ray spectroscopists are still electron volts or kiloelectron
volts for energy and angstroms for wavelength. When joules and nanometers are used, as
required by the international rules, the numerical value of the conversion constant be-
comes 1.98645610
16
(nm=J).
The presence of dierent wavelengths of dierent order on the same goniometer
position has a particular consequence for wavelength-dispersive spectrometry. For a given
position of the goniometer (and detector), one may have a rst-order wavelength, say
0.6 A
, or values
of energy 20.66, 41.32, or 61.98 keV, respectively. Figures 5 and 6, reproduced from the
work of Arai (1987) demonstrate a practical situation and Table 1 reprints a fragment
from x-ray tables by Cauchois and Senemaud (1978). Similar compendia of spectral lines
and attenuation coecients are easily available (Birks, 1974), as a rule covering the range
of elements between lithium (Z 3) to plutonium (Z 94). Much more controversial are
the values of the mass attenuation coecients for low energies and for low-Z elements at
the same time (Henke et al. 1982, 1988). If necessary, one can have access to the calcu-
lation for even heavier elements (So et al., 1977). It is easy to understand the practical
value of such calculations for the analysis. However, it is much more dicult to estimate
the accuracy of numerical calculations for transuranium elements. To sort out x-rays with
Copyright 2002 Marcel Dekker, Inc.
wavelengths belonging to dierent spectral orders, a pulse-amplitude selector or detector is
needed that can make the distinction between energy levels. A detector is a device of which
the principle can be most easily explained by assuming the particle character of electro-
magnetic radiation. This is one reason that the characteristic of the impacting photons is
often expressed in terms of energy, not wavelength.
Figure 5 Higher-order overlapping of L series x-rays of Cr, Fe, and Ni to Ka of carbon. (From
Arai, 1987.)
Figure 6 Overlapping of CKa by higher-order x-rays of Ni, Fe, and Cr. (From Arai, 1987.)
Copyright 2002 Marcel Dekker, Inc.
The dierences between wavelength-dispersive (WD) and energy-dispersive (ED)
XRF are due not only to the dierent detectors but also to other factors:
1. The brightness of a WD spectrometer is very low, the attenuation in a crystal
being responsible for an important part of losses. This problem may be
overcome by the use of radiation sources of signicant intensity, the synchrotron
being the best known but bound to the availability of this large facility.
2. The crystal is only the dispersive device, not the detecting device. The situation is
dierent in EDXRF, for which detectors play a double role: as the dispersive
device and as the detector at the same time.
Table1 Fragment of Detailed Tables of X-ray Lines
l (uX) l (mA
) Element Transition
Notation
usuelle Ordre E (keV) n=R n=R
1=2
1251.56 1254.19 42 Mo KM
IV
Kb
II
5
2
1251.64(
a
) 1254.27 58 Ce KM
IV
Kb
II
5
4
1252.192 1254.824 62 Sm KL
II
Ka
2
4
1252.58 1255.21 73 Ta abs. L
III
1 9.8776 725.99 26.9441
1252.6 1255.2 90 Th L
I
N
IV
2
1252.82 1255.45 73 Ta L
III
O
IV,V
Lb
5
1 9.8758 725.85 26.9416
1255.17 1257.81 73 Ta L
III
N
VI,VII
Lb
7
1 9.8573 724.49 26.9163
1255.43 1258.07 32 Ge KL
II
Ka
2
1 9.8552 724.34 26.9135
1255.50 1258.14 69 Tm KL
I
5
1256.60 1259.24 75 Re L
I
M
II
Lb
4
1 9.8460 723.66 26.9010
1256.71 1259.35 73 Ta L
III
O
III
1 9.8452 723.60 26.8998
1257.1(
a
) 1259.7 71 Lu L
II
N
I
Lg
4
1 9.842 723.4 26.8956
1257.12 1259.76 90 Th abs. L
II
2
1257.66 1260.30 73 Ta L
III
O
II
1 9.8377 723.05 26.8897
1258.74 1261.39 54 Xe KL
II
Ka
2
3
1259.210 1261.856 68 Er KL
III
Ka
1
5
1259.8 1262.4 90 Th L
II
P
II,III
2
1260.30 1262.95 74 W L
I
M
III
Lb
3
1 9.8171 721.54 26.8615
1260.6 1263.2 90 Th L
II
P
I
2
1260.620(
a
) 1263.269 58 Ce KM
III
Kb
1
4
1261.15(
a
) 1263.80 64 Gd KM
IV,V
Kb
5
5
1261.23 1263.88 73 Ta L
III
O
I
Lb
7
1 9.8099 721.01 26.8516
1261.956 1264.608 42 Mo KM
III
Kb
1
2
1262.68 1265.33 90 Th L
II
O
IV
Lg
6
2
1263.086 1265.741 42 Mo KM
II
Kb
3
2
1263.428(
a
) 1266.083 58 Ce KM
II
Kb
3
4
1264.6 1267.3 74 W L
III
N
III
1 9.784 719.1 26.8158
1264.99(
a
) 1267.65 70 Yb L
II
N
IV
Lg
1
1 9.7807 718.86 26.8116
1265.0 1267.7 91 Pa L
II
N
IV
Lg
1
2
1265.24(
a
) 1267.90 69 Tm L
I
N
III
Lg
3
1 9.7788 718.72 26.8090
1268.520(
s
) 1271.186 92 U L
II
N
I
Lg
5
2
1268.66 1271.33 90 Th L
I
N
III
Lg
3
2
1268.68(
a
) 1271.35 68 Er L
I
O
IV,V
1 9.7523 716.77 26.7726
1270.2 1272.9 74 W L
II
M
V
1 9.741 715.9 26.7566
1270.35(
a
) 1273.02 64 Gd KM
III
Kb
1
5
Source: From Cauchois and Senemaud, 1978.
Copyright 2002 Marcel Dekker, Inc.
3. Because Braggs law is geometric in character, the angular conditions for the
collimation of primary and secondary beams are very severe for WDXRF (8
25 msr), not the same as for EDXRF (150 msr) (Wollman et al., 1997). This
restriction on the geometrical eciency still makes the overall photon collection
eciency of a wavelength-dispersive spectrometer worse.
4. An attractive aspect of EDXRF is the simultaneous collection of the whole
spectrum, whereas a typical WDXRF device is exclusively sequential.
However, the maximum count rate for an EDXRF instrument is 30 kcps
for the whole spectrum, which severely limits the total number of accumulated
counts and, consequently, limits the precision (counting statistics). Similarly,
the limitation on the total count rate leaves a small margin for trace element
analysis. From the assumption, the trace element, if present in a sample, can
emit only a very small part of the total radiation emitted by a whole sample
(and there is also the background, participating in the total amount of
counts).
Simultaneous WD instruments are composed of a series of individual crystal spectrometer
(channels) operating simultaneously, but the number of channels is limited.
A comparison of wavelength- and energy-dispersive versions of x-ray spectrometers
is performed in a recent publication by Brill (1996).
III. LAYOUT OFASPECTROMETER
The main parts of a spectrometer can be represented in the simplest form by Figure 7. The
analyzing crystal is the central point of the wavelength-dispersive instrument. On the left-
hand side of the crystal, we nd (1) the source of excitation, (2) the lters and devices for
shaping the exciting radiation (collimators and masks), and (3) the sample. On the right-
hand side, we nd (4) devices for shaping the diracted beam (collimators) and (5) the
detector. Signals from the detector are fed into the electronic circuitry where they are
shaped to be processed by the computer software for data analysis. This arrangement can
be reduced or made more complicated according to the demand. However, all possible
reductions lead to less exible devices, and complication does not necessarily enhance the
quality of the instrument.
Figure 7 A wavelength-dispersive spectrometer: FCproportional flow counter; SDscintilla-
tion detector. (Reprinted with permission of Siemens AG.)
Copyright 2002 Marcel Dekker, Inc.
A. Sources
A variety of radiation sources, emitting either charged particles or g- or x-rays of sucient
energy, are used for excitation of some or all elements of the periodic table and some or
most of the spectral lines of analytical interest. Other chapters deal with excitation by
protons [proton-induced or sometimes particle-induced x-ray emission (PIXE) (Chapter
12)], by electrons [electron microprobe (Chapter 13)], or by x-rays emitted from secondary
targets or from a synchrotron (Chapter 8). Excitation by x-rays or soft g-rays from
radioisotopes and x-rays from low-power tubes is mainly restricted to energy-dispersive
spectrometers (Chapter 3). The ideal excitation source would be a tunable x-ray laser
(monochromatic and intense, allowing the best choice of exciting wavelength and often
selective excitation), but this cannot be expected in the near future. Nevertheless, it is
worth reading some treatises about recent and future progress in the eld (Nagel, 1982;
Jamelot, 1995; Fill, 1995; Crasemann, 1994; London, 1993). There are serious reasons for
the slow progress in x-ray laser construction. The primary reason is that the population
inversion of electrons means a much larger deviation from the energetical equilibrium if
done between the levels, allowing the emission of x-rays. Consequently, it demands an
excessive pumping power, turning the pumped matter into a plasma. One must be aware
that the x-ray laser would be even much more useful in the elds of x-ray microscopy,
holography, litography, or for the research of time-resolved phenomena than as a source
for XRF analysis on a routine basis.
The synchrotron is another modern source of x-ray radiation, but it cannot be
considered as a tabletop instrument and a device for easy, inexpensive, and routine-based
applications. From a practical point of view, vacuum x-ray tubes are the overwhelming
choice among other potential excitation sources. High-power tubes are the only ones dealt
with in detail in this chapter; low-power tubes are discussed in Chapter 3.
All modern tubes owe their existence to Coolidges hot-cathode x-ray tube as pre-
sented in Physical Review some 85 years ago (Coolidge, 1913). They essentially consist of
a sealed glass tube containing a hot tungsten lament for the production of electrons, a
cooled anode, and a beryllium window. From a variety of modications proposed over
more than three-quarters of the century, two geometries have emerged as the most sui-
table for all practical purposes [the end-window tube (EWT) and the side-window tube
(SWT)], but now the preponderance of the EWT pushes the SWT out of the market.
Perhaps, a future deeper orientation of the WDXRF toward the microprobe applications
will renew the interest in SWT, which is much better in deriving parallel x-ray beams.
The general requirements of the x-ray tube as a source are as follows:
1. Sucient photon ux over a wide spectral range, with increasing emphasis on
the intensity of the long-wavelength tail of the white spectrum. The actual vivid
interest in low-Z element analysis will certainly activate research in this
direction.
2. Good stability of the photon ux (<0.1% at least). Long-term stability reduces
the frequency of recalibration in routine analyses; short-term stability is an
absolute requirement for obtaining an acceptable precision.
3. Switchable tube potential (10100 kV), allowing the creation of the most
eective excitation conditions for each element. Still, in more high-power
constructions available on the market, putting a lter in the beam path is an
easier solution than switching the voltage (Shimadzu XRF-1700). For sure, the
selection of the spectral region for the targeted excitation by the use of a single
x-ray tube will never be as good as in dedicated synchrotron beam lines with
Copyright 2002 Marcel Dekker, Inc.
monochromators of the adequate quality, but the low cost of tube operation is
an obvious advantage. The intensity of the analyte lines varies considerably with
excitation conditions. An extreme example is given by Vrebos and Helsen
(1985a, 1985b) for simulated AlMo alloys.
4. Freedom from too many interfering lines from the characteristic spectrum of the
tube anode.
Freedom from interfering lines is important. The scattered characteristic lines of
the anode may spectroscopically interfere with analyte lines, disturbing the qualitative
recognition, the peak intensity estimation for the line of interest, and the accuracy of
subsequent conversion of intensity to concentration. This is found in the results obtained
after correction by some algorithms as well as by fundamental parameter calculations.
Although there is some kind of a remedy by the use of more ecient spectral decon-
volution methods (Remond et al., 1993; 1996), it occurs at the expense of a greatly
increased computational eort and depends heavily on the assumptions made a priori.
An x-ray tube is characterized by its anode element (a single element or two
elements as in dual-anode tubes), its input power [expressed in watts (W) or kilowatts
(kW), typically between 0.2 and 5 kW for high-power tubes], the voltage range between
anode and cathode (10100 kV for a SWT; limited to 60 kV for an EWT for the ma-
jority of commercial instruments), tube current [milliampere (mA), typically up to
60 mA or sometimes to 150 mA for high-power tubes], and an open (SWT) or closed
(EWT) anode cooling circuit. The photon output of the tube or, more importantly, the
photon ux hitting the sample (expressed in counts per second per watt of input power)
is determined by a, the incidence angle of an electron beam on the anode, the takeo
angle b (for SWT), the distance t to the beryllium window, the thickness d of the
beryllium window, and the distance between window and sample, t
0
(Fig. 8). The ra-
diation output power is rather poor with respect to the input power and is of the order
of 1%, making the device a very inecient transformer of electron current to electro-
magnetic radiation. For an energy balance, see Bertin (1975) and some remarks by
Wollman et al. (1997).
The impact of the electrons creates an excitation volume from which the white and
the characteristic radiation of the target escape. This phenomenon represents the rst
distinct dierence between a SWT and an EWT. As shown in Figure 9a, the escape path
of the photons in SWT geometry is, on average, less than in the EWT geometry. An
immediate consequence of this characteristic is that dual anodes are used only in a SWT,
where a light element (e.g., scandium) covers a heavier element (e.g., molybdenum). By
switching the excitation voltage, x-rays are produced either in the upper (lighter) element
layer or in the substrate (higher-Z element), resulting in two distinct tube spectra with
dierent yields in the low- and high-wavelength regions (Fig. 9b). The popular dual-target
tube arrangements are now Rh=Cr and Rh=W.
The smaller the distance t, the higher the output. A decrease of t to half of the
original value increases the intensity roughly by a factor of 4. However, the reduction of
t in the classical setup is limited. The smaller the value of t, the more intense the bom-
bardment by the electrons and subsequent heating of the window by the scattered elec-
trons (Fig. 8). In a SWT, this bombardment is rather intensive because both the anode and
the window are in this geometry at ground potential. In a EWT, to the contrary, the -
lament and the window are both at ground potential, and heating of the window is neg-
ligible. For an EWT, however, t can be reduced to a certain extent only because it faces the
anode at high potential (Fig. 10).
Copyright 2002 Marcel Dekker, Inc.
1. Alternative Congurations
To overcome the absorption of the low-wavelength tail of the continuum, a windowless
conguration was considered. This is an obvious solution, but it requires that the whole
spectrometer with sample, collimators, crystals, and ow counter be evacuated to the low
pressure suitable for securing an acceptably long life for the tube lament and conducting
analyses in the required range of the soft x-rays.
Nordfors (1956) advocated a dual-anode tube with two anodes physically separated.
They were excited by deviating the electron beam to either of the two anodes (Fig. 11).
Lack of stability prevented this solution from gaining commercial interest. Probably, a
modied version of this idea was introduced in the series 2000 x-ray spectrometers of
Diano (dual target dual lament; see Section VI).
Tubes with exchangeable anodes are another possibility for solving the problem of
comfortable switching between dierent anodes. To the best of our knowledge, this
solution is not commercially available because of the tedious exchange operation and
diculties in rigorous repeating of previous conditions of the tube action.
Using the sample as an anode is another solution, but then the application is
bound to conducting materials. In a modied form, this solution is implemented in
scanning electron microprobes. Because this anode cannot be adequately cooled, the
input power must be restricted to very low values, entailing low count rates and,
consequently, reducing the precision that can be obtained within reasonable counting
times.
Another possibility is to apply an existing 18-kW rotating anode x-ray tube gene-
rator. Sucient intensity is obtained. Part of this gain can be sacriced and Boehme (1987)
and Nichols et al. (1987) introduced the microdiraction beam collimator into the beam
Figure 8 Details of the geometry of a side-window x-ray tube and sample.
Copyright 2002 Marcel Dekker, Inc.
derived from such a high-power tube. This collimates the beam to about 30 mm and allows
microuorescence. Squeezing the beam to more or less the same diameter is possible when
applying the polycapillary lens, even more so with the single capillary. These solutions are
very restrictive for the beam intensity; thus, a reasonable device conguration includes an
energy-dispersive detector instead.
2. Optimization
Because the takeo angle is a very important parameter, the surface of the anode can be
formed in steps in such a way that the yield of low-wavelength radiation is optimized. As is
clear from Figure 12, the step-shaped surface decreases the escape depths.
In a conventional EWT, the reduction of the anode to window distance t is limited
(Fig. 10). According to an old idea of Thordarson (1939), Botden et al. (1952) developed
an x-ray tube with a gold anode evaporated directly on the beryllium window. The tube
was used for surface radiotherapy. After many years, Phillips returned to Botdens
construction and again reduced t to zero by sputtering the anode target element directly
on the beryllium window (Fig. 13). In a SWT, the anode is at ground potential, which
Figure 9 (a) Principle of excitation in a side-window tube. (b) Principle of excitation in a dual-
anode x-ray tube (side window); Mo, W, or Au (for A) and Sc, Rh, or Cr (for B) are commercially
available anode components. The spectral feature of the tube spectrum depends on the energy of the
incident electrons.
Copyright 2002 Marcel Dekker, Inc.
considerably facilitates and simplies the safety requirements. This geometry enhances
the intensity by an order of magnitude, allowing a much lower input power: 200 W for
an intensity equivalent to the earlier 3-kW EWT. As a result of the low input power, the
heat produced in the target by the electron beam is easily dissipated and no supple-
mentary water cooling is required. The tube construction is reduced and only the most
basic components are left. The commercial name for this development is T
3
or TTT
(target transmission tube; Fig. 14c). It is used exclusively in simultaneous WDXRF
instruments, especially in the new version of the PW1660 (Philips Technical Materials
Simultaneous X-Ray Spectrometer System PW 1660) (see Section VI).
Figure10 Principle of end-window x-ray tubes.
Figure 11 X-ray tube with two separate anodes: (1) anodes: (2) filaments; (3) cooling;
(4) separating brass wall; (5) insulator. (After Norfords, 1956.)
Figure12 X-ray tube with stepwise configuration of the anode.
Copyright 2002 Marcel Dekker, Inc.
3. End-WindowTube
The essential parts of an EWT are displayed in Figure 14a. The ring-shaped cathode and
the cooling circuit of the anode are typical features of EWT construction. The latter
consists of a closed circuit lled with deionized water. Deionized water is necessary to
reduce the conductivity and enhance the safety because the circuit is at anode potential!
Electrons are diverted to the anode surface by electron-optics elements. Because the la-
ment and window are at equal potential, no electrons hit the window. Heating is mainly by
radiation from the anode, and an external cooling circuit is provided, which may be
connected to normal water supplies because it is not in contact with a voltage source.
Although the construction just described is in a common use now, it cannot be
considered as an ideal source of x-rays. Recently, van Sprang and Bekkers (1998) and
Kuczumow et al. (2000) have discussed the deviations from the homogeneity in the end-
window x-ray tube output. The geometry of the excitation system, as described by an
angle supported on the tube axis the sample point the goniometer entrance axis,
deviates from the formal geometry (about 83
, which have slightly dierent orientations and lead to widening the diracted
peaks. This happens to widely diering degress: topaz, EDDT, ADP, and gypsum exhibit
low mosaicity, quartz and LiF a little higher mosaicity, and PET a signicantly higher
Table 2 LiF: d Spacings for Different Orientations and Angular
Dispersion
Crystal 2d (A
) Dispersion
LiF (422) 1.652
LiF (420) 1.800
LiF (220) 2.848 0.5154
LiF (200) 4.027 0.2902
Note: Calculated for W angles corresponding to MnKa (2.102 A
) and n 1.
Copyright 2002 Marcel Dekker, Inc.
mosaicity. A result is a slight angular widening of the diracted peak on one hand, but on
the other hand, the intensity of the peak is generally enhanced. Widening is smaller than
widening due to other geometric factors of the whole spectrometer and the loss can be
tolerated while compensated by the increase in intensity. Some mechanical treatments
enhance the mosaicity, and LiF crystals are easily improved in that way. The angular
intensity distribution of a diracted peak, the rocking curve, can be determined by double-
crystal spectrometer. In Figure 17, the peak height of the represented curve obtained by a
high-resolution spectrometer is determined by atomic scatter factors and the space group
of the analyzing crystal, whereas the width at half-maximum is determined by the mo-
saicity of the analyzing crystal and the divergence of the incident and diracted beams.
The surface under the curve is the integral reection or, when the curve is normalized to
the intensity of the incident beam, the integral reection coecient. This coecient is very
much dependent on mosaicity. Abraded LiF has a 10-fold increase in the reection
coecient compared to freshly cleaved LiF (3610
5
to 4610
4
rad), measured by using
CuKa radiation (Birks, 1969). LiF crystals have also the great advantage of low ab-
sorption by their constituting atoms. Topaz and quartz, otherwise good crystals, have
Table 3 Currently Available Crystals
Crystal 2d (nm) Element range Remarks
LiF(420) 0.180 NiU High resolution, special applications for
heavy-element K lines
LiF(220)
a
0.2848 VU High resolution, good intensity
LiF(200)
a
0.4027 KU General purpose, wide range, best diffracted
intensity
Si(111) 0.6271 P, S, Cl Supresses even orders
Ge(111)
a,b
0.6532 P, S, Cl Supresses even orders, good for
intermediate- and low-Z elements
PG(002)
a
0.6708 P, S, Cl Good intensity, poor resolution
InSb(111)
a,b
0.784 Si
PET(002)
b
0.8742 AlCl Good intensity, low spectral contamination,
soft
EDDT(020) 0.8808 AlCl Lower intensity than PET, problems with
stability
ADP(101) 1.064 Mg Still lower intensity
TIAP(100) 2.575 OMg Especially F, Na, Mg, poisonous
PbSt 10.0 F, C Good resolution, low intensity
LTC 15.6 Up to Be Longest wavelengths known for
semicrystals structures
PX-1
a
4.93 OMg Low resolution, good intensity
PX-3
a
19.5 B Low resolution, good intensity
PX-4
a
12.2 C Low resolution, good intensity
OVO 55 5.5 Mg, Na, F Low resolution, good intensity
OVO 100 10.0 C, O Low resolution, good intensity
OVO 160 16.0 B, C Low resolution, good intensity
OVO H300 30.0 Be, B One of the largest 2d available up to now
PX6
a
Also available as curved in simultaneous instruments.
b
Also available as transversely curved for sequential instruments.
Copyright 2002 Marcel Dekker, Inc.
poor reectivity properties, whereas PET is very good in this respect. Reection constants
as a function of wavelength are given in Figure 18.
c. Stability and Temperature
The mechanical stability of most crystals is satisfactory, but there are exceptions. Gypsum
can eoresce (especially in a high vacuum); PET has a tendency to change phases on aging
Figure 17 Rocking curve that would be obtained from a real crystal-diffracting parallel
monochromatic radiation (rocking curves are actually measured on double-crystal spectrometers):
Pthe peak diffraction coefficient; Rthe integral reflection coefficient. (From Birks, 1969.)
Figure18 Single-crystal integral reflection coefficients of graphite, LiF (200), LiF (220), and KAP.
[From Gilfrich et al., 1971). Reprinted by permission of Analytical Chemistry. Copyright #
American Chemical Society.]
Copyright 2002 Marcel Dekker, Inc.
and it is soft so that it is easily damaged when manipulated. The temperature inside the
measuring chamber of a spectrometer is kept as constant as necessary (<1 or <0.01
C if
chemical peak shifts have to be measured). In these circumstances, damage of the organic
crystals by high temperature is not likely to occur. When very precise measurements of
peak position are to be made for speciation purposes, the thermal-expansion coecient is
important. PET is very disadvantageous in this respect; topaz is in the opposite situation.
It should be noted that the characteristic spectral lines of the crystal can be super-
imposed on the spectrum, making it obscure or even unt to use, especially when it
suggests the presence of some elements in a false way. Crystals of excellent resolving
features, Be and Ge (111), have their own lines in the soft x-ray region.
d. Spectral Range
The useful spectral range covered by one crystal is limited. The maximum wavelength is of
trigonometric origin and is imposed by Braggs law, namely by the condition that sin W < 1
and l
max
< 2d. In practice, the goniometer is scanned for values of 2W varying from a few
degrees to about 150
) or illitemontmorillonite (2530 A
) type, are
not suciently ordered in this respect (Weaver and Pollard, 1973). Unfortunately, natural
crystals, irrespective of their dierent crystallographic planes, oer only limited possibi-
lities in this spectral range; specically, their reectivity is very low. The class of structures
called LangmuirBlodgett lms was more promising, but the reectivity and thermal,
mechanical, and chemical stabilities were rather poor.
Figure 21 Cross section of a typical multicapillary lens: (a) focusing configuration, from wide
spot (divergent source) to another spot (semifocus); (b) configuration transforming the divergent
into the parallel beam. (Courtesy of X-Ray Optical Systems Inc., Albany, NY. and personally from
Dr. J. P. Bly.)
Copyright 2002 Marcel Dekker, Inc.
A correct solution resulted from generalization of Braggs law, namely that dif-
fraction also occurs in media consisting of layers of dierent refraction coecientsin
other words, in sites of dierent periodically changing electron densities. Such structures
may be formed in an articial way. They may be treated as crystallike substances ordered
along the c axis (i.e., spacing d or 2d), but with uncontrolled arrangement within the
ab plane. The structure in the ab plane can be amorphous, provided that the optical and
mechanical properties are uniform in these directions.
The rst eort of synthesis was made in 1940, but the AuCu multilayer of Du
Mond and Youtz (1940) did not survive more than a few days because of deterioration by
interdiusion. The problems of thermal, radiative, diusional, and chemical instabilities
of multilayers have recently been solved. Signicant progress in this eld has taken place
in the last three decades, hence some years before the application of multilayers in
WDXRF.
The second approach was the result of an increasing interest in x-ray optics. As a
branch of physics, x-ray optics is nearly as old as the discovery by Ro ntgen, but because of
the specic properties of x-rays, its progress has been slow. All existing substances exhibit
complex refractive indices:
m 1 d ib 9
where m is the refraction index, d is the so-called unit decreament, a real number, and the
imaginary term b is related to absorption. The refraction indices in the x-ray region are all
below but very close to 1. This makes all eorts of x-ray focusing by lenses in a traditional
sense an impractical task. Franks (1977) gives an example of a lens with a curvature radius
of 10 mm built of a material with d 5 10
5
. The focal length of this lens is 100 m. In
such a situation, only glancing, grazing, and diracting features of materials have been of
practical interest for a long time. Otherwise, the grazing optics conditions for x-rays are
extremely severe. However, the need for mirrors, monochromators, and focusing devices
for the x-ray region is increasing rapidly in the elds of synchrotron radiation, plasma
research and diagnostics, x-ray microscopy, x-ray uorescence analysis, diraction, thermo-
nuclear fusion, x-ray laser, x-ray astronomy, and even x-ray waveguides. The invention of
multilayers dened this time as a complex set of particular layers consisting of materials of
periodically dierent refractive indices gave great impetus to progress in this branch of
optics. Some important properties of the materials used to produce the multilayers are
discussed in the next section.
a. Nature of Materials
In principle, three kinds of material are necessary for the synthesis of multilayers. First,
there must be a support (sometimes called a substrate), such as a piece of at or curved
silica layer, with an ideally polished surface. This condition is very important because the
roughness of the surface is translated to the deposited layers, correctable almost ex-
clusively when one of the layers consists of amorphous materials such as carbon. After
years of progress in the eld, when the roughness of the interfaces have probably reached
the limit of about 2 A
, adviced
for the determination of boron. However, when a new multilayer should be synthesized
Figure 22 Mass attenuation coefficient for the BKa line drawn against the possible absorbers in
the periodic table. Arrows show the deep minima in coefficient value. (Courtesy of Bruno Vrebos
inspiration.)
Copyright 2002 Marcel Dekker, Inc.
with high reectivity in mind, then the attention is put, instead, on a pair of materials with
a small value of the scattering DebyeWallner factor [see Eq. (10)]. Here, the W=B
4
C
tandem is a very good one. As a rule, the low-electron-density compounds are selected
from the elements with K absorption edges above the energy of spectral lines to be
measured. In that sense, boron and carbon are selected for multilayers of spacing above
50 A
.
b. Smoothness
This parameter is of great importance on each level, starting from the supporting
substrate to the last upper layer. All rough places, uneven layers, disturbances in in-
terfaces, interdiusion, and chemical interaction regions or defects, act in a manner
analogous to the thermal motions in classical diraction analysis. During synthesis, the
extent of roughness can vary in a dierent manner: In some cases, it tends to
smoothness, in others to further growth. The use of amorphous materials to form some
layers favors smoothing. The greater interplanar distances are, the more pronounced
the roughness is. On the other hand, the smoothness probably has its limits, even in
the case of multilayers with relatively small interplanar distances, with the Debye
Wallner parameter s tending to reach a value of about 2 A
oxygen to magnesium
PX3 2d 195 A
carbon
PX5 2d 112 A
nitrogen
PX6 2d 300 A
beryllium
and from the Ovonic Synthetic Materials Company (1987):
OV-040A 2d 40 A
fluorine to silicon
with multilayers changing 2d spacing every 20 A
up to
OV-140B 2d 140 A
carbon to oxygen
and also the MoB
4
C structure:
OV-H series 2d 244 A
beryllium to boron
up to
OV-300H 2d 300 A
Probably, materials with even greater interplanar distances can also be supplied. However,
recent scientic (except the XRF eld) interest in multilayers is centered, instead, more on
multilayers with periods of 1020 A
C).
The relative spectral resolution of dispersive devices is often expressed in the form of
the important parameter DE=E or Dl=l, which is the basis for their choice in XRF ap-
plications. Unfortunately, this is the weakest point of the multilayer. In a commercially
available specimen, the relative spectral resolution changes from about 0.025 to over 0.1
while passing from structures with 2d 40 A
, respectively, for
dierent measured lines (Ovonic Synthetic Materials Co., 1987), and for the structure with
2d 75 A
(Henke et al.,
1988), which is about three times worse (Ovonic Synthetic Materials Co., 1987) or even
more (Henke et al., 1986) than for a good classical pseudocrystal. The comparison of
spectra of the same sample by using multilayers with dierent spacings, as in Figure 23, is
very instructive (Van Eerbergen and Volbert, 1987). Relatively broad internal bandpasses
of multilayers diminish the resolution of the whole spectrographic device, seriously in-
creasing the detection limits. This is of special importance to the analysis of the L or M
series, and avoids the eorts to search the chemical shifts in x-ray spectra [although see
Habulibaz et al. (1996), where chemical shift is investigated for the SiL spectrum, dif-
fracted on a multilayer with an interplanar distance 300 A
, respectively, and should be analyzed in the region otherwise reserved for the
Figure 25 2W scans for Mg in cement using TlAP and PX1. (From Nicolosi et al., 1986.)
Copyright 2002 Marcel Dekker, Inc.
K series of oxygen and uorine, hence by the use of the PX-1, PX-4 or OV-120B pseudo-
crystals. The question is whether the limited spectral resolution of multilayers is sucient
to show such subtle eects as chemical shifts and to separate the mentioned lines of metals
from the disturbing lines of oxygen or uorine if present in the sample.
The wavelengths of characteristic x-rays in the K series are longer than atomic radii
starting from Ga (GaKa 1.337 A
=min. The slewing speed is the maximum rotation speed of the goni-
ometer arm for passing from one position to another without recording. The 2W range
covered is from 4
to 152
inclination
dierence. The diracted beams are detected by two detectors 30
120
and 10
100
The electron imparts its kinetic energy to other atoms, creating a series of electroncation
pairs on its path through the gas in a number equal to E
kin
divided by the energy needed to
expel an electron. Values for dierent gaseous and solid detector media are included in
Table 4 Effective Ionization Potentials (eV)
He 27.8
Ne 27.4
Ar 26.4
Kr 22.8
Xe 20.8
NaI (Tl) 50
Si (Li) 3.8
Ge 3
Copyright 2002 Marcel Dekker, Inc.
Table 4. Subsequently, the cations and electrons are accelerated toward the cathode and
anode of the detector by the gradient of the electrical eld between both electrodes,
forming the avalanches of new ion pairs on their way to the electrodes. The discharge at
the electrodes gives rise to a current pulse in the external electrical circuit. For a range of
electrode potentials, the current pulse is proportional to the energy of the photon, and in
spectrometers, the gas-lled detectors are exclusively used in this range. This is why these
detectors are called proportional counters. Above this potential, in the GeigerMu ller
plateau (Fig. 28), any photon entering the detector gives rise to a pulse of maximum
amplitude without relation to the initial photon energy. The ratio of the number of ion
pairs discharged at the electrodes to the ion pairs originally formed is called the ampli-
cation factor, generally denoted A. It is a complex exponential function of ionization
potential of the gas, anode potential, radii of detector chamber and anode wire, and the
mean free path of the electron. It implies that stability of applied potential and gas
pressure is very important for reducing statistical errors on pulse amplitude.
The basic parts of a gas-lled detector are a metal rectangular housing as cathode, a
thin wire passing through the central axis as an anode isolated from the cathode, a lling
gas, and a window either on one side or on two opposite sides of the housing. This gas
ows through or is enclosed in the housing (Fig. 29). The diameter of the anode wire
should be small. Realistic diameters are in the range of 4080 mm.
a. Flow-Proportional Counters
These detectors cover the widest wavelength range and are used in sequential spec-
trometers exclusively for long wavelength ( >2 A
around 2W 37
or >6 keV). The outer orbital electron for an iodide ion requires about 30 eV to be
knocked out. Hence, the originally ejected electron is imparted with almost all the
initial energy of the photon. The ejected electron dissipates its energy by promoting
valence-band electrons to an excited state 3 eV above ground level, an energy emitted
on deexcitation as a photon of 3 eV or l 410 nm. The intensity of the emitted light
pulse is proportional to the number of electrons excited by the x-ray photon
( hv=3 eV). The pulses are detected by a photomultiplier in which the light pulse
produces a few photoelectrons from the cathode material (e.g., indium antimonide).
These electrons are accelerated inside the vacuum tube between the cathode and the
rst intermediary anode, called a dynode (Fig. 31). They gain kinetic energy and
generate, in turn, a higher number of electrons from the dynode. This process is most
often repeated 10 times, resulting in a substantial multiplication of the original number
of electrons. The multiplication depends on the potential dierences between the suc-
cessive dynodes (100150 V), and the order of magnitude of this multiplication is 10
6
,
which is called the amplication factor A. In general, the photomultiplier is a front-end
Copyright 2002 Marcel Dekker, Inc.
tube, and the iodide crystal is in close contact with the tube window. The crystal itself
has on its outer faces a reective coating (but transparent for the photons above
6 keV) and a coating against moisture. In the external electronic circuit, the electrons
provoke a current pulse of which the amplitude is ultimately proportional to the energy
of the photon captured in the crystal. The conversion of the x-ray photons into a
current pulse occurs over dierent steps: the initial formation of a photoelectron
(I
!I
0
e
2W)
Pm La
2
10 0.05 Ho Ll <10 0.04
La Lb
2
30 0.31 Pr Lb
7
20 0.24
Ba Lg
5
40 0.60 Mn Ka 50 0.61
La Lb
7
30 0.23 Pm Lb
1
10 0.17
Ce Lb
6
20 0.31 Ba Lg
4
30 0.30
Pm La
1
20 0.31
V Kb 10 0.56
Copyright 2002 Marcel Dekker, Inc.
spectral resolution of the system is essential because only in this case are peaks very
narrow and do not overlap, and the background level is easily reached in the intervals
between peaks.
Finally, we refer to a method based on the relationship of uorescent and scattered
radiation intensities versus the value of attenuation coecients. No measurements of back-
ground for the estimation of net line intensities is required in this case (Bougault et al., 1977).
If we cannot overcome the background problem, it would be advantageous to use the
background or scattered tube lines for increasing the information about the sample. There
are very interesting examples of such work and they will be shown in Sec. IV.C.
Figure 34 Spectra of Pd excited by a Rh tube and recorded with and without a primary beam
molybdenum filter (0.1 mm Mo). The peaks are labeled PdKa, PdKb, RhKa, and RhKb with their
respective order indices. (Courtesy of ARL.)
Copyright 2002 Marcel Dekker, Inc.
B. Qualitative Spectrometry
Although the number of spectral lines is very limited compared to atomic emission
spectra in the ultraviolet or visible region, overlapping of lines in x-ray spectra does
exist. A classic example can be found in the works of Gentry et al. (1976), Sparks et al.
(1978), and Sparks (1980), and an example is reproduced in Figure 36. At energies at
which K lines of elements between palladium and cesium (about 22 and up to 30 keV)
can be present, additionally the tails of L lines of uranium and thorium are found, and
also the theoretically calculated M lines of the superheavy elements of Z110 [in the
time which passed since 1976 this element has been discovered (Hofmann et al., 1994)]
and 126 should be expected. The erroneous attribution of some lines to these hy-
pothetical elements was the subject of extended research and long discussions (Sparks
et al., 1978; Sparks, 1980). In the software of modern spectrometers, special searching
blocks exist for the assignment of lines: The detected line is compared with data com-
piled in the program. After the initial assignment, other lines of the element are iden-
tied according to data from the library. The initial assignment is eventually rejected if
not conrmed by other lines (Fig. 37). Multichannel wavelength-dispersive instruments
are not suited for this type of work, except for elements implemented on the spectro-
meter (maximum 28 elements). Sequential instruments can cover all elements (from
beryllium on). The spectra can be displayed on the monitor screen of all modern
spectrometers, documented, memorized as les, and printed. Particular parts of the
spectrum can be scanned occasionally with higher resolution and sensitivity (step
scanning) according to the time available for the analysis and the information required.
Parts of particular interest can be displayed separately with a linear intensity axis or
rescaled to square root or logarithmic scales. The peaks can be identied on the display
by the angle 2W, by element name, either by wavelength or energy.
Figure35 Synchrotron-excited spectra of gaseous nitrogen with a small amount of xenon taken in
the direction of polarization and perpendicular to the orbit plane. Counts normalized to stored
electron current are integrated over the lifetime of the detection system. DORIS was operated at
3.3 GeV and about 55 mA. (From Kno chel et al., 1983. Reprinted by permission of the author and
Elsevier Science Publishers.)
Copyright 2002 Marcel Dekker, Inc.
Figure 36 Fluorescence spectrum from a monazite particle excited with 37-keV synchrotron
radiation (ORNL DATA) shows an improved signal-to-background ratio over that excited with
5.7-MeV protons (FSU DATA). Data points every 20 eV in ORNL spectrum and every 61.8 eV in
FSU spectrum. (From Sparks, 1980.)
Figure 37 Example of qualitative analysis by rejection. The suspected MoLa peak in the lower
spectrum was rejected after comparison with the standard containing molybdenum (upper
spectrum). (Courtesy of Phillips Analytical.)
Copyright 2002 Marcel Dekker, Inc.
The spectrum can be scanned linearly with respect to 2W or in a more optimal way. If
Braggs and Moseleys laws are combined, the following relationship among Z, the atomic
number, and W is found:
Z
m
k2d sin W
_
s 14
Z is proportional to the inverse square root of sin W. Jenkins et al. (1979) proposed
computer monitoring of the scan speed according to this equation. It optimizes the scan
speed, which represents a high gain in time especially in the low-Z range where the angular
distances between the lines are long.
The newest applications of WDXRF softwares try to avoid the old-fashioned treat-
ment of data. One such traditional manual estimation was that concerning the possible
spectral interferences. Normally, the analyst has to decide which place in a spectrum is
relatively empty of disturbing peaks and compromising background contributions. This job
was rather time-consuming, with the wide use of spectral tables and other datasets con-
cerning, for example, intensity ratios. The results were not always encouraging. Now,
virtual synthesis of wavelength-dispersive spectra is in progress (Reed and Buckley, 1996),
in the Internet demonstration version also (Buckley, 1998). The expected pure-element
spectra are loaded from the memory and installed in a region of interest of the anticipated
spectrum. Those partial spectra are apparently corrected for the expected concentrations of
the sample constituents. From the result shown on a screen of the computer, one can easily
estimate whether the region of interest is suitable for a given analysis or should be excluded.
Such a procedure makes the selection of lines for the qualitative analysis much easier, as
well as the estimation of the background and also the assessment of the detection limits.
C. Special Quantitative Applications
Although the main quantitative treatment of XRF has been left to Chapter 5, still there are
some special and nonconventional applications which proved to have potential practical
meaning. Their signicance becomes obvious when, considering the excitation process, one
wonders how great a part of the radiation is transformed into the scattered form and how it
can supplement the essential quantication based on characteristic signals.
As stated earlier, monochromatic excitation simplies the quantitative treatment of
spectral data considerably, but this case is only seldom met in WDXRF. The most
common excitation source is the x-ray tube, in which a set of characteristic lines super-
imposed on a white spectrum is generated. For this reason, only the fundamental coe-
cient method can be considered as rigorous and somewhat simplied in comparison with a
complete fundamental parameter method. How the correction terms can be derived in a
rigorous manner for the case of monochromatic excitation was shown in an article by
Kuczumow (1982). In its original formulation, the LachanceTraill equation is (Lachance
and Traill, 1966)
W
i
R
i
1
j6i
a
ij
W
j
_ _
15
with a
ij
dened as a constant written as
a
ij
m
jC
1
E m
jC
2
E
i
m
iC
1
E m
iC
2
E
i
m
iC
1
E m
iC
2
E
i
15a
Copyright 2002 Marcel Dekker, Inc.
where m denotes the mass absorption coecients for primary or secondary photons of
energy E or E
i
, respectively, corrected for the incident C
1
and takeo C
2
angles. It has
been proved (Claisse and Quintin, 1967) that this expression is valid only for boundary
conditions: monochromatic excitation and absence of enhancement. In all other condi-
tions, as are not constant, as experimentally shown (Tertian and Claisse, 1982; Lachance,
1981). It is impossible to derive a more exible algorithm with constant coecients directly
from Shiraiwa and Fujinos equation (Kuczumow, 1982; Rousseau and Claisse, 1974).
Other algorithms similar to that of Tertian (Rousseau, 1974) and Lachance COLA
(Lachance, 1981) are not recognized as fundamental. An algorithm that can be considered
partially fundamental is that of Claisse and Quintin (1967), but it suers from inaccuracies
in the approximations introduced. Later, Rousseau (1984, 1987) and Kuczumow and
Holland (1989) gave examples of algorithms derived from fundamental assumptions.
Kuczumow and Holland describe the change in a when enhancement is included
(monochromatic excitation) as follows:
W
i
R
i
1
j6i;k
a
ij
W
ij
Da
ik
W
k
_ _
16
where all as are dened as previously but Da is dened as
Da
ik
m
0
kC
2
E
i
m
kC
2
E
i
m
iC
1
E m
iC
2
E
i
16a
The new term m
0
kC
2
E
i
was introduced in Eq. (16). Although dependent on the excitation
condition E and composition, this new term has the great advantage of being dependent
on the latter in a simple way. The possibility of deriving the LachanceTraill equation
directly from the Shiraiwa and Fujino equation under the conditions applicable to Eq. (16)
is also well established (Kuczumow and Holland, 1989).
The next interesting case concerns scattered radiation as a source of analytical al-
gorithms. Similar to the derivation made by Kuczumow (1982), an expression on the basis
of coherently scattered radiation is obtained (Kuczumow, 1988):
W
i
R
coh
i
j6i
R
coh
i
a
coh
ij
b
coh
ij
_ _
W
j
17
where
a
coh
ij
m
jC
1
E
0
m
jC
2
E
0
m
iC
1
E
0
m
iC
2
E
0
m
iC
1
E
0
m
iC
2
E
0
17a
b
coh
ij
s
coh
j
E
0
s
coh
i
E
0
17b
and R
i
coh
results from the comparison of a coherently scattered line from the sample and
from the pure element i and s
coh
is the mass coherent scatter coecient (for the set of
incident and takeo angles, as always in the conditions of the analysis). For Compton-
scattered radiation, analogous expressions are obtained:
W
i
R
com
i
j6i
R
com
i
a
com
ij
b
com
ij
_ _
W
j
18
Copyright 2002 Marcel Dekker, Inc.
with
a
com
ij
m
jC
1
E
0
m
jC
2
E
com
m
iC
1
E
0
m
iC
2
E
com
m
iC
1
E
0
m
iC
2
E
com
18a
b
com
ij
s
com
j
E
0
s
com
i
E
0
18b
R
i
com
is the ratio of the Compton peak intensity of the sample to that of the pure ele-
ment. Equations (17) and (18) may be called LachanceTraill equations for scattered
radiation by their general appearance as well as by the way they have been derived.
These equations are less sensitive to the type of components and changes in composi-
tions than their analogs for uorescent radiation, which otherwise is an advantage in
analyses of samples widely varying in composition. From assumption, they include only
the terms for strictly dened energy values, even if the analysis is carried out with a
polychromatic source! In Table 6, the coecients a and b from Eqs. (15)(18) are col-
lected, calculated for the K lines of La and Sm, the main components of samples con-
sisting of La
2
O
3
Sm
2
O
3
excited by 60-keV photons at the incidence and emergence
angles equal to 80
and 70
(Kb
1
) and
0.615 A
(Lg)! For everyday work, however, only two lines, 1.936 and 1.940 A
(Ka
1,2
)
and 1.757 A
(Kb), play a role. The densely populated L series spectrum such as that of
gold has a set of about 20 lines between 0.9 and 1.5 A
C
0
i
C
i
2
n p
20
where C
i
is the true weight fraction and C
0
i
the result obtained; p is the number of para-
meters used in the t: p 2 for a linear and p 3 for a parabolic t. If the concentration C
a
of an element in one specimen diers considerably from C
0
a
, the presence of an outlier is
suspected. The suspected value is eliminated, and a new standard deviation is calculated.
C
a
is an outlier with a probability of P% according to the inequality (F-test) (S
1
=S
2
)
2
>F
p
.
The threshold values for F
p
are listed in the current statistical literature. Application of the
F-test in fact checks whether a statistically signicant dierence exists between the stan-
dard deviation calculated with and without the suspected result. Possible causes of outliers
are numerous, including all kinds of human mistake, temporary instrument failure, and
heterogeneity of specimens. Recently, the numerical programs helping to segregate the
data on the true ones and outliers are available. They are based on the Grubbs test
(Grubbs, 1969; Rousseeuw and Leroy, 1987).
A measurement of uorescent radiation is, as in any other experimental determi-
nation, subject to minimal error. The standard deviation of a radiation measurement is
equal to the square root of the number of accumulated counts N. The real standard de-
viation is greater than
N
p
because of errors of instrumental origin. Another source of
error is the conversion of intensity into concentration and, last but not least, inherent to
the sample, microheterogeneity, which is dicult to foresee and to quantify and is present
even when all normal precautions for careful sample preparation are taken.
Heterogeneities caused by sample preparation are dealt with elsewhere. These arti-
facts should be absent here. Segregation, however, which is the origin of what we call
microheterogeneity, is a bulk (or sometimes surface) phenomenon of crystallographic
origin: solid solutions of compounds or elements that are not perfectly soluble in each
other, segregating into multiphase components. The degree of heterogeneity in the
composition of the segregated phase as well as in the size of the phases depends on the
history of the solution (thermal, chemical, or mechanical treatment). It is not easily
mastered and is dicult to quantify. Its importance for XRF analysis has been ac-
knowledged by several authors; Examples are tin and lead in solders (Glade and Post,
1970), silicates in fused beads (Novosel-Radovic et al., 1984), and silicon in SiAl alloys
(Michaelis and Kilday, 1962). The rst report we are aware of is an article by Claisse
(1957) on the determination of FeS in a sulfur matrix. A systematic study has been devoted
to this phenomenon by Helsen and Vrebos (1984a, 1984b, 1986a, 1986b) and Vrebos and
Helsen (1983, 1985a, 1985b, 1986), both by Monte Carlo simulation on hypothetical
mixtures and by measurements on samples of known segregation. It was found by si-
mulation on a (hypothetical) dispersion of spherical particles of iron in a chromium matrix
Copyright 2002 Marcel Dekker, Inc.
that even for a particle diameter of 1 mm, an error is introduced in the relative intensity of
FeKa of 6% (on the level of c
Fe
0.5) with respect to the perfectly homogeneous solution.
The eect was proved experimentally on alloys of AlSi, AlMo, and AlZn. The extent of
the eect changes with time, as in fused beads. This was demonstrated by the determi-
nations of Novosel-Radovic et al. (1984) on uxed silicates and also by Monte Carlo si-
mulations by Vrebos and Helsen (1985a, 1985b). Microsegregation imposes a true physical
limit to the accuracy obtainable by XRF spectrometry on quite a substantial number of
samples, including minerals and alloys.
3. High-ZLimit
We already insisted on the low-Z limit in Secs. III.D.1 and III.D.2, but what about the
high-energy end of the spectrum?
Commercially available generators and side-window tubes allow working at high
voltages up to 100 kV, bringing even the K line of bismuth into the potential analytical
spectrometric range. If a tungsten target is used, however, characteristic lines represent
only a minority of the integrated emitted intensity, thus, excitation is bound mostly to
continuous radiation, the maximum of which lies at 50 keV. Thus, only elements with K
absorption edges below this value can be eciently excited (Eu, Z63, and all elements
below). The above remark is even more valid for the spectrometers with end-window
tubes, most commonly used in present-day instruments, of which the high voltage is
limited to 60 kV. Possible excitation of K lines for all the elements from the periodic table
by the tubes operating in voltages up to 160 kV is bound to metalceramic low-power
tubes, cooperating with EDXRF systems and is outside the scope of this chapter. Thus,
the available voltage limitation is the rst one.
If we return to Figure 26b and compare the spectral resolution of the LiF spectro-
meter to spectral distances between analogous lines in the K series, it is obvious that spectral
resolution is worse above 35 keV (for K lines of Pr, Z59). This is a second limitation.
Most of goniometers have their optimal 2W range between 10
15
and 70
: At the
higher 2W end, the angular dispersion becomes insucient; at the lower 2W end, the in-
tensity becomes very low. If we choose LiF (420) with the shortest interplanar distance,
2d 1.802 A
corresponding to photons of
39.6 keV (Sm, Z62 in the rst-order spectrum). This is a third limitation.
Studying Figure 26 further, we note that the spectral resolution of the LiF spec-
trometer is below the spectral resolution of the semiconductor SiLi detector for photons
of energy exceeding 17 keV, which corresponds to MoKa, Z42. This is the fourth and
most serious limitation in the K series. For analysis in the L series, there is no such
limitation because the whole energy range lies within the capabilities of wavelength-dis-
persive devices. However, using the new microcalorimetric detectors with extraordinary
and relatively constant spectral resolution, one can anticipate the preponderance of the
spectral resolution factor for wavelength-dispersive crystals only in a region below
45 keV, which means the Ka line of V, Z23 (see Fig. 27). Then, the ability to make the
ecient trace analysis (because of higher count rates) and the availability of the device are
still the advantageous sides of application of the crystal dispersion in the future.
The wide voltage range of generators should not be used without some awareness of
inherent drawbacks. For example, the K absorption edge of Mo is 20 keV. Thus, each
x-ray tube with characteristic lines above 20 keV and working above twice this 20 kV for
the maximum of bremsstrahlung generation is quite good for the excitation of Mo. The
potential, however, should not exceed this value substantially much because the maximum
Copyright 2002 Marcel Dekker, Inc.
of the white radiation would shift more and more away from the molybdenum absorption
edge and even the spectral characteristics do not improve. If we divide the mass coecient
of photoelectron excitation by the sum of mass coecients for Compton and coherent
scatters (this is a kind of measure of peak-to-background ratio), we obtain the values 76
and 18 for tube-excitation potentials of 40 and 100 kV, respectively! This is a substantial
loss in peak-to-background ratio while passing to higher voltages.
The discussion in this section shows that the advantage of high-energy photons
(generated by very high voltages in SWT) does not compensate for the drawbacks and this
probably explains why most manufacturers t EWT tubes to the instruments. During the
last years, manufacturers rather resigned from high-voltage generators and tubes, even in
high-power 4-kW constructions. The use of a higher current is now a preferred option.
Thus, we expect that only spectrometers with a high voltage up to 60 kV will dominate the
market in the future. The eective spectral range above 3040 keV should probably be left
for typical energy-dispersive detectors.
In the preceding paragraphs, we tried to show that although the spectrometers are
almost perfect, there are limits to precision and accuracy due to the nature of the sample
and to the mathematical conversion of intensity to concentration. These uncertainties are
generally greater than those introduced by instrumental parameters.
V. CHEMICAL SHIFTANDSPECIATION
The photoelectric excitation of atoms and subsequent specic deexcitation by emission of
x-rays is the basis of all qualitative and quantitative uses of XRF. The deexcitation process
was considered to take place between unique excited and ground quantum states. If these
states were always stable, x-ray emission lines were to be found at strictly determined
wavelengths or energies. However, electron density and quantum states are inuenced by
the environment of the atom with consequent wavelength shifts, or so-called chemical
shifts, of the emitted lines. Such techniques as nuclear magnetic resonance or Mo ssbauer
spectroscopy collect their spectroscopic information essentially from measurement of
chemical shifts. As a matter of fact, all kinds of spectral lines of electromagnetic radiation
are subject to chemical shifts. All these shifts are small and detectable only if the spectral
linewidths are suciently small and the instrument has the required spectral resolution.
For x-rays, the line shift is at most a few electron volts. An angular error DW 0.02
on the
determination of a peak position at l 1 A
C gives rise to peak shifts greater than the chemical shifts. Helsen and
Wijnhoven (1972) solved this problem by inserting standards at regular intervals and by
using statistical techniques for the determination of the peak position. Sumbaev and Co-
workers (1968, 1970, 1976) made experiments with an instrument in a Cauchois ar-
rangement. In an article by Kataria et al. (1986), it was proved that even using a common
solid-state SiLi detector with moderate energy resolution, it was possible to detect che-
mical eects. Similarly, Habulibaz et al. (1996) performed experiments with a multilayer
with an exact 2d spacing of 300 A
2y=min)
0.25128 0.110 0.006120 0.1240 1200 0.01180
Step scanning (
2y,
2y,
y)
0.0002 ? 0.0001 0.0005 0.001 0.0003
Temperature stabilization
(
C)
0.1 None <0.05 ? 0.01 0.3
Additional fixed channels
(yes, no, or number)
y None 2 None None None
Parts included in vaccum All All All All except sc All except sc All
Counting electronics
Count rate (kcps): fpc 2000 ? 2000 2000 2000 2000
sc 1500 ? 1500 1000 2000 1000
Sealed 1500 4000
Duplex 3500
Pulse-shift correction Automatic Graphical display Automatic,
dynamic
Pulse-height
discrim.
Automatic Automatic
Dead-time correction Automatic Graphical display Automatic Automatic Automatic Automatic
Computer instrument
monitoring peripherals
Full Partly Full Full Full Full
Remote servicing (yes, no) y n y ? y n
Sample changer: type (x, y),
positions
xy, 1298 x, 4 or 10 xy, 6, 24168 x(?), 50, 100 xy, 8, 58, 100,
110, 138
?
Copyright 2002 Marcel Dekker, Inc.
Table 9 Continued
Manufacturer
1 2 3, 3
0
4, 4
0
5 6, 6
0
Installation
Instrument dimensions:
H
*
W
*
D (mm)
920
*
1080
*
760 1400
*
860
*
1346 1092
*
1410
*
920 1400
*
1420
*
1054 1070
*
1890
*
880 1350
*
1770
*
1080
Weight (kg) 450 590 580 750 700 780
Electricity supply (V, kVA) 220, 6 208480, 10 200240, 8, 5 200220, 10 115400, 8 200220, 8 or 9
Cooling water: bar, max.
temp. (
, 0.15
, 0.46
, 0.93
, 1.0
, 1.90
, and 1.54
.
Crystals: All sequential spectrometers make use of at crystals.
Counters: A gas-ow-proportional counter (fpc) and a scintillation counter (sc) are
standard on all instruments, except Diano, in which the sealed Xe detector is
standard and a scintillation counter is an option. The gas ow (Ar 10%
CH
4
) is 12 L=min.
The electronic counters are all based on a multichannel analyzer (Shimadzu and
Rigaku probably still use a pulse-height discriminator) with an automatic pulse
shift and dead-time correction.
Available on all spectrometers is the choice of vacuum or He as the spectrometer
atmosphere, with nitrogen as option for many.
All companies supply the control unit by personal computer in more or less
sophisticated versions with software for monitoring the instrument, for
qualitative analysis, and for a number of possibilities for conversion of
intensity to concentration. All of these items are subject to change and not
discussed in detail here. The conversion algorithms are discussed in detail in
Chapter 5. The software systems supplied by Bruker, ARL, Rigaku, and
Shimadzu are all very user friendly.
Installation data are added to give the potential buyer an idea of the required space
and infrastructure. The dimensions and weight of the instruments are given
without an external sample changed t.
b. Remarks About the Individual Instruments (in Alphabetical Order as in Table 8)
ARL 9400: As far as we are aware, ARL was the rst manufacturer (and still the only
one) to use counting Moire fringes for its goniometer positioning. The
goniometer is a complete gearless construction and it has the highest slew rates
Copyright 2002 Marcel Dekker, Inc.
Table10 Presentation of Simultaneous Instruments
Manufacturer
1 2 3 4 5
Generator
Power (kW) 3 3 3 0.2 3
Voltage (kV), step (kV) ?60=? ?60=? 560=1 (a) 50 2060=1
Current (mA), step (mA) ?100=? ?75=? 5125=1 4 5150=1
Stability D%/% mains 0.0001 0.0005 0.0005 0.0005 0.0001
Source
X-ray tube:
Load (kW) 3 3 3 TTT
*
Rh 66 G
Anodes Rh, W Rh, Cr, Mo Rh, other on
request
Cr Rh, Mo, W,
Au, Cr
Be windows (mm) 125, 75 125, 75 125, 75 125, 75
Primary beam filters 4 (a) 10 3
positions
Elements Cu, Al, Fe
Primary beam attenuators y, 14 y
(Yes, No)
Samples
Sample surface Up Up Down (b)
Size: diameter, height (mm) 75, 50 51, 40 51, 40 40, ? 52, 57
Sample changer (positions) 11, 32, or 8 (b) 6, 72, 300 8, 58, 100, 138
Monochromator (fixed
channels)
Takeoff angle (
2y) 0.001 ? ? ?
Angular reproducibility (
2y)
17155 5120 10100 8547
sc (
C, L=min)
>3, <18, 5 3058, <20,
610
38, 20, 3
Compressed air (bar, L=min) ? 45, ? 610, 20
Copyright 2002 Marcel Dekker, Inc.
of all spectrometers. In the series 9420S, two goniometers or additional xed
channels can be tted, adaptable to a great variety of analytical situations,
excellent for rapid nonroutine industrial control, reducing the analysis time by a
factor of 2 with respect to a conventional sequential spectrometer. Separate
vacuumpumps support the vacuumfor the sample lockandthe monochromator.
Diano 2000: This oers dual-target tubes. The targets are push-button selected and
no moving parts are present (dual lament). Two collimators are oered: a
0.032-in. spacing beam tunnel or a 0.010-in.-spacing diracted beam soller
slit. The instrument is not temperature stabilized because the 45-kg bronze
bull gear goniometer has a suciently high thermal capacity, avoiding rapid
temperature changes. It is, however, not sucient when one wants to measure
peak shifts. The crystal in the turret must be individually tuned. The W and 2W
axes are coupled but with a bisector arrangement and moved by a dc motor.
The standard detectors are a ow-proportional counter and sealed Xe, the
scintillation counter is an option. The counting is also by a pulse-height
selector. Diano oers a xed-sample and xed-element mode for analysis.
Table11 Hybrid Instruments
Manufacturer
1 2, 2
0
3
Generator
Voltage (kV)=current
(mA)=power (W)
4050=4=200 50=4=200 id.
Stability D%=%mains 0.007 0.0005 id.
Max. mains
variation (V)
210250 180264 id.
X-ray source
Tube Similar (a) Target transmission id.
Anode Rh Sc Cr
External Cooling None None id.
Sample
Sample surface up, down Down Down id.
Spinning (rpm) 30
Dimension (diameter,
height in mm)
48, 40 2647, 40 id.
Sample changer ? Single, 10 Single
Spectrometer chamber Vac, He optional Vac, He Vac
Monochromator (fixed)
Element covered FU S, S-bg., Zr, Ni,
Ni-bg., Mn,
Mn-bg., V, V-bg
Al, Si, Na, Mg,
Ti, Mn, K, Ca,
P, S, Fe
Crystals Flat, curved optional Curved id.
Additional
spectrometers
Flexi-channel
(EDXRF)
None id.
Dimensions
(H
*
W
*
D in mm)
1090
*
800
*
800 id. 1200
*
750
*
960
Weight (kg) 160 id. 180
Copyright 2002 Marcel Dekker, Inc.
Philips PW 2400, 2404: Both types dier only in the power output of the generator.
The tube power is switched along the isowatt curve (i.e., by keeping the power
constant when either the voltage or current is changed). The xed channels
that can be added have as the detector either a ow-proportional counter
(3000 kcps), Xe-sealed (1000 kcps), Kr-sealed (2000 kcps), Ne-sealed
(2000 kcps) proportional counters, or a scintillation counter (1500 kcps). The
positioning of the goniometer is done by optical disk, a sophisticated high-
performance system, giving the highest angular accuracy and reproducibility.
The goniometer is driven by dc motors. The external sample changer accepts
samples of nonuniform size.
Rigaku 2100, 3100: Both types dier only in power output of 3 and 4 kW. The
collimators are either standard=high resolution or standard high resolu-
tion=ultralight elements. The attenuator in front of the monochromator crystal
reduces the intensity by a factor 10. The advantage should be that it avoids
power switching of the x-ray tube. The sample-surface-up geometry makes it
dicult, if not impossible, to analyze solutions. A diaphragm of 1 mm
diameter allows small-spot analysis. A built-in cleaning system removes the
impurities of the wire in seconds in this instrument because the ow-counter
eciency deteriorates by contamination of the wire electrode.
Siemens (Bruker) SRS 3400: The instrument is rather conventional and easy to
service. The positioning of the decoupled arms of the goniometer is done by
stepper motors. The positioning of the sample holders is very reproducible. The
external sample changer accepts nonuniform samples also. An excellent
software SPECTRA PLUS is supplied for qualitative, semiquantitative, and
quantitative analyses. The instrument was formerly sold under the name of
Siemens but is taken over by Bruker Analytical X-Ray Systems (Bruker AXS).
Shimadzu XRF 1700, 1500: The dierence between both instruments is that the
selection of lters, diaphragms, attenuator, and crystals is automated in the
1700 model. A special feature of this instrument is the presence of diaphragms
of diameter 1, 3, 10, 20, and 30 mm. The 1-mm diaphragm together with an
appropriate sample positioning system allows element mapping.
2. Simultaneous Instrumentswith Sample Surface Up or Sample Surface Down
A summary of the characteristics of the instruments capable to determine simultaneously a
number of the elements is given in Table 10. The instruments are composed by a few till
maximum 30 small-size high-eciency spectrometers as discussed earlier, optimized for
one particular element. The set of spectrometers is dedicated to a particular series of
samples as used in, for example, the cement or steel industry. To attribute some exibility
to these instruments, one or more scanning goniometers are added.
a. Remarks and Features Common to All Instruments
All manufacturers use x-ray tubes with either a 75- or 125-mm-thick Be window.
The detectors oered are ow-proportional gas counter (Ar), sealed gas counters
lled with Kr, Ne, or Xe, and a scintillation counter.
The linearity of the photon count rate is equal or similar to the one given for
sequential instruments.
Temperature stabilization is either individual for each spectrometer or for the
ensemble.
Copyright 2002 Marcel Dekker, Inc.
Electricity power supply is single phase, 1015 kVA.
ARL oers the possibility of adding an additional x-ray diraction device within the
uorescence spectrometer.
Other manufacturers oer integrated sample transport from the spectrometer to an
x-ray diraction instrument. The CubiX XRF from Philips is a simultaneous
spectrometer for up to 14 elements, similar to the XCem. It is integrated in the
same module with an x-ray diraction instrument; both can be juxtaposed,
allowing a rapid sample transport from one instrument to another. Both devices
are operated by a single personal computer. Cubix represents a modular concept
for integrated analysis, including elemental and crystallographic determination.
b. Remarks About the Individual Instruments (in Alphabetical Order as in Table 8)
APR 8600S
1. The primary beam lter takes the place of three monochromators and cannot be
t alongside an XRD channel.
2. Eight positions for cassettes for large samples (75 mm and height of 50 mm);
11 and 32 positions for small samples (52 mm and height of 30 mm).
3. Thirty monochromators can be tted or 20 plus 1 goniometer, or 11 plus an
XRD channel.
4. The goniometer is the same as the module of the sequential instrument.
Philips PW1600: Maximum two xed channels can be substituted by a program-
mable goniometer (with at crystals and parallel beam collimation). One
programmable channel occupies space as for one xed channel.
Philips PW2600
1. Switching time of the generator is 2 s and the tube is switched along the
isowatt curve.
2. Particular feature of this instrument is that it allows liquid samples to be handled
also.
3. Two dierent goniometers can be tted: (1) one with a at crystal and parallel
beam collimation and it takes the space of one xed channel; (2) one with a
cylindrically curved crystal (logarithmic spiral optics) and two slits and it takes
the space of three xed channels. Optional beam attenuators can be tted,
enabling all channels to handle concentrations from ppm levels up to 100%.
Philips X9 Cem: The x-ray tube is a low-power Target Transmission Tube (TTT),
(Sec. IIIA). Because of the low power (200W), this instrument does not need
external water cooling.
Siemens MRS 4000: High power (4 kW) and high speed instrument.
3. Hybrid Instruments
The instrument from Oxford MDX1000 is classied as hybrid because it combines a WD xed
channel with what is called a Flexi channel, an EDXRF unit. The x-ray tube is similar to the
target transmission tube, produced by Philips, but no other details are given in their brochures.
The instrument series Venus 100 from Philips are called hybrid because Philips uses
xed channels, which are, however, switched sequentially in front of the sample. The in-
struments are delivered for specic purposes: The PW4110=10 is provided for the analysis
of industrial minerals, particularly for the ceramic industry, and the PW4100=20 and
Copyright 2002 Marcel Dekker, Inc.
PW4120=20 for the determination of, for example, nickel, sulfur, and vanadium in pet-
roleum products. The element range can be adapted to specic purposes.
Both instruments from Oxford and Philips are simple, low-cost, but high-perfor-
mance devices which can be used in harsh industrial environments; they are simple to
install and are characterized by small dimensions, light weight, no water cooling, no
compressed air, and no severe requirements for the stability of the mains.
Hybrid instruments are presented in Table 11.
4. Instruments Dedicated to Special Applicationsand Supplements
Two manufacturers, Siemens and Philips, supply for the semiconductor industry trans-
formed versions of their sequential spectrometers SRS 3300 (power up to 4 kW, mapping
possible) and PW 2800, respectively. The main feature is the sample area which must
accept silicon disks as large as 10 in. in diameter. The sample introduction system is au-
tomated and meets the standards for working in a cleanroom environment.
The manufacturers also provides as many auxiliary devices as possible. Advanced
analysis lines are available from Rigaku, Philips, and Siemens, which oer a complete
set of the mills, uxers, and polishing devices for the XRF and XRD laboratory.
Worth noting is that, on the other hand, during the last years, some new companies
appeared on the market, oering not whole instruments as such but only some specialized
parts of instrumentation. Among the most important participants in this kind of activity,
the following should be mentioned:
Advance Hivolt, oering x-ray generators, especially with 4 kW power
X-Ray Optical Systems Inc., supplying x-ray focusing and collimating polycapillary
optical devices plus designing the optical systems with x-ray semilens
Osmic Inc., the main producer of multilayers
Corporation Scientique Claisse Inc., producing uxers for sample preparation
Contact with these producers can be very meaningful alternative in case of constructing
new, home-made, and dedicated version of an instrument.
We conclude this subsection on the intercomparison of instruments with a list of the
addresses of the headquarters of the manufacturers:
ARL Applied Research Laboratories S.A., En Vallaire Ouest C, case postale, CH-
1024 Ecublens, Switzerland
Diano Corporation, 271 Salem Street, Woburn, MA 01801, USA
Oxford Instruments, 19=20 Nueld Way, Abingdon, Oxon OX14 1 TX, England
Philips Analytical X-Ray BV, Lelyweg 1, 7602 EA Almelo, The Netherlands
Rigaku International Corporation, 3-9-12, Matsubara-cho, Akishima-shi, Tokyo
196, Japan
Shimadzu Corporation, International Marketing Division, 3, Kanda-Nishikicho
1-chome, Chiyoda-ku, Tokyo 101, Japan
Siemens, A.G., now: Bruker AXS GMBH, Oestliche Rheinbrueckenstr. 50, D-76187,
Karlsruhe, Germany
VII. FUTURE PROSPECTS
The time passed since the rst edition of this handbook has been one of the most stormy
periods in research and applications of x-rays since their discovery! It is well known how
Copyright 2002 Marcel Dekker, Inc.
much Ro ntgen was confused, although totally convinced, by the impossibility to de-
monstrate the electromagnetic nature of x-rays through reection, diraction, or inter-
ference experiments! For many years, it remained impossible and further progress was very
slow and tedious. Recent years have brought the real breakthrough. We know of three
recent realizations of x-ray lenses:
Two diractive:
1. Zone plates (in other words, Fresnel diracting structures), earlier restricted to
the very soft x-rays. They are very important for the construction of the x-ray
microscopes and for the observation of living biological objects in the so-called
water window. But now, by combining the Fresnel diraction patterns with
Braggs phenomenon, the special BraggFresnel lens was constructed, well
suited for focusing hard synchrotron radiation (Aristov et al., 1986; Snigirev,
1995; Snigirev et al., 1995; Chevallier and Dhez, 1997).
2. Snigirev refractive lens (Snigirev et al., 1996; 1998; Elleaume, 1998; Hastings,
1996; Eskildsen, et al., 1998), consisting of one or two mutually perpendicular
arrays of holes drilled in low-Z dense material (Be, Al). Depending on the
number and shape of the holes, they can focus hard x-ray beams,
One version using total reection:
1. Kumakhov polycapillary lenslike device, which might be somewhat poor in
focusing capability (spot is $50500 mm wide) but very ecient with respect to
the fraction of the primary input power turned into a small spot on the analyzed
surface. Interesting products were announced by X-Ray Optical Systems Inc.,
Advertising Materials.
In parallel, great progress was realized in the construction of energy-dispersive de-
tectors with increasingly better spectral resolution. The most promising device seems to be
the microcalorimeter detector, a device with very good spectral resolution parameters.
Moreover, its spectral resolution is more or less constant over the whole spectral range
covered (120 keV). With such a device, the spectrum deconvolution problems, so char-
acteristic for devices with poor spectral resolution, should vanish. Potentially, it lowers the
detection limits, by narrowing the energy region in which the very sharp peak is placed.
Only a very restricted number of background counts is detected in that region, thus im-
proving the peak-to-background count rate ratio. However, the detection limit depends on
the maximum count rate allowed for these detectors, which is, at the moment, very re-
stricted. Recent discussions show (Ladbury, 1998) that intensive work is going on to
overcome this shortcoming in the nearest future. The aim is to get the device with an area
of 40 mm
2
, with count rates up to 10
5
cps.
Another progress took place in the construction of time-resolved excitation sources.
The classic radiation generator for the purpose is synchrotron. Probably much more
convenient, the portable and switchable version of a time-resolved x-ray source will be a
compact-ash source. Furthermore, the construction of the x-ray laseran old dream!is
also progressing, maybe slowly, but in a promising direction.
Collected together, all available information about this progress justies hope on the
construction of new versions of existing instruments as well as the advent of new ones. It is
worth noting that in the present chapter we considered an XRF spectrometer mainly as a
device used for the determination of elemental composition. What is now discussed tes-
ties that future instruments will give both compositional and spatial resolution (mi-
croprobe) and time resolution (kinetic probe). New kinds of lenses and normal incidence
Copyright 2002 Marcel Dekker, Inc.
mirrors give an impetus to the construction of x-ray microscopes also in the hard part of
the x-ray spectrum. One can easily imagine a microscope (but one easily construct!)
working in the hard x-rays regime supported by two dierent x-ray lensesone acting as
an ocular lens and one as an objective lens.
The new version of the microprobe could then eciently concentrate x-rays (e.g., by
Kumakhov polycapillary) on the conned areas and then record the uorescent radiation
with the excellent spectral resolution of a microcalorimeter as the energy-dispersive de-
tector with the whole spectrum completed in one operation. Such a device is now under
construction in the National Institute of Standards and Technology (Wollman et al.,
1997a, 1997b; Anonymous, 1997). The microcalorimeter detector, attached to a scanning
electron microscope, has, for the time being, a spectral resolution of 7.2 eV0.4 eV,
calculated in the standard way for the MnKa line, or 2.0 eV for 1.5 keV (Wollman et al.,
2000). Count rates are below 150 cps, but it is, by far, not the last word in this matter.
A newcomer in the generation of possible applications of the classical WDXRF,
rather unexpectedly brought to the awareness of scientists, was that the extended ne
structure of uorescent spectra emitted by solid samples can be observed (Hayashi et al.,
1997; Kawai et al., 1998). Up to now, this eld was reserved for the absorption mode,
mostly in combination with synchrotron radiation. The discovery of the extended ne
structure in experiments performed under normal laboratory conditions opens a new eld
in the study of local structure in condensed matter and makes this method available to
lower-budget laboratories.
ACKNOWLEDGMENTS
All manufacturers of instruments are gratefully acknowledged for supplying documents
about their instruments. Mr. P. Van der Aa, Dr. B. Volbert, and Dr. ir. B. A. R. Vrebos
from Philips Analytical and Mr. S. Ornigg, Dr. K. E. Mauser, and Dr. N. Broll from
Siemens (Bruker) AG deserve our gratitude for invaluable advice and critical reading of
the manuscript. Dr. D. de Boer is acknowledged for supplying us with his knowledge on
the most recent progress in multilayer research. A. Kuczumow thanks the Catholic Uni-
versity of Leuven for a fellowship that enabled him to participate in the rst edition of this
work and the Flemish Government for a fellowship helpful in the present revised version.
REFERENCES
Adamson BW, Juchli K, Matula G. Adv X-Ray Anal 34:131, 1991.
Agarwal BK. X-Ray Spectroscopy. Berlin: Springer-Verlag, 1979.
Anonymous. J Res NIST 102:392, 1997.
Anonymous. Rigaku J 14:37, 1997.
Arai T. Adv X-Ray Anal 30:213, 1987.
Arai T. X-Ray Spectrom 20:9, 1991.
Arai T, Omote K. Adv X-Ray Anal 31:507, 1988.
Arai T, Shoji T. Adv X-Ray Anal 33:521, 1990.
Aristov VV, Gaponov SV, Erko AI, Salaschenko NN. Optics News 12:128, 1986.
Asada E. Jpn J Appl Phys 12:1946, 1973.
Asada E. Jpn J Appl Phys 15:1417, 1976.
Asada E, Takiguchi T, Suzuki Y. X-Ray Spectrom 4:186, 1975.
Attaelmanan A, Larsson S, Rindby A, Voglis P, Kuczumow A. Rev Sci Instrum 65:7, 1994.
Copyright 2002 Marcel Dekker, Inc.
Bandas D, Kellogg E, Murray S, Enck E Jr. Rev Sci Instrum 49:1273, 1978.
Barbee, TW Jr. In: Atwood DT, Henke BL, eds. Low Energy X-Ray Diagnostics. AIP Conf
Proceedings Vol. 75. New York: American Institute of Physics, 1981, p 124.
Barbee TW Jr. Opt Eng 25:898, 1986.
Becker RS, Golovchenko JA, Patel JR. Phys Rev Lett 50:153, 1983.
Beers Y. Theory of Error. Reading, MA: Wesley 1958.
Bent H. Spectrochim Acta 25B:613, 1970.
Bertin EP. Principles and Practice of X-Ray Spectrometric Analysis 2nd ed. New York: Plenum, 1975.
Bilderback DH, Homan SA, Thiel DJ. Science 263:201, 1994.
Birks LS. X-Ray Spectrochemical Analysis, 2nd ed. New York: Interscience, 1969.
Birks LS. In: Robinson JW, ed. Handbook of Spectroscopy. Vol. I. Cleveland, OH: CRC Press,
1974, pp 1254.
Birks LS, Gilfrich JV. Spectrochim Acta 33B:305, 1978.
Bishop C. Rev Sci Instrum 63:4772, 1992.
Boehme DR. Adv X-Ray Anal 30:39, 1987.
Bonetto RD, Riveros JA. X-Ray Spectrom 13:44, 1984.
Bonvin D, Juchli K, Adamson BW. Adv X-Ray Anal 38:711, 1995.
Bos M, Weber HT. Anal Chim Acta 247:97, 1991.
Botden PJM, Combee B, Houtman J Philips Tech Rev 14:165, 1952.
Bougault H, Cambon P, Toulhoat H. X-ray Spectrom 6:66, 1977.
Bragg WH, Bragg WL. Proc Roy Soc 88A:428, 1913.
Brill M. Metall 50:504, 662, 1996.
Broll N. Proceedings C. R. Coll. Rayons X, Siemens. Grenoble, 1985, pp 284304.
Brown DB, Gilfrich JV, Peckerar MC. J Appl Phys 46:4537, 1975.
Brunetto MG, Riveros JA. X-Ray Spectrom 13:60, 1984.
Buckley A. 15th International Congress on X-ray Optics Microanalysis, ICXOM, Antwerp, 1998.
Buhrke VE, Jenkins R, Smith DK. Preparation of Specimens for X-Ray Fluorescence and X-Ray
Diraction Analysis. New York: WileyVCH, 1998.
Burkhalter PG, Campbell WJ. Oak Ridge National Laboratory Report ORNL-IIC-10. Vol. 1, 1967.
Caciuo R, Melone S, Rustichelli F, Boeuf A. Phys Rep 152:1, 1987.
Carpenter DA. X-Ray Spectrom 18:253, 1989.
Cauchois T, Senemaud C. International Tables of Selected Constants. 18. Wavelengths of X-Ray
Emission Lines and Absorption Edges. Oxford: Pergamon Press, 1978, p 67.
Chang WZ, Wittry DB, Microbeam Anal 3:24, 1994.
Chevallier P, Dhez P. Hard X-Ray and Microbeam Production and Application. In Shafroth JM,
Austin JC, eds. Accelerator-Based Atomic Physics Techniques and Applications. Woodbury,
NY: American Institute of Physics, 1997, pp 309348.
Cholewa M. Book of Abstracts 15th International Congress on X-ray Optics Microanalysis.
Antwerp, 1998, p 58.
Chumakov AI, Metge J, Baron AQR, Gru ndsteudel H, Gru ndsteudel HF, Ru er R, Ishikawa T.
Nucl Instrum Methods A 383:642, 1996.
Claes M, Van Grieken R, De Bokx P. X-Ray Spectrom 26:153; 1997.
Claisse F. Norelco Rep 4:95, 1957.
Claisse F, Quintin M. Can Spectrosc 12:129, 1967.
Claisse F, Thinh TP. Anal Chem 51:954, 1979.
Clayton CA, Hines JW, Elkins PD. Anal Chem 59:2506, 1987.
Compton AH, Allison SK. X-Rays in Theory and Experiment. New York: Van Nostrand, 1935.
Coolidge WD. Phys Rev 2:409, 1913.
Crasemann B. In: Matthews D, Eder D, eds. Proceedings 4th International Colloquium on X-Ray
Lasers. New York: American Institute of Physics, 1994, pp 18.
Croke JE, Nicolosi J. Adv X-Ray Anal 30:225, 1987.
Dagabov SB, Nikitina SV, Kumakhov MA, Ibraimov NS, Vartaniants GA, Nikitin AN,
Spielberger L. Nucl Instrum Methods B103:99, 1995.
Copyright 2002 Marcel Dekker, Inc.
Dane AD, Timmermans PAM, van Sprang HA, Buydens LMC. Anal Chem 68:2419, 1996.
de Boer DKG. X-Ray Spectrom 19:145, 1990.
de Boer DKG, Brouwer PN. Adv X-Ray Anal 33:237, 1990.
de Bokx PK, Urbach HP. Rev Sci Instrum 66:15, 1995.
de Korte PAJ. Proceedings ESA Symposium on Photon Detectors for Space Instruments.
Noordwijk, 1992, p 41.
de Vries JL. Fresenius Z Anal Chem 324:492, 1986.
Dos Santos JMF, Bento ACSS, Conde CAN. X-Ray Spectrom 22:328, 1993.
Du Mond J, Youtz JP. J Appl Phys 11:357, 1940.
Dyson NA. X-Rays in Atomic and Nuclear Physics. London: Longman, 1973.
Ebel H, Mantler M, Gurker N, Wernisch J. X-Ray Spectrom 12:47, 1983.
Ebel H, Ebel MF, Wernisch J, Poehn CH, Wiederschwinger H. X-Ray Spectrom 18:89, 1989.
Eisenberger P, Platzman PM, Winick H. Phys Rev Lett 36:623, 1976a.
Eisenberger P, Platzman PM, Winick H. Phys Rev B, Solid State 13:2377, 1976b.
Elleaume P. J. Synchrotr Rad 5:1, 1998.
Ellis AT, Potts PJ, Holmes M, Oliver GJ, Streli CH. Wobrauschek P. J. Anal Atomic Spectrom
12:461R, 1997.
Feldman LC, Mayer JW. Fundamentals of Surface and Thin Film Analysis. New York: North-
Holland, 1986.
Fernandez JE. X-Ray Spectrom 18:271, 1989.
Fernandez JE, Molinari VG. Adv X-Ray Anal 33:573, 1990.
Fewster PF. Rep Prog Phys 59:1339, 1996.
Fill EE. In: Proceedings of Laser Interaction with Matter Conference, Oxford, 1994. Inst Phys Conf
Ser No 140. Bristol: Institute of Physics Publishing Ltd., 1995, pp 111116.
Finkbeiner F, Hahn A, W. Heeringa, Jany P, Klages HO, Strobelt T. Nucl Instrum Methods
A306:215, 1991.
Fiori C, Myklebust R, Heinrich K, Yakowitz H. Anal Chem 48:233, 1976.
Fitzgerald R, Gantzel P. Energy Dispersion X-Ray Analysis: X-Ray and Electron Probe Analysis.
ASTM Spec Techn Pub 485. Philadelphia: ASTM, 1971.
Folkmann F, Frederiksen F. Nucl Instrum Methods B49:126, 1990.
Franks A. Sci Prog (Oxford) 64:371, 1977
Freiser H, Nancollas GH. IUPAC Compendium of Analytical Nomenclature: Denitive Rules 1987.
2nd ed. Oxford: Blackwell Scientic, 1987.
Gardner RP, Hawthorne AP. X-Ray Spectrom 4:138, 1975.
Gentry RV, Cahill TA, Fletcher NR, Kaufman HC, Medsker LR, Nelson JW, Flocchini RG. Phys
Rev Lett 37:11, 1976.
Germer R. J Phys E: Sci Instrum 12:336, 1979.
Gilfrich JV. In: Robinson JW, ed. Handbook of Spectroscopy. Vol. I. Cleveland, OH: CRC Press,
1974, pp 232237.
Gilfrich JV. Adv X-Ray Anal 30:35, 1987.
Gilfrich JV, Birks LS. Anal Chem 40:1077, 1968.
Gilfrich JV, Nagel DJ, Loter NG, Barbee TW Jr. Adv X-Ray Anal 25:355, 1982.
Gilfrich JV, Burkhalter PG, Whitlock RR, Warden ES, Birks LS. Anal Chem 43:943, 1971.
Gillam E, Heal HT. Br J Appl Phys 3:353, 1952.
Glade GH, Post HR. Appl Spectrom 24:193, 1970.
Gohshi Y. Spectrochim Acta 36B:763, 1981.
Gohshi Y, Hukao Y, Hori K. Spectrochim Acta 27B:135, 1982a.
Gohshi Y, Kamada H, Kohra K, Utaka T, Arai T. Appl Spectrosc 36:171, 1982b.
Green M, Cosslett VE. Proc Phys Soc 78:1206, 1961.
Gries WH. Microchim Acta 107:117, 1992.
Grubbs FE. Technometrics 11:1, 1969.
Habulihaz B, Martins E, Gamblin S, Urch DS. X-Ray Spectrom 25:15, 1996.
Hastings J. Nature 384:22, 1996.
Copyright 2002 Marcel Dekker, Inc.
Haycock DE, Urch DS. X-Ray Spectrom 7:206, 1978.
Heinrich KFJ. In: Heinrich KFJ, Newbury DE, Myklebust RL, Fiori CE, eds. Energy Dispersive X-
Ray Spectrometry. NBC Spec Publ 604. Washington, DC: National Bureau of Standards,
1981.
Helsen JA, Vrebos BAR. Analyst 109:295, 1984a.
Helsen JA, and Vrebos BAR. Spectrochim Acta 39B:751, 1984b.
Helsen JA, Vrebos BAR. Int Lab 16(10):66, 1986a.
Helsen JA, Vrebos BAR. X-Ray Spectrom 15:173, 1986b.
Helsen J, Wijnhoven J. Bull Groupe Fr Argiles 24:15, 1972.
Henke BL, Uejio JY, Yamada HT, Tackaberry RE. Opt Eng 25:937, 1986.
Henke BL, Lee P, Tanaka TJ, Shimabukuro RL, Fujikawa BK. Atomic and Nuclear Data Table.
Vol. 27. New York: Academic Press, 1982.
Henke BL, Davis JC, Gullicksen EM, Perera RC. LBL Report 26259 UC-411, November 1988.
Hilton GC, Martinis JM, Wollman DA, Irwin KD, Dulcie LL, Gerber D, Gillevet PM,
Twerenbold D. Nature 391:672, 1998.
Hofmann S, Ninov N, Hessberger FP, Armbruster P, Folger H, Mu nzenberg G, Scho tt HJ,
Popeko AG, Yeremin AV, Andreyev AN, Saro S, Janik R, Leino M. Report GSI-94-82,
Darmstadt, 1994.
Ice GE, Sparks CJ Jr. Nucl Instrum Methods 222:121, 1984.
Irvin KD, Hilton GC, Wollman DA, Martinis JM. Appl Phys Lett 69:1945, 1996.
Iwatsuki M, Fukasawa T. X-Ray Spectrom 16:73, 1987.
Jaklevic JM, Giauque RD, Thompson AC. Anal Chem 60:482, 1988.
Jamelot G. In: Liesen D, ed. Physics with Multiply Charged Ions. New York: Plenum Press, 1995,
pp 291317.
Jenkins R. An Introduction to X-Ray Spectrometry. London: Heyden, 1974.
Jenkins R. Anal Chem 56:1099A, 1984.
Jenkins R, De Vries JL. Practical X-Ray Spectrometry. 2nd ed. New York: Springer-Verlag, 1967.
Jenkins R, Gould RW, Gedcke D. Quantitative X-Ray Spectrometry. 2nd ed. New York: Marcel
Dekker, 1995.
Jenkins R, Hahm YH, Pearlman S. Norelco Rep 26:27, 1979.
Jenkins R, Hammell B, Cruz A, Nicolosi J. Norelco Rep 31: 1984.
Kataria SK, Govil R, Saxena A, Bajpei HN. X-Ray Spectrom 15:49, 1986.
Kearney PA, Slaughter JM, Falco CM. Opt Eng 30:1076, 1991.
Kevex Corp, Kevex Material Analyzer Model 6500 Brochure, p 2.
Kikkert JN. Spectrochim Acta 38B:1497, 1983.
Kirz J, Jacobsen C, Howells M. Q Rev Biophys 28:33, 1995.
Klimasara AJ, Berry TL. Adv X-Ray Anal 30:325, 1987.
Klockenka mper R. Spectrochim Acta 42B:423, 1987.
Klockenka mper R. Total-Reection X-Ray Fluorescence Analysis. New York: Wiley, 1997.
Kloyber H, Ebel H, Mantler M, Koitz S. X-Ray Spectrom 9:170, 1980.
Kno chel A, Petersen W, Tolkiehn G. Nucl Instrum Methods 208:659, 1983.
Kohler D, Guttman JL, Watson BA. Rev Sci Instrum 56:812, 1985.
Kortright JB. J Magn Magn Mater 156:271, 1996.
Kortright JB, Joksch ST, Ziegler E. J Appl Phys 69:168, 1991.
Kowalski BR, Schatzki FT, Stross FH. Anal Chem 44:2176, 1972.
Kozhevnikov IV, Fedorenko AI, Kondratenko VV, Pershin YP, Yulin SA, Zubarev EN,
Pardmore HA, Cheung KC, van Dorssen GE, Roper M, Balakireva LL, Serov RV,
Vinogradov AV. Nucl Instrum Methods A 345:594, 1994.
Kramers HA. Philos Mag 46:836, 1923.
Kuczumow A. X-Ray Spectrom 11:112, 1982.
Kuczumow A. X-Ray Spectrom 13:16, 1984.
Kuczumow A. Spectrochim Acta 43B:737, 1988.
Kuczumow A, Helsen JA. X-Ray Spectrom 18:63, 1989.
Copyright 2002 Marcel Dekker, Inc.
Kuczumow A, Holland G. X-Ray Spectrom 18:5, 1989.
Kuczumow A, Helsen JA, Wouters R. Spectrochim Acta 47B:971, 1992.
Kuczumow A, Rindby A, Larsson S. X-Ray Spectrom 24:19, 1995.
Kuczumow A, Schmeling M, Van Grieken R. J Anal At Spectrom 15:535, 2000.
Kuczumow A, Vekemans B, Claes M, Schalm O, Vincze L, Dorrine W, Gysels K, Van Grieken R.
X-Ray Spectrom 28:282, 1999.
Kulenkamp H. Ann Phys 69:548, 1922.
Kumakhov MA, Komarov FF. Phys Rep 191:289, 1990.
Lachance GR., Proceedings of the International Conference on Industrial Inorganic Analysis, Metz,
France, 1981.
Lachance GR, Claisse F. Quantitative X-Ray Fluorescence Analysis: Theory and Application.
Bans Lane, UK: Wiley, 1995.
Lachance GR, Traill RJ. Can Spectrosc 11:43, 1966.
Ladbury R. Phys Today 51(7):19, 1998.
Lagarde G, Brissaud I, Cailleret J, Fontes P, Heitz CH. Nucl Instrum Methods B1:45, 1984.
Larsen JT, Morgan WL, Goldstein WL. Rev Sci Instrum 63:4775, 1992.
Le Grand JB, Bruijn MP, Frericks M, Laauwen W, Nilsson D, de Korte PAJ, Valko P,
Gijsbertsen JG, Houwman EP, Flokstra J. J Low Temp Phys 93:573, 1993.
Lepe tre Y, Schiller JK, Rasigni G, Rivoira R, Philip R, Dhez P. Opt Eng 25:948, 1986.
Lessyna L, Di Marzio D, Gottesman S, Kesselman M. J Low Temp Phys 93:779, 1993.
Lindh AE, Lundquist O. Ark Mat Astro Fys 18:14, 34, 35, 1924.
Li-Xing Z. X-Ray Spectrom 13:52, 1984.
Lo I Yin, Trombka JI, Seltzer SM. Nucl Instrum Methods A277:619, 1989a.
Lo I Yin, Trombka JI, Seltzer SM, Johnson RG, Philpotts JA. J Geophys Res 94:13,611, 1989b.
London RA. Phys Fluids B 5:2707, 1993.
Long GL, Winefordner JD. Anal Chem 55:712A, 1983.
Loomis TC, Keith HD. X-Ray Spectrom 5:104, 1976.
Luck SR, Urch DS, Zheng DH. X-Ray Spectrom 21:77, 1992.
Luggar RD, Gilboy WB. Nucl Instrum Methods A353:650, 1994.
Luo L, Guo C, Ma G, Ji A. X-Ray Spectrom 26:15, 1997.
MacGowan BJ, Mrowka S, Barbee TW Jr, Da Silva LB, Eder DC, Koch JA, Pan LS, Turner JA,
Underwood JH, Young PE. J X-Ray Sci Technol 3:231, 1993.
Manne R. In: Fabian DJ, Kleinpoppen H, Watson LM, eds. Inner-Shell and X-Ray Physics of
Atoms and Solids. New York: Plenum, 1981, p 699.
Mantler M. Adv X-Ray Anal 27:433, 1984.
Markowicz AA, Van Grieken RE. Anal Chem 56:2049, 1984.
Markowicz AA, Van Grieken RE. Anal Chem 58:279R, 1986.
Martins E, Urch DS. Adv X-Ray Anal 35:1069, 1992.
McCammon D, Cui W, Juda M, Morgenthaler J, Zhang J, Kelley RL, Holt SS, Madejski GM,
Moseley SH, Szymkowiak AE. Nucl Instrum Methods A326:157, 1993.
Michaelis RE, Kilday BA. Adv X-Ray Anal 5:405, 1962.
Moseley HGJ. Philos Mag 26:1024, 1913.
Moseley HGJ. Philos Mag 27:703, 1914.
Moseley SH, Mother JC, McCammon D. J Appl Phys 56:1257, 1984.
Murata M, Shibahara H. X-Ray Spectrom 10:41, 1981.
Nagel DJ. In: Bonelle C, Mande C, eds. Advances in X-Ray Spectroscopy. Oxford: Pergamon Press,
1982.
Nichols MC, Boehme DR, Ryon RW, Wherry D, Cross B, Aden G. Adv X-Ray Anal 30:45, 1987.
Nicolosi JA, Groven JP, Merlo D. Adv X-Ray Anal 30:183, 1987.
Nicolosi JA, Groven JP, Merlo D, Jenkins R. Opt Eng 25:964, 1986.
Nomenclature, symbols, units and their usage in spectrochemical analysisII. Spectrochim Acta
39B:242, 1978.
Nordfors N. Ark Fys 10:279, 1956.
Copyright 2002 Marcel Dekker, Inc.
Novosel-Radovic V, Maljkovic D, Nenadic N. X-Ray Spectrom 13:148, 1984.
Omote K, Arai T. Adv X-Ray Anal 32:83, 1989.
Otto M, George T, Schierle C, Wegscheider W. Pure Appl Chem 64:497, 1992.
Ovonic Synthetic Materials Co. Advertising matter and letter to customers. Troy, MI, March 30,
1987.
Palicarpo AJPL. Space Sci Instrum 3:77, 1978.
Payne AP, Clemens BM. Phys Rev B 47:2289, 16,068, 1993.
Pella PA, Feng L, Small JA. X-Ray Spectrom 14:125, 1985.
Pella PA, Tao GY, Lachance G. X-Ray Spectrom 15:251, 1986.
Pe rez RD, Sa nchez HJ. Rev Sci Instrum 68:2681, 1997.
Philips Application Note. No 745.
Philips Application Note. No. 737 and No. 803.
Philips Application Note. Nos 504, 622, 623, 808, 913, and 916.
Philips Technical Materials. Sequential X-Ray Spectrometer System PW2404, 7.
Philips Technical Materials. Simultaneous X-Ray Spectrometer System PW1660.
Plesch R. X-Ray Spectrom 10:8, 1981.
Plotnikow RJ, Pszenicznyj GA. Fluorescentnyj Rentgenoradiometriczeskij Analiz. Moscow:
Atomizdat, 1973.
Potts PJ, Webb PC. J Geochem Explor 44:251, 1992.
Pouvesle JM, Cachoncinlle C, Viladrosa R, E.Robert E, Khacef A. Nucl Instrum Methods
B113:134, 1996.
Quiset JP. Spectra Anal 25:188, 1996.
Rao-Sahib TS, Wittry DB. In: Shinoda G, Kohra K, Ichinokawa T, eds. Proceedings 6th
International Conference on X-Ray Optics Microanalysis. Tokyo: University of Tokyo Press,
1972, p 131.
Reed SJB, Buckley A. Microchim Acta 13(suppl.):479, 1996.
Reed SJB, Ware NG. J Phys E 6:286, 1973.
Remond G, Campbell JL, Packwood RH, Fialin M. Scan Microsc(suppl)7:89, 1993.
Remond G, Gilles C, Fialin M, Rouer O, Marinenko R, Myklebust R, Newbury D. Microchim Acta
13(suppl):61, 1996.
Rindby A. Nucl Instrum Methods A249:536, 1986.
Ro ntgen WC. Nature 53:274, 1896.
Rosen RS, Stearns DG, Viliardos MA, Kassner ME, Vernon SP, Cheng Y. Appl Opt 32:6975, 1993.
Rousseau RM. X-Ray Spectrom 3:102, 1974.
Rousseau RM. X-Ray Spectrom 13:115, 1984.
Rousseau RM. X-Ray Spectrom 16:103, 1987.
Rousseau RM, Claisse F. X-Ray Spectrom 3:31, 1974.
Rousseau RM, Willis JP, and Duncan AR. X-Ray Spectrom 25:179, 1996.
Rousseaus Software for XRF Analysis of Homogeneous and Non-homogeneous Materials,
Advertising materials of Les Logiciels R. Rousseau Inc., Cantley Canada, 1996.
Rousseuw PJ, Leroy AM. Robust Regression and Outlier Detection. New York: Wiley, 1987.
Salem SI, Lee PL. In: Atomic Data and Nuclear Data Tables. Vol. 18. 1976, p 233.
Sasaki YC. Jpn J Appl Phys 30:761, 1990.
Scha fers F, Mertins H-Ch, Schmolla F, Packe I, Salashchenko NN, Shamov EA. Appl Opt 37:719,
1998.
Schmitt W, Rothe J, Hormes J, Gries WH, Fresenius J Anal Chem 349:144, 1994.
Schomann B, Wernisch J, Ebel H. Adv X-Ray Anal 38:291, 1995.
Schomann B, Wiederschwinger H, Ebel H, Wernisch J. Adv X-Ray Anal 39:127, 1997.
Schwenke H, Knoth J. Nucl Instrum Methods 193:239, 1982.
Seltzer SM, Berger MJ. Nucl Instrum Methods B 12:95, 1985.
Sherman J. ASTM Spec Tech Publ 157. Philadelphia: ASTM, 1954.
Sherman J. Spectrochim Acta 7:283, 1955.
Shiraiwa T, Fujino N. Jpn J Appl Phys 5:886, 1966.
Copyright 2002 Marcel Dekker, Inc.
Shiraiwa T, Fujino N. Bull Chem Soc Jpn 40:2289, 1967.
Short MA. Adv. X-Ray Anal 34:319, 1991.
Siemens SRS Anwendung. Booklet nos 84=8, 86=1, 86=5, 86=6.
Siemens Lab Report X-Ray Anal: XRF Analysis of Beryllium in Copper. 1994.
Siemens Booklet. Factors Aecting X-Ray Intensities (Sensitivity) in Sequential X-Ray Spectro-
meters.
Sigmund P. In: Behrisch R, ed. Sputtering by Particle Bombardment I. New York: Springer-Verlag,
1981, Chap 2.
Silver E, LeGros M, Madden N, Beeman J, Haller E. X-Ray Spectrom 25:115, 1996.
Silver E, LeGros M, Austin G, Madden N, Beeman J, Haler E. X-Ray Spectrom 26:265, 1997.
Simo es PCPS, dos Santos JMF, Conde CAN. X-Ray Spectrom 26:182, 1997.
Small JA, Leigh SD, Newbury DE, Myklebust RL. J Appl Phys 61:459, 1987.
Snigirev A. Rev Sci Instrum 66:1053, 1995.
Snigirev A, Kohn V, Snigireva I, Lengeler B. Nature 384:49, 1996.
Snigirev A, Kohn V, Snigireva I, Souvorov A, Lengeler B. Appl Opt 37:653, 1998.
Snigirev A, Snigireva I, Engstro m P, Lequien S, Suvorov A, Hartman Y, Chevallier P, Idir M,
Legrand F, Soullie G, Engrand C. Rev Sci Instrum 66:1461, 1995.
So G, Lustig H-J, Mu ller B. H Phys G: Nucl Phys 3:583, 1977.
Sparks CJ Jr. Phys Rev Lett 33:262, 1974.
Sparks CJ Jr. Adv X-Ray Anal 19:19, 1976.
Sparks CJ Jr. In: Winick H, Doniach S, eds. Synchrotron Radiation Research. New York: Plenum,
1980, p 459.
Sparks CJ Jr, Raman S, Ricci E, Gentry RV, Krauze MO. Phys Rev Lett 40:507, 1978.
Spiller E. In: Atwood DT, Henke BL, eds. Low Energy X-Ray Diagnostics. AIP Conference
Proceedings Vol 75. New York: American Institute of Physics, 1981, p 124.
Spiller E, Rosenbluth AE. Opt Eng 25:954, 1986.
Statham PJ. X-Ray Spectrom 5:16, 1976a.
Statham PJ. X-Ray Spectrom 5: 154, 1976b.
Sumbaev OI. Sov Phys JETP (Engl Transl) 30:927, 1970.
Sumbaev OI. In: Fluck E, Goldanskii VI, eds Modern Physics in Chemistry. Vol I. London:
Academic Press, 1976, p 31.
Sumbaev OI, Petrovic EV, Smirnov YP. Ergorov AI, Zykov US, Gruskhko AI. Sov Phys JETP
(Engl. Transl) 26:891, 1968.
Suzuki T, Kishimoto T, Kaji T. J Phys Soc Jpn 29: 730, 1970.
Szalo ki I, To ro k SB, Chul-Un Ro, Injuk J, Van Grieken RE. Anal Chem 72:211R, 2000.
Tanninen VP, Mikkola E, Hyva rinen HK, Grekula A, Kallioma ki PL. X-Ray Spectrom
14:188, 1985.
Tertian R, Broll N. X-Ray Spectrom 13:134, 1984.
Tertian R, Claisse F. Principles of Quantitative X-Ray Fluorescence Analysis. London: Heyden,
1982.
Thiel DJ, Stern EA, Bilderback DH, Lewis A. Physica B 158:314, 1989.
Thordarson S. Ann Phys 35: 135, 1939.
To ro k SB, La ba r J, Injuk J, Van Grieken R. Anal Chem 68:467R, 1996.
To ro k SB, La ba r J, Schmeling M, Van Grieken RE. Anal Chem 70: 495R, 1998.
Trincavelli J, Castellano G, Riveros JA. X-Ray Spectrom 27:81, 1998.
Tsuji K, Wagatsuma K. Jpn J Appl Phys 35:L1535, 1996.
Tsuji K, Sato S, Hirokawa K. Rev Sci Instrum 66:4847, 1995.
Tsuji K, Wagatsuma K, Hirokawa K. T. Trace Micropobe Tech 15:1, 1997.
Uda M, Benka O, Fuwa K, Maeda K, Sasa Y. Nucl Instrum Methods B22:5, 1987.
Uhlig S, Mu ller L. Siemens Anal Appl Note. No.326, 1991.
Underwood JH. Optics News 12: March 1986.
Underwood JH, Atwood DTA. Phys Today 37:44, 1984.
Underwood JH, Barbee TW Jr. Appl Opt 20:3027, 1981.
Copyright 2002 Marcel Dekker, Inc.
Urbach HP, de Bokx PK. Phys Rev B 53:3752, 1996.
Urch DS, Wood PR. X-Ray Spectrom 7:9, 1978.
Van Eenbergen A, Volbert B. Adv X-Ray Anal 30:201, 1987.
Van Grieken R, Markowicz A, To ro k S. Fresenius Z. Anal Chem 324:825, 1986.
van Kan JA, Vis RD. Nucl Instrum Methods B109=110:85, 1996a.
van Kan JA, Vis RD. Nucl Instrum Methods B113:373, 1996b.
van Kan JA, Vis RD. Spectrochim Acta B52:847, 1997.
van Sprang HA, Bekkers MHJ. X-Ray Spectrom 27:31, 37, 1998.
Vekemans B. PhD thesis, University of Antwerp, 2000.
Verbeni R, Sette, F, Krish MH, Bergmann U, Gorges B, Halcoussis C, Martel K, Maschoviecco C,
Ribois JP, Ruocco G, Sinn H. J Synchrotr Radiat 3:62, 1996.
Vis RD, van Langevelde F. Nucl Instrum Methods B61:515, 1991.
Vrebos B, Helsen JA. Spectrochim Acta 38B:835, 1983.
Vrebos BAR, Helsen JA. Adv X-Ray Anal 28:37, 1985a.
Vrebos B, Helsen JA. X-Ray Spectrom 14:27, 1985b.
Vrebos BAR, Helsen JA. X-Ray Spectrom 15:167, 1986.
Watson RL, Leeper AK, Sonobe BI, Nucl Instrum Methods 142:311, 1977.
Weaver CE, Pollard LD. The Chemistry of Clay Minerals. Amsterdam: Elsevier, 1973.
Weebs ME, Leicester HM. Discovery of the Elements. 7th ed. Easton, PA: J. Chem. Educ., 1967.
Wobrauschek P, Aiginger H. X-Ray Spectrom 9:57, 1980.
Wobrauschek P, Aiginger H. X-Ray Spectrom 12:72, 1983.
Wobrauschek P, Aiginger H. Adv. X-Ray Anal 28:64, 1985.
Wobrauschek P, Aiginger H. Fresenius Z Anal Chem 324:865, 1986.
Wollman DA, Irwin KD, Hilton GC, Dulcie LL, Newbury DE, Martinis JM. J Microsc
188:196, 1997.
Wollman DA, Nam SW, Newburry DE, Hilton GC, Irwin KD, Bergren NF, Deiker S, Rudman DA,
Martinis JM. Nucl Instrum Methods A 444:145, 2000.
X-Ray Optical Systems Inc. Advertising Materials. November 1997 and July 1998.
Xunliang D, Wei L, Yiming YJ. Beijing Norm Univ 31(suppl):40, 1995.
Yap CT. X-Ray Spectrom 16:229, 1987.
Yap CT, M. Tang SM. Appl Spectrom 38:527, 1984.
Yap CT, Tang SM. X-Ray Spectrom 14:157, 1985.
Yiming Y, Xunliang D, Dachun W. Adv X-Ray Anal 37:507, 1994.
Copyright 2002 Marcel Dekker, Inc.
3
Energy-Dispersive X-ray Fluorescence
Analysis Using X-ray Tube Excitation
Andrew T. Ellis
Oxford Instruments Analytical Ltd., High Wycombe, Buckinghamshire, England
I. INTRODUCTION
The scope of this chapter is laboratory and industrial x-ray uorescence (XRF) analysis
systems in which x-ray tubes are used for excitation and energy-dispersive (ED) semi-
conductor detectors, as opposed to Bragg-diraction (wavelength) dispersion devices,
are used. Such ED detectors directly measure the energy of the x-rays by collecting the
ionization produced in a suitable semiconductor material. The coverage of this chapter
specically excludes the following topics, which are covered in other chapters: on-line
EDXRF systems (see Chapter 7), total-reection XRF (TXRF) (see Chapter 9), polarized
beam XRF (see Chapter 10), and x-ray microuorescence (XRMF) (see Chapter 11).
Early approaches to EDXRF used gas proportional counters or scintillation de-
tectors to determine directly the energy of the x-rays. Such systems were limited in their
application because of the inherently poor energy resolution, which precluded the se-
paration of characteristic x-rays of adjacent elements in the periodic table. The limited
actual energy resolution of such detectors led them to be known as nondispersive, but this
drawback has been eectively countered in some benchtop instruments through the use of
novel arrangements of primary and secondary beam lters (Ross, 1928; Kirkpatrick, 1939;
Kirkpatrick, 1944; Field, 1993). The real breakthrough in EDXRF came in the late 1960s
(Bertolini et al., 1965; Bowman et al., 1966; Elad and Nakamura, 1966; Aitken, 1968) with
the arrival of solid-state semiconductor diode detectors and their associated pulse pro-
cessing circuits. These detection systems were developed through the early 1970s, often in
electron microscopes, to the point at which practical x-ray spectrometry with an energy
resolution of 200 eV or less became possible (Frankel and Aitken, 1970; Landis et al.,
1971; Heath, 1972). Although the energy-resolution capability of semiconductor detectors
remained greatly inferior to that achieved by wavelength-dispersive (WD) XRF systems,
the increased eciency inherent in the energy-dispersive method compensated in many
analytical applications and permitted the use of a multiplicity of experimental geometries
not practical with WDXRF. A wide variety of EDXRF analytical systems based on
radioisotopoe sources, x-ray tubes, charged-particle accelerators, microprobe electron
beams, and synchrotron light sources have been developed in recent years.
Copyright 2002 Marcel Dekker, Inc.
Within the scope of this chapter, only EDXRF systems in which x-ray tubes are used
for excitation will be considered further. The material in this chapter discusses the in-
strumental aspects of the energy-dispersive method, with particular emphasis on the use of
low-power, compact x-ray tubes in combination with semiconductor detectors.
The basic components of an EDXRF instrument within the scope of this chapter are
depicted schematically in Figure 1. Each of the main subsystems of such an instrument is
covered in detail in the following sections.
II. X-RAY TUBE EXCITATIONSYSTEMS
In EDXRF spectrometry, there is no physical discrimination of the secondary radiation
that leaves the sample and enters the detector. This means that all photons of all energies
in this secondary beam interact with the detector. The detector and its associated signal
processing chain have a limited capacity to process these events and this is typically in the
range 150 kcps. As a direct consequence, EDXRF has a limited total counting capacity
Figure1 Basic components of a tube-excited EDXRF instrument.
Copyright 2002 Marcel Dekker, Inc.
for all element x-ray emissions and it is, therefore, essential that the content of the incident
spectrum be optimized for its useful information. This may require that more than one
acquisition be made under dierent excitation conditions in order to control the range of
energies needing to be collected in the detection system. Alternatively, the excitation
system may be optimized to increase the intrinsic peak-to-background ratio through the
use of alternative geometries such as TXRF or polarization EDXRF. In both these cases,
however, higher-power x-ray tubes are typically required and there is still the need to apply
some type of selectivity to the excitation system. This selectivity can be achieved in TXRF
using more than one tube target materialtypically Mo and W. In the case of polarization
EDXRF, more than one tube high voltage or polarizing element is likely to be applied.
This selectivity of excitation employed in EDXRF is completely the reverse of
conventional WDXRF practice. In WDXRF, the high discrimination power of the dis-
persion and detection process and its associated losses coupled with very simple detection
and counting systems allow very high count rates to be processed for each narrow wa-
velength band. Thus, the detection process is highly selective and the count rates in an
individual element channel with a modern WDXRF system can be well in excess of 10
6
CPS. Typically, the simple rule of thumb in WDXRF would be to use twice to three times
the kilovoltage (kV) of the absorption edge of the highest element of interest and then to
apply the maximum allowed current at that setting. In summary, we may regard the es-
sential operational dierence between EDXRF and WDXRF spectrometry as follows:
WDXRF uses broad-band excitation and selective detection.
EDXRF uses selective excitation and broad-band detection.
There are various ways in which selective excitation is achieved in EDXRF spec-
trometry. The use of TXRF and polarization is covered elsewhere in this volume and will
not be considered further here. The most important modes in which selective excitation is
realized in EDXRF are as follows:
Selection of tube anode material
Variation of tube voltage (kV)
Use of primary beam lters
Use of secondary targets (and associated lters)
The selection of tube anode material is typically a once-o decision made at the time of
purchase in view of the intended application. The high purchase cost of x-ray tubes
generally precludes operational changes and low-power dual or multiple (Kis-Varga, 1988)
anode x-ray tubes are not commercially available.
The types of x-ray excitation arrangements are shown schematically in Figure 2.
The following subsections describe typical arrangements for excitation and describe
how optimum excitation conditions can be determined. Earlier work on this subject may
be found in the XRF literature (Sandborg and Shen, 1984; Vane and Stewart, 1980;
Gedcke et al., 1977).
A. Direct and Filtered Direct Excitation
In Figure 2, the beam(1) represents the conguration used for direct excitation of the sample
by the x-rays emitted fromthe anode. Aprimary x-ray beamlter can be used to modify the
spectrum from the x-ray tube that is nally used to excite the elements in the sample.
The optimum selection of kV and primary beam lter are critically important for
obtaining the best data from an EDXRF system. As with all spectrometry methods, the
Copyright 2002 Marcel Dekker, Inc.
principal driver for best precision and detection limits is peak-to-background (P/B) ratio.
However, as mentioned earlier, the limited total counting capacity of the EDXRF system
and the analysis of multielement samples are added complications in deriving optimum
excitation conditions. In general, the kV governs the sensitivity and the primary beam
lter governs the background.
The energy distribution of the spectrum arriving at the sample governs the eec-
tiveness of excitation for element XRF lines. In order to excite x-ray uorescence, it is
necessary to have incident x-ray energies above the absorption edge for the elements line
series to be excited. In order to have adequate excitation, there needs to be a high intensity
of x-rays with energies higher than the edge and conventional wisdom has suggested that
the tube kV should be 1.52 times the absorption edge of interest. This overvoltage en-
sures that there is a substantial proportion of the x-ray tube output spectrum available for
excitation of the element line(s) of interest. Figure 3 shows the unltered x-ray tube
spectrum scattered from a thin-polymer-lm sample when the x-ray tube is operated at
5, 10, and 15 kV. The observed scatter spectrum is a reasonable representation of the
spectrum striking the sample and clearly shows the following:
Figure 2 An x-ray tube and typical excitation geometries. The high-energy electron beam strikes
the anode to produce x-rays. (1) Direct or direct filtered: the flux (1) is used directly to strike sample
(1) or is first passed through a primary beam x-ray filter. (2) Secondary target: the primary flux (1)
strikes a secondary target 2, which generates a flux of x-rays (2), which strikes the sample at 2.
(3) Transmission geometry: the beam (3) is directed through the anode to the sample.
Copyright 2002 Marcel Dekker, Inc.
The high-energy cuto at the operating kV. This is also known as the DuaneHunt
limit.
The bremsstrahlung hump.
The intense characteristic AgLlines (around 3 keV) fromthe tube anode. The intensity
of the AgL lines increases with the kV (i.e., as the overpotential increases).
The tube L lines of Ag and Rh are particularly eective for exciting the light elements in
the range 12.5 keV. Once higher-energy lines (e.g., FeK at 6.4 keV) are sought, the L
lines provide no excitation, as they are lower in energy than the absorption edge (FeK at
7.11 keV) and the bremsstrahlung hump provides the excitation. Despite the highest in-
tensity of the bremsstrahlung being lower than AgL lines, it is the integrated excitation
spectrum, higher in energy than the absorption edge, which is important. When the x-ray
tube kV exceeds the energy of the K edge of the tube anode material, the characteristic K
lines will begin to dominate the excitation spectrum. This is shown in Figure 6, which
depicts the scattered excitation spectrum with the tube operated at 35 kV, at which the
AgK characteristic lines are strongly excited.
The basic sensitivity for analysis is governed by the kV applied to the x-ray tube and
an example of this is seen in Figure 4. The sample was a pressed pellet and the kV was set
in the range 616 kV using a vacuum path in the spectrometer and no primary beam lter.
At each kV, the rhodium-target x-ray tube current was adjusted to give 50% deadtime.
The K-series line intensities were used and the sensitivity for each element has been nor-
malized to the value at 10 kV. From the plots, the following are clear:
The Si intensity falls o as the kV increases. This is due to the increasing distance of
the tube main spectrum from the SiK absorption edge and to the increased relative
contribution of the other elements to the total counts in the spectrum. This latter
Figure 3 Scattered, unfiltered excitation spectrum provided by a silver target x-ray tube operated
at (1) 5, (2) 10, and (3) 15 kV.
Copyright 2002 Marcel Dekker, Inc.
factor is important and derives from the limited total count rate for the whole
spectrum.
The FeK lines are not excited until at least 8 kV is applied. The FeK absorption edge
is at 7.111 keV and the maximum sensitivity for Fe is at approximately twice this
value.
The general rule of thumb in which the kV is set to twice the highest absorption-edge
energy is a reasonable place to start. However, as more elements are sought across a wider
energy range, there usually comes a point at which the balance of the spectrum is un-
favorable for the analytical problem being solved. When this happens, split the spectrum
into two or more areas of interest and apply a kV to best suit each of these element ranges.
In any event, as the kV and the number of elements increase, the importance of using a
primary beam lter and splitting the spectrum into optimized energy regions increases.
B. Primary BeamFilter Selection
A primary beam lter acts as an x-ray absorber and it is placed between the x-ray tube and
the sample to modify the x-ray tube output spectrum to which the sample is exposed. In
general, the kV is selected rst to ensure high excitation sensitivity and then an appropriate
lter is chosen. The lter acts to reduce scattered background in the region of interest and
to reduce the excited intensity of lower-energy peaks. The characteristics of the lter are
dened by its x-ray absorption curve, which is controlled by selecting the material and its
thickness. Filters are typical thin, pure metal foils of thickness in the range 10500 mm.
Some typical lters, the range of kV used with each, and the applicable optimum element
range (K lines) are shown in Table 1.
Figure 4 Effect of kV on sensitivity for the major elements in a geological sample (see text for
details).
Copyright 2002 Marcel Dekker, Inc.
The x-ray tube spectrum provided by unltered radiation can be seen in Figure 3.
The eect on the x-ray tube spectrum of applying a thin and thick aluminum lter is seen
in Figure 5, where a tube voltage of 15 kV is used.
The aluminum acts as a simple absorption lter, having a single absorption edge at
1.56 keV. From the plots, it is clear that the aluminum lters both completely absorb the
AgL lines as a consequence of the high-mass-absorption coecient of around 700 cm
2
=g.
The hump at lower energy provided by the spectrum ltered though thin Al yields good
P=B ratios for elements in the S to V range (2.35 keV). The absence of the AgL (or RhL,
MoL) lines from the tube removes their spectral interference and avoids the need for these
x-rays to consume counting capacity. The thick aluminum lter pinches the low-energy
side of the bremsstrahlung hump, providing high P=B ratios for the elements with lines in
the range 38 keV.
The combination of kV and an absorption lter provides an energy region in which
the P=B ratio is optimal. On the low-energy side of this region, the excitation is suppressed,
which allows more of the systems counting capacity to be used in the region itself. On the
high-energy side of the region, the excitation from the tube is tuned to provide a high
integrated intensity above the energy of the absorption edges of the element lines of in-
terest. The use of a thick copper lter with maximum kV is another example of this type of
lter. The copper lter used is generally thick enough to completely absorb the x-ray tube
anode K lines, which are of high energy (AgKa is 22.1 keV). Such an arrangement provides
for the eective determination of Ag and Cd using a silver anode x-ray tube. At the other
end of the energy scale, a cellulose lter is a weakly absorbing lter which suppresses the
AgL lines.
The so-called Regenerative Monochromatizing Filter (RMF) is a particularly im-
portant lter in tube-excited EDXRF spectrometry. The lter acts on the x-ray tube anode
K lines so the kV must be suciently high above the K absorption edge of the
anode material to generate intense K characteristic emission. The same element as the
anode material is used to preferentially transmit the characteristic K x-rays generated in
the anode. The lter takes its benet from the fact that an element has a low-mass-
Table1 Typical Primary Beam Filters and Their Range of Use
Filter Thickness kV range Elements Comments
No filter N=A 450 All, NaCa Optimum for light elements
with 48 kV excitation
a
Cellulose Single sheet 510 SiTi Suppresses tube L lines
a
Thin aluminum 2575 mm 812 SV Removes tube L lines
a
Thick aluminum 75200 mm 1020 CaCu Good for transition element
metal alloys
Thin anode element 2575 mm 2540 CaMo RMF
b
, wide range and good
trace analysis
Thick anode element 100150 mm 4050 CuMo RMF
b
, good for trace analysis
of heavy-element L lines.
Copper 200500 mm 50 >Fe Suppresses tube K lines
a
Needs to be used in conjunction with vacuum or He path to remove the air in the spectrometer that would
severely attenuate the low-energy lines and result in ArK lines in the detected spectrum.
b
Regenerative monochromatizing lter (RMF). The same element as the anode material is used as the lter
material to preferentially transmit the characteristic K x-rays generated in the anode.
Copyright 2002 Marcel Dekker, Inc.
absorption coecient for its own characteristic lines, as these lie immediately below the
absorption-edge jump. In the case of Ag, the AgKa
1,2
lines are at 22.1 keV and the
absorption coecient of Ag for this energy is only 14 cm
2
=g.
The eect of two silver primary lters, of dierent thickness, on the spectrum from a
silver anode x-ray tube operated at 35 kV is shown in Figure 6. The unltered 35-kV
excitation (curve 1) provides equally intense AgL and AgK characteristic lines and a broad
bremsstrahlung hump centered on 1215 keV. Any element in the 515 keV range will
have the benet of excitation by both the AgK characteristic lines and the bremsstrahlung.
However, there exists the disadvantage of sitting on scattered background derived from
the radiation in the bremsstrahlung hump. With the thin Ag lter in place (curve 2), the
AgL characteristic lines are completely absorbed and the bremsstrahlung hump is reduced
to a low-energy tail on the AgK characteristic lines. The resulting excitation yields high
P=B ratios for lines in the energy range 412 keV. Use of the thicker Ag lter (curve 3),
reduces even further the residual bremsstrahlung below the AgK lines and provides
pseudomonochromatic radiation, which yields high P=B ratios for lines in the energy range
515 keV. The strongly ltered AgK line excitation provides excellent sensitivity and
background reduction and a clear, intense Compton-scatter peak, which is good for trace
analysis where a matrix correction based on Compton scatter is used. The same can be said
for anode and lter materials of Mo, Rh, and Pd, which also provide characteristic K lines
in this important region of the spectrum.
A specic and well-documented example of optimum selection of x-ray tube anode,
excitation kV, and primary beam lter can be seen in work by Potts et al. (1986). In that
work, a Co anode x-ray tube and an iron primary beam lter were used to remove the
Figure 5 Scattered excitation spectrum provided by a silver target x-ray tube operated at 15 kV.
The plots show the effect of two thicknesses of an Al primary beam filter: (1) unfiltered; (2) thin Al
filter; (3) thick Al filter.
Copyright 2002 Marcel Dekker, Inc.
CoKb line which would otherwise strongly excite the Fe that typically dominates this part
of the spectrum in geological analysis.
Most modern EDXRF systems provide the ability to use more than one set of
conditions per analysis. Depending on the analytical needs, the kV and lter are chosen to
optimize the important parts of the spectrum and the data from spectra at each of these
conditions are consolidated to provide the best P=B values. In order to establish the op-
timum conditions and make the best use of analysis time, the best strategy is to select a
typical sample and then to make measurements with the dierent conditions available. Use
criteria such as detection limits to decide upon the optimum conditions for the required
analysis problem. Often, the instrument manufacturers recommendations can con-
veniently be taken as the starting point for any investigation.
C. SecondaryTarget Excitation
The basic geometry of secondary target excitation is shown in Figure 2. Use of this mode
of excitation for EDXRF spectrometry was rst described by Jaklevic et al. (1972), who
incorporated the secondary target inside a small, low-power x-ray tube. The high P=B
ratios that they reported were encouraging and Porter (1973) subsequently described the
rst work in which the secondary target was external to the x-ray tube. The primary ra-
diation from the x-ray tube strikes the secondary target, which then emits characteristic
radiation and generates some scattered bremsstrahlung radiation. The emitted
radiation is pseudomonochromatic and the degree of monochromaticity is further
enhanced, as in the case of the RMF in direct excitation, with a lter between the
secondary target and the sample. Filters are of the same material as the target and are
Figure 6 Scattered excitation spectrum provided by a silver target x-ray tube operated at 35 kV.
The plots show the effect of two thicknesses of Ag primary beam filter: (1) unfiltered; (2) thin Ag
filter; (3) thick Ag filter.
Copyright 2002 Marcel Dekker, Inc.
typically only used with the higher-Z targets (e.g., Mo, Ag, and Gd). With increasing
distance below the exciting lines, the absence of bremsstrahlung lower in energy than the
exciting lines results in a very rapid reduction in sensitivity. The provision of a number of
secondary targets with associated lters, which can usually be selected automatically,
provides a range of discrete and narrow excitation bands. This arrangement can be con-
veniently thought of as providing the equivalent of a number of transmission target x-ray
tubes, each with dierent anode materials. In order to generate intense characteristic ra-
diation from the secondary target, the tube kV should be two to three times the K ab-
sorption-edge energy of the element from which the secondary target is made. The K lines
are used, as they oer the highest uorescence yield, which means that there are lower
losses than with other line series. Typical secondary targets are, in decreasing energy order
of K line emission: Gd, Sn, Ag, Mo, Ge, Cu, Fe, Ti, and Al. The optimization of x-ray tube
kV, secondary target, and lter was described by Spatz and Lieser (1979).
The uorescence yield decreases as the target atomic number decreases and the ef-
ciency of x-ray production falls o rapidly below Ti. The result of this is that the lighter
elements Si to K are not well excited by secondary targets and there are few suitable
elements available to provide targets with K lines in this important region. One approach
is to use a reduced kV and a scattering target which scatters the x-ray tube L lines and
bremsstrahlung, which, as was seen for direct excitation, is eective in this energy region.
A stable polymer block provides a suitable target for this arrangement, although the
overall eciency is much lower than with direct excitation. There are two other ar-
rangements that eliminate this limitation. The rst is to use a second x-ray tube whose
geometry provides for direct, and perhaps direct ltered excitation. Such an arrangement
adds considerable cost and complexity but does oer optimum performance. It is possible
for both tubes to be run from the same high-voltage power supply. The second approach is
to provide mechanical repositioning of the x-ray tube such that it can be operated either in
secondary target or in direct and direct ltered mode. The positional repeatability of the
mechanical arrangement needs to be excellent if this approach is adopted. The mechanical
complexity, which inevitably adds cost and some possibility for unreliability, has not
stopped this approach from being used in the past with some success in one commercially
available instrument. One disadvantage of this approach is the time that is needed for the
mechanical switching, which can be a severe penalty if a number of acquisitions are re-
quired for a complete analysiswhich is usually the case. Very recently, a so-called Wide
Angle Geometry was described by Yokhin (2000), in which a single, xed x-ray tube was
used to deliver direct unltered or ltered and secondary target excitation. This ar-
rangement was made possible using the wide cone x-ray beam from the special x-ray tube
and a proprietary x-ray beam distributor.
The reduction in overall excitation eciency when using secondary targets is usually
oset by increasing the x-ray tube power available. A tube providing close geometric
coupling and power in the range 100400 W is typically used to oset the low overall
eciency of excitation. In order to eciently excite the GdK lines, it is preferable to use an
x-ray tube capable of being operated at 60 kV. A high-Z anode such as tungsten or gold
provides the highest bremsstrahlung intensity and, therefore, the best excitation eciency
for the higher-Z secondary targets.
The high degree of monochromatization in ltered secondary target systems pro-
vides for very high P=B ratios for elements with absorption edges immediately below the
target x-ray emission lines. The background in secondary target geometry can be further
reduced using 3-axis or Cartesian geometry as described by Standzenieks and Selin (1979),
Christensen et al. (1980), and Bisgard et al. (1981).
Copyright 2002 Marcel Dekker, Inc.
Despite secondary target excitation gaining wide acceptance through the early 1980s,
the use of direct and direct ltered excitation has remained as the most common excitation
arrangement. This is, perhaps, partly due to the restricted energy range oered by each
secondary target+lter combination and the general need for unltered direct excitation
to obtain the highest sensitivity for the light elements. In the 1990s, the use of polarized
EDXRF (see Chapter 10) tends to have supplanted secondary target EDXRF for the
provision of high P=B ratios in some applications. In addition, improved detection systems
have continued to improve the performance of instruments using direct ltered excitation.
An application in which secondary target systems have a valuable advantage is when the
secondary target lines are selected to be just below the absorption edge of a major element,
but just above the absorption edges of trace elements. One example of this is the use of a
copper secondary target for the determination of low levels of Mn and Fe in Cu alloys. In
this example, the CuK lines from the target do not excite the Cu but are highly eective for
exciting Fe. The detection and counting system is not dominated by CuK lines, neither is
the FeKa peak interfered with by the Si escape peak from CuKa. Such applications are
few in number and, in the case of the metals industry, analysts may often prefer to use
WDXRF spectrometers.
D. X-rayTubes
X-rays were rst discovered over a century ago (see Michette and Pfauntsch, 1996) by
Ro ntgen (1896a, 1896b), and since that time, they have become used in a great variety of
applications. These applications encompass medical and dental x-ray radiology and
imaging, industrial and baggage inspection, industrial nondestructive testing (NDT),
industrial thickness gauging, and x-ray spectrometry in all its guises. In most, if not all,
of these elds, the convenience of the high-voltage x-ray tube has substantial benets in
terms of convenience, safety, and availability of appropriate spectral and intensity output.
The modern x-ray tube used for EDXRF analysis is based on a design introduced by
Coolidge as long ago as 1913 (Coolidge, 1913). The basic design of x-ray tubes typically
used in EDXRF instrumentation consists of the following:
A radiation-shielded glass envelope
An independent tungsten lament with current control unit
A high-melting-point metallic anode
A high-voltage connection to the anode
A beryllium foil exit window
Some means of dissipating the heat generated
In order to operate, the x-ray tube requires a controllable, stable high-voltage power
supply capable of providing typically 560 kV. In addition, there is a low-voltage power
supply for providing current to heat the lament, thereby controlling the electron beam
current and output intensity. X-ray tubes used for EDXRF analysis are typically operated
with a power in the range of 1400 W. For direct excitation, the power output is typically,
no greater than 50 W. Reproducibility of analytical results is achieved for x-ray tube ex-
citation by operating for a measured amount of time at a xed emission current (dc op-
eration). This approach depends on the reasonable assumption that the total output x-ray
ux is proportional to the number of electrons striking the anode.
In some types of x-ray tube, a control grid can be used to modulate or pulse the
emission current. This can be used to modulate the electron beam current to ensure
constant output of radiation from the tube in dc operation or it can be used to rapidly
Copyright 2002 Marcel Dekker, Inc.
switch the electron beam on and o for pulsed operation. If the pulsing is synchronized
with x-ray detection, increased output countrate can be achieved by reducing pulse pileup
(see Sec. IV.C).
The key features of the x-ray tube used in EDXRF instruments are as follows:
Easily shielded and safe to operate
Compact and able to be placed close to the sample
Stable output over an extended period, typically months
Well-dened spectral output
High-voltage controllable in small increments, typically 1 kV, throughout the
operating range (typically 550 kV)
Beam current controllable in small increments, typically 10 mA, throughout the
operating range (typically 101000 mA).
The geometry and designs typically used and the characteristics of each are described in
the following sections. Further reading on the subject can be found in Skillicorn (1982,
1983).
1. Side-WindowGeometry
Side-window geometry is the oldest and, perhaps, the most common x-ray tube geometry
used in EDXRF instruments and it is suitable for both high-kV and high-power operation.
Figure 7 shows the essential details of this design.
The geometric coupling of tube anode spot to sample is limited rst by the distance
of the spot itself from the exit window and then, potentially, by the tube encapsulation
itself. Typically, these tubes are operated in the range 550 kV. They can dissipate 50 W
with the need only for forced-air cooling in order to keep the tube external body below a
temperature of, say, 55
35
from
normal to the incident electron beam and the exit window is usually made from beryllium
foil of thickness in the range 50250 mm.
The tube output spectrum can be measured or estimated and such information is
important in analytical correction procedures, particularly those using fundamental
parameter calculations for which accurate and computationally convenient tube output
spectrum models are required (see Chapter 2, Sec. I.B for further details.)
Copyright 2002 Marcel Dekker, Inc.
Figure 7 Side-window x-ray tube.
Copyright 2002 Marcel Dekker, Inc.
It should be remembered that the tube output spectrum, especially with higher-
power tubes, may change with time. One way in which this can occur is through deposition
of a thin tungsten lm on the inside of the tube output window through evaporation of
tungsten from the hot lament. Such deposition is unlikely to aect the low-power tubes
typically used in EDXRF systems and, in any event, is usually taken care of through the
normal restandardization process.
2. End-WindowGeometry
The design features of such an x-ray tube are shown in Figure 8. The same factors of
operation and tube spectrum apply to this type of tube as was applied to side-window
tubes. The tube spectrum is altered by the fact that the takeo angle is normal to the anode
surface and so there is minimal self-absorption by the target material. This high takeo
angle makes such a design particularly good for low-energy excitation. One of the im-
portant benets of the end-window design is the very tight coupling of anode spot to
sample that can be achieved with a sharp-nose design. The spot-to window coupling can
be close with suitable electron optics (e.g., 12 mm in the Varian EG50 design). The use of a
high positive voltage on the target minimizes the eect which repelled electrons would
have of causing an unacceptable amount of exit-window heating. This reduced heating
allows thinner beryllium windows to be used, which further improves the low-energy
x-ray output.
3. TransmissionTarget Geometry
The design features of such an x-ray tube are shown in Figure 9. In this type of x-ray tube,
the geometry is basically end window, but in this embodiment, there is a metal foil target
parallel to and inside the beryllium exit window. The incident high-voltage electron beam
Figure 8 End-window x-ray tube.
Copyright 2002 Marcel Dekker, Inc.
strikes and enters one side of the thin-lm target and the resulting radiation is heavily
ltered as it passes through the foil and exits the far side of the target. This novel design
was initially developed (Jaklevic et al., 1972) in order to reduce the bremsstrahlung from
the x-ray tube, which would then reduce the scattered background in the spectrum and
thereby improve detection limits. The initial design used a molybdenum transmission
target of 0.12 mm thickness and, when operated at 42 kV and up to 25 W, was shown to
give an output spectrum consisting principally of characteristic MoK radiation. The
output spectrum was comparable to an
125
I radioisotope source coupled to a molyb-
denum-foil secondary target.
This type of x-ray tube was made commercially available by WatkinsJohnson
(Hershyn, 1975a, 1975b). The spectrum output from this type of tube was investigated by
Zulliger and Stewart (1975), who calculated and measured spectra from Cr, Mo, and W
transmission target x-ray tubes. The targets were either thin (50 mm) metal foils or thin
lms (510 mm) on a beryllium substrate of 0.250.5 mm thickness. The thin targets mean
that only low-power tubes (<100 W typically) can be made successfully. These authors
found, as predicted, a substantial reduction in bremsstrahlung compared to the char-
acteristic lines and quoted detection limits for a number of applications. The Cr target was
found, not unsurprisingly, to be best for the light elements S and Ca. Although the energy
range covered was claimed to be wide, the intrinsic xed ltration of this design makes it a
less valuable general excitation source compared to a side- or end-window tube with an
external, selectable lter. A novel, multiple-target transmission tube has even been de-
scribed (Kis-Varga, 1988) in which the ve anode materials (Fe, Cu, Ge, Mo, and Ag) were
changed by a magnetic mechanism.
The intrinsic ltration of the low-energy bremsstrahlung makes this type of tube
much less suitable for excitation of low energy (i.e., <2-keV) x-ray lines. Consequently,
their usage is limited to specic applications where the intrinsic ltration and potential for
close geometric coupling of the tube anode spot to the sample are major benets. In fact,
in one design, the WatkinsJohnson tubes did not have a target closely coupled to the
exit window, which, when combined with the intrinsic lossess in the target, led to low
output ux from the tubes and, consequently, reduced sensitivity. Finally, these tubes are
not useful when operated at low voltages (Skillicorn, 1982) due to strong absorption in the
target. This makes them much less sensitive than solid-target, side- or end-window tubes
for selectively exciting the light elements in the presence of large amounts of heavier
elements.
Figure 9 Schematic of a transmission target x-ray tube.
Copyright 2002 Marcel Dekker, Inc.
III. SEMICONDUCTOR DETECTORS
In this type of detector, the total ionization produced by each x-ray photon striking the
detector is converted to a voltage signal with amplitude proportional to the incident en-
ergy. Specially designed preamplication and processing electronics are employed to
maintain the linearity of the voltage signal with respect to the original charge pulse.
A multichannel analyzer accumulates an energy spectrum of the sequential events in a
histogram memory. Because the energy analysis does not depend in any way on the dif-
fraction or focusing of the x-rays by the detector, the geometry of the system is relatively
insensitive to the placement of the detector with respect to the sample. An exception to this
is in polarized excitation XRF (Chapter 10), where the orientation angle can be used to
suppress scattered background. This insensitivity to geometry and ability to place the
detector element close to the sample provides for a large solid angle and high geometric
eciency. In addition, the mechanism by which the ionization signal is measured is not
restricted to a narrow energy region, thus allowing the simultaneous detection of x-rays
over a wide dynamic range of the emission spectrum. The principal advantages of EDXRF
derive from its capability for simultaneous detection, with high geometric eciency, of
characteristic x-rays from multiple elements.
A. Fabrication and Operating Principle
The energy-dispersive detecting element is based on the simple, semiconductor diode
structure shown in Figure 10. The example shown is a structure typical for lithium-drifted
silicon, Si(Li), detectors, although the basic elements are similar for high-purity
germanium (HPGe) and other semiconductor detectors. The device is fabricated on a
cylindrical wafer of high-quality semiconductor material with rectifying p or n contacts on
opposing end surfaces. The bulk of the material is characterized by a very low con-
centration of free-charge carriers. This reduced free-carrier concentration is achieved
either with the use of extremely pure material, in the case of HPGe detectors, or through
the charge compensation of p-type silicon with lithium donor atoms in the case of Si(Li)
detectors. A typical dimension for such a detector crystal is 1080 mm
2
active area with a
thickness of 35 mm.
In the geometry shown in Figure 10, the lithium-diused region acts as an n contact
and the metal surface barrier (typically an evaporated Au layer) serves as the p rectifying
contact. When the diode is reverse-biased, any remaining free carriers are swept out of the
bulk by the applied eld and an active depletion region is created. In this condition,
the only current that ows between electrodes on the respective contacts at each end of the
diode cylinder is due to thermally generated charge carriers, which are excited above the
narrow band gap of the semiconducting material. In order to minimize the noise from
these thermally generated charges, the detector crystal is operated at reduced temperature,
which is typically achieved using cryogenic cooling by means of a Dewar vacuum ask
lled with liquid nitrogen. Other cooling strategies can be adopted (see Sec. III.C).
When an x-ray photon is incident upon the active volume (i.e., depletion region) of
the diode, it normally interacts by photoelectric absorption to create an inner-shell va-
cancy in the semiconductor material together with an energetic photoelectron.
This photoelectron interacts with the atoms in the semiconductor crystal lattice to produce
multiple low-energy ionization events. This process continues until the electron comes to
rest at the end of a total range, which is short compared with the dimensions of the crystal.
The energy associated with the inner-shell vacancy is also absorbed in the crystal, in most
Copyright 2002 Marcel Dekker, Inc.
cases following the emission of Auger electrons or multiple low-energy x-rays and sub-
sequent re-absorption. The result of these multiple ionization process is the essentially
instantaneous production of a large number of free electronhole pairs in the sensitive
volume of the detector structure. These free-charge carriers form as a cloud from which
they are separated by the eld gradient. The front detector electrical contact, being biased
at 500 V, for example, attracts the hole and repels the electrons, the latter negative
charge cloud is swept to the rear contact and converted, by means of the FET, to an
amplied voltage pulse. The number of charge carriers produced is directly proportional
to the energy of the x-ray photon incident on the detector. Thus, the number of charges
collected results in a voltage pulse whose magnitude is, in turn, directly proportional to the
energy of the detected photon.
B. Crystal Materials
The characteristics of energy-dispersive semiconductor x-ray spectrometers derive pri-
marily from the properties of the materials used to make the device itself. Although the
most common material for semiconductor detectors in EDXRF is silicon, devices based on
the use of germanium or compound semiconductors such as mercury(II) iodide (HgI
2
),
CdTe, CdZnTe, or GaAs have also been used to varying degrees. In particular, mercury
(II) iodide detectors (Swierkowski et al., 1974; Slapa et al., 1976; Ponpon and Sieskind,
1996) have been used successfully in both research and commercial products for many
years. Early devices with adequate performance at room temperature were own on space
missions. More recently, devices have been used in hand-held EDXRF spectrometers
where the benet of ambient temperature operation, with or without thermoelectric
(Peltier) cooling of the eld-eect transistor (FET), have been successfully exploited.
Iwanczyk et al. (1996) compared the performance of HgI
2
, Si PIN photoiodide, and CdTe
detectors for this latter, important application. Large arrays of up to 100 HgI
2
detectors
Figure10 Cross section of a typical Si(Li) detector showing electrode contacts and active volume.
Copyright 2002 Marcel Dekker, Inc.
have also been fabricated (Iwanczyk et al., 1995; Patt et al., 1995). Dabrowski et al. (1981)
have been particularly successful in designing, fabricating, and characterizing HgI
2
de-
tectors throughout the past two decades and their work should be consulted for further
details of this important type of EDXRF detector. Although high-purity germanium
(HPGe) is widely used for detectors in electron microscopes as a result of its slightly better
resolution than Si(Li) detectors, the use of HPGe detectors is less popular in photon-
excited EDXRF spectrometry. One reason for this lack of acceptance may lie in the high
escape peak intensities (see Sec. III.E.1) which can strongly interfere in trace analysis.
Table 2 lists properties relevant to x-ray detection for Si, Ge, and HgI
2
, which are the
materials most commonly employed in commercial EDXRF spectrometers. A variety of
other compound semiconductor materials, such as GaAs, CdTe, and CdZnTe, have been
used for applications in which specic properties are particularly benecial. These prop-
erties include the following:
Higher atomic number for increased photoelectric cross section
Larger band gap for lower thermal leakage at room temperature
A lower band gap corresponding to a smaller average energy for creation of
holeelectron pairs (Knoll, 1979), which oers improved energy resolution.
Recalling the direct relationship described earlier between photon energy and number of
charge carriers and, therefore, the detectors signal resulting from detection of a particular
x-ray photon, one can use the data from Table 2 in the following example for detection of
CaKa
1
(KL3), which has an energy of 3.691 keV:
No. electronhole pairs produced 3691=3.86 956
Charge on electron 1.6610
19
C
Therefore,
Charge from CaKa
1
interaction 1.53610
16
C
Assuming a feedback capacitance in the rst stage amplication of 0.1 pF, the
resulting output voltage pulse is 1.53 mV.
From this, we can see the very small magnitude of the signals involved. This has led to very
great eorts having been made in the design of detector and signal processing systems to
minimize all sources of electronic noise which would otherwise interfere very strongly with
the signal being sought.
In addition to the conventional Si(Li) detector, Si PIN (Haselberger et al., 1996;
Cesareo et al., 1996) and Si drift detectors (Lechner et al., 1996; Bertuccio et al., 1996)
have also become available in recent years. The active area of these devices is generally
Table 2 Semiconductor Detector Material Properties
Detector properties Si Ge HgI
2
Atomic number 14 32 80.53
Atomic=molecular weight 28.09 72.60 454.45
Density (300 K) g=cm
3
2.33 5.33 6.3
Band gap (300 K) (eV) 1.115 0.665 2.13
Average energy E per electronhole pair (77 kV)
(eV)
3.86 2.96 4.15
a
Fano factor (77 K) 0.12 0.08
a
At 300 K.
Copyright 2002 Marcel Dekker, Inc.
310 mm
2
and the depletion thickness is of the order of 300 mm, which reduces detection
eciency (see Sec. III.F) compared to Si(Li) detectors.
In the following section, the details of detector operation are discussed, with parti-
cular emphasis on those aspects that impact EDXRF spectrometry most directly. These
include the detector energy resolution, detection eciency, spectral response, and counting
system throughput.
C. Cooling Systems
The need to minimize the electronic noise of the detection system is paramount for ob-
taining the best energy resolution. The main source of noise in the detector itself is leakage
current, which derives from the generation of charge carriers in the absence of x-rays
through, for example, thermal vibrations of the detector crystal lattice. Whatever the
cooling system used, the intention of it is to minimize leakage, and the lower the tem-
perature, the less leakage will occur. The boiling point of liquid nitrogen (77 K) is a typical
temperature used for the detector crystal, although the FET may require a dierent and
often higher temperature to minimize its noise contribution and, therefore, obtain opti-
mum detector resolution.
The need for reduced operating temperature results in the need for careful design in
the insulation from ambient temperatures of the detector crystal. Eective insulation is
typically achieved by enclosing the detector head (crystal and FET) assembly in a evac-
uated cryostat. In order for the low-energy x-rays (<2 keV) to reach the detector crystal,
the vacuum is usually retained by a thin (typically 550 mm) beryllium entrance window
(Fig. 11). In recent years, the thinnest windows (5, 8, and 12 mm) have become available
with coatings which ensure that there is no leakage of He through the Be window. Any
helium getting through the window would degrade the high vacuum within the cryostat
and lead to increased temperature, leakage current, and deteriorating performance.
The original and perhaps simplest way of cooling a detector unit is to attach the
measurement head via a cold-nger assembly to a dewar vessel containing liquid nitrogen
(LN). The dewar itself is typically a complex technological component if it is to provide
high performance and long operational lifetime. The modern LN cryostat delivers opti-
mum resolution and the ability to be cycled through multiple warm-up=cooldown cycles
without degradation in performance. The inevitable and natural boiling-o from the
dewar of LN over a period (1 L per day is typical) means that the dewar needs regular
replenishment. In order to minimize the frequency of replenishing LN, a large dewar is
preferable, although this desire for a long LN holding time may lead to a rather bulky
system. LN volumes of 715 L are typical, although some large systems may use volumes
up to 25 L and some small, hand-held units may use 1 L or less. In any event, the fol-
lowing are important:
Use only liquid nitrogen: Liquid air, although delivering the required temperature
range, will fractionate, yielding O
2
, which is an explosion hazard in the chemical
laboratory.
Ensure that there is no ice or dirt in the LN used. The presence of ice or dirt in the
dewar can cause microphony, which can degrade energy resolution. Ice and dirt
can be eliminated by pouring the LN through a funnel equipped with a wire
screen (e.g., a motor fuel metal lter funnel).
Observe safe LN handling procedures. Avoid overlling and spillages as liquid
nitrogen can cause embrittlement of polymers and cable insulation or result in
Copyright 2002 Marcel Dekker, Inc.
condensation from the atmosphere being deposited on components that are
sensitive to moisture.
It is essential that a Si(Li) detector be kept cold throughout the time that the bias is
applied. If this is not done, then there is the possibility of damage to the detector as-
sembly by, for example, migration of the drifted Li ions from the compensated region
of the crystal. Should this happen, the detector will fail and, for this reason, it is usual
for an automatic bias shuto system to be implemented. Such systems typically use a
temperature sensor or a LN level sensor to switch o the detector bias in case of the
dewar warming up. Other detectors, such as HPGe or high-purity Si, for which drifting
is not required, do not exhibit this particular problem but may still suer some pro-
blems if thermally cycled repeatedly, unless the cryostat assembly is specically designed
to do this.
The simplicity of the LN-lled dewar has led to its widespread use in EDXRFsystems.
However, the diculty in obtaining LN in remote areas, or the need for a no-LN solution,
has led to the introduction in the past decade or so of thermoelectric (Peltier eect) coolers or
mechanical refrigerating units. The mechanical units may employ the JouleThompson
(JT) eect or some other cycle such as the Stirling cycle to obtain essentially LN temper-
atures at the detector head. The major benet of this approach is that there is no dierence in
performance in terms of detector energy resolution from the conventional dewar LN sys-
tems, but the coolers come with penalties of increased complexity, cost, and bulk. HPGe
detectors, which typically need liquid-nitrogen temperature to operate, may use these me-
chanical coolers, but other detectors which will function at higher temperatures, albeit with
degraded energy resolution, are suitable for use with thermoelectric (Peltier eect) coolers.
Figure11 Cross section of a groove-type Si(Li) detector mounted in its cryostat end cap.
Copyright 2002 Marcel Dekker, Inc.
Peltier coolers can oer a very compact cryostat=detector assembly that requires
only electrical power to provide its cooling. Multistage (three to six) coolers may be used
to provide a DT cooling, from ambient, of 50
C. In an
ambient room temperature of 20
C to 90
DE
2
Det
DE
2
Elec
q
1
The DE
Det
component is determined by the statistics of the free-charge production process
occurring in the depleted volume of the diode. The average number of electronhole pairs
Copyright 2002 Marcel Dekker, Inc.
produced by an incident photon can be calculated as the total photon energy divided by the
mean energy required to produce a single electronhole pair (see Sec. III.B). If the uc-
tuation in this average were governed by Poisson statistics, the standard deviation would be
s
n
p
E
e
r
2
In semiconductor devices, the details of the energy-loss process are such that the individual
events are not strictly independent and a departure from Poisson behavior is observed.
This departure is taken into account by the inclusion of the Fano factor in the expression
for the detector contribution to the FWHM:
s
Fn
p
FE
e
r
3
Taking
sE
E
Fn
p
4
and, rearranging,
DE
Det
2:35
FeE
p
5
where e is the average energy required to produce a free electronhole pair, E is the energy
of the photon, F is the Fano factor (see Table 2), and the factor 2.35 converts the standard
deviation to FWHM for a Poisson distribution.
A typical value for the detector contribution to resolution, also known as the disper-
sion, is $120. Examination of the values of e and F listed in Table 2 shows that for an
Figure12 Pulseheight spectrum of MnK x-rays taken with a 30-mm
2
Si(Li) detector.
Copyright 2002 Marcel Dekker, Inc.
equivalent energy, the detector contribution to the resolution is 28% less for the case of Ge
compared to Si. This results in better energy resolution being obtained for HPGe detectors
and could be an important consideration in the choice of detector for certain applications.
However, this theoretical advantage in energy resolution of Ge over Si(Li) is mitigated
in many systems because of the dominance of electronic noise (e.g., from the FET). In ad-
dition, the very high GeK escape peak probability (see Sec. III.E.1) is likely to be a serious
limitation to the use of HPGe detectors in most EDXRF work. HPGe detectors are widely
used in electron microscopes, where energies <10 keV are likely to be of most interest.
E. SpectrumArtifacts
The simplicity of the element line spectra encountered in XRF spectrometry is one of its
major benets when compared with other atomic spectroscopy methods. However, a few
peaks can arise in the EDXRF spectrum from sources other than the elements in the
sample and it is wise to be aware of their nature. Unlike basic detector eciency and
spectrum background, which are dened by the selection of detector window and the
detector manufacturing process, the end user can exert some control over the eects of
spectrum artifacts. From the spectra presented so far, it is clear that the bremsstrahlung
continuum and the Compton (and Rayleigh)-scatter peaks arise from the interaction of the
exciting radiation with the sample. These features are not regarded as artifacts.
1. Escape Peaks
For an incident x-ray with energy higher than the SiK absorption edge, the detection
process will involve the generation, through x-ray uorescence of the detector material, of
SiK x-rays. The vast majority of these will themselves immediately be absorbed in the
detector volume and contribute to the overall charge collected for the original incident
x-ray photon. This is the normal mechanism of x-ray detection, which was described
earlier. There is, however, a nite probability that the SiKa x-rays produced will escape
from the detector volume and not contribute to the charge collected for the original
photon that was detected. The resulting energy detected will be reduced by 1.740 keV,
which is the energy of the SiKa x-ray that escaped. The greatest probability of escape arises
close to the front of the detector crystal from which SiK x-rays can more easily escape.
X-rays detected far within the volume will still generate SiK x-rays, but they will all be
absorbed before they can reach the outer surface of the crystal and escape. The probability
of generating the SiKb x-ray and of it escaping is extremely low and this eect can safely be
ignored. The result of the SiK escape process is a peak in the collected spectrum, which is
1.74 keV lower in energy than the parent peak. The Si escape peaks for TiKa and TiKb are
clearly seen, 1.74 keV below the TiK lines, in Figure 13 (indicated by 1 and 2, respectively)
(see also Table 3). In many detectors, the incomplete charge-collection tail on the low-
energy side of the parent peak will obscure the Si escape peak from TiKb.
The escape probability is highest for lines closest to the SiK absorption edge at
1.838 keV and may be calculated as described by Statham (1976) and Dyson (1974). The
intensity falls rapidly from around 3% of the parent peak at 23 keV to around 0.1% of
the parent peak at 10 keV. At energies above 10 keV, the eect is negligible, but below this
energy, it is important that spectrum-processing software takes account of the peak.
Typically, the processing software will make a correction in which the Si escape peak
is removed and the most comprehensive packages will reinstate the lost intensity to the
parent peak. If such packages are available, then make use of them before carrying out
Copyright 2002 Marcel Dekker, Inc.
qualitative analysis in order to avoid classic errors such as the Si escape peaks from ZnK
being identied as Co!
In the case of HPGe, the escape probability for GeL lines is negligible in XRF and
can be ignored. The escape of the GeK lines has, however, a very high probability, and for
Se, the escape peak intensity is around 16% of the parent peak. This is due to the com-
bination of the high detector eciency for GeK x-rays and the much greater depth within
the crystal from which the GeK lines can escape. There is further complication in that
there are many GeK lines and each of these lines will result in an escape peak. The
complexity and intensity of Ge escape peaks can easily be seen in the spectrum from a
molybdenum oxide sample, shown in Figure 14.
The principal GeK lines and their characteristic energies are as follows:
GeKa: 9.876 keV
GeKb
1
: 10.981 keV
GeKb
2
: 11.100 keV
The assignments of the numbered GeK escape peaks arising from the detection of the
MoK lines are listed in Table 4.
Figure13 Peak artifacts from a Ti spectrum collected by a Si(Li) detector (see Table 3 for details).
Table 3 Artifact Peak Energy and Assignments for Figure 13
Peak No. Energy (keV) Assignment
1 2.8 TiKaSiK escape
2 3.2 TiKbSiK escape
3 9.0 TiKa TiKa sum peak
4 9.4 TiKa TiKb sum peak
5 9.9 TiKbTiKb sum peak
Copyright 2002 Marcel Dekker, Inc.
The energy and intensity of these GeK escape peaks interferes very strongly over a
wide and important region of the spectrum. The interference is considerable from the GeK
escape peaks of Nb and Mo in stainless steels and corrections for these intense escape
peaks are not suciently accurate to retain the intrinsic performance of EDXRF for what
would be a reliable and straightforward analysis using a Si(Li) detector. The severe lim-
itation posed by the Ge escape peaks in EDXRF is also demonstrated in Figure 15 (see
also Table 5) for a typical determination of minor and trace heavy metals in a low-Z
matrix. In this case, the complex GeK escape peaks from the Rh backscatter peaks
completely dominate the important part of the spectrum for trace analysis.
In conclusion, the Si escape peak is small and easily corrected, but the number and
high intensity of the GeK escape peaks severely limit the value of the HPGe detector for
tube-excited EDXRF. In the case of energy-dispersive (ED) microanalysis, however,
Figure 14 GeK escape peaks from Mo spectrum collected by HPGe detector (see Table 4 for
details).
Table 4 GeK Escape Peak Assignments and Relative Intensity for Figure 14
Peak No. Energy (keV) Assignment % of parent
1 6.4 MoKaGeKb
1
1.5
2 7.57 MoKaGeKa 9.6
3 8.6 MoKbGeKb 1.0
4 9.74 MoKb
1
GeKa 8.8
5 10.12 MoKb
2
GeKa 2.5
Copyright 2002 Marcel Dekker, Inc.
the HPGe detector has benets in that the intensity of the GeL escape peaks is negligible
and the energy resolution is better than that of Si(Li) detectors.
2. SumPeaks
Sum peaks arise from a specic form of peak pileup (see Sec. IV.C) where two events from
high-intensity peaks arrive in the pulse processing electronics so close together in time that
the pileup inspector cannot recognize themas twoevents. The eect of this is for the signals to
be seen as one and for themto be registered at the energy that is the sumof the two. The sum
peaks of Ti are seen in Figure 13, in which the most intense peak (3) is from the sum of two
TiKa events. The smallest peak (5), fromthe sumof two TiKb events is only visible when the
parent peak is very intense. At rst glance, peaks 3 and 4 may appear like a typical Ka, Kb
Figure 15 Comparison of Si(Li) and HPGe detectors for tube-excited EDXRF trace analysis of
low-Z sample matrix (see Table 5 for details).
Table 5 GeK Escape Peak Assignments for Figure 15
Peak Assignment
1 RhKa ComptonGeKa
2 RhKa RayleighGeKa
3 RhKb ComptonGeKa
4 RhKb RayleighGeKa
Copyright 2002 Marcel Dekker, Inc.
doublet, but the peak separation is too small for Klines in the region of the spectrumin which
they occur. In addition, the relative line ratio is not the same as for a characteristic K line
series at this energy. Consequently, sum peaks are unlikely to be incorrectly identied as
elements, but they may interfere with important lines in a particular analysis. This is parti-
cularly the case in some environmental analyses, (e.g., where high concentrations of Fe lead
tointense FeKlines). This will result inFeKsumpeaks inthe regionof the Klines of Se andBr
and the L lines of toxic heavy elements such as Hg, Tl, and Pb.
Some spectrum-processing software packages are able to correct for sum peaks and
should be used to minimize potential errors in analysis (See Chapter 4). If it is found that
such features are not available, then reducing the count rate will substantially reduce the
eect of the sum peaks. In order not to lose analytical performance, however, selection of
alternate excitation conditions (e.g., using a more absorbing beam lter) will rebalance
the spectrum in favor of the peaks at higher energy and minimize the interference.
3. Diffraction Peaks
Whenever a crystalline sample is measured in an EDXRF spectrometer, there is the
possibility that the conditions for Bragg diraction will be met. This is exacerbated by the
divergent geometry of the incident and detected beam and the variety of excitation con-
ditions available. The worst cases are where unltered primary radiation is used, as there
are then many energies and angles which increase the probability of meeting the Bragg
condition for a crystalline component in the sample. The more monochromatic the ex-
citing beam is, the less likely it is that the Bragg condition will be met. Figure 16 shows the
diraction peak obtained from a boric acid pellet which was excited with a silver anode
x-ray tube operated at 5 kV with no primary beam lter and in an approximately 90
geometry.
The spectrum is overlaid with one from a sulfur pellet measured under exactly the
same conditions. It can be seen that the diraction peak is broader than a K series line at
that energy and the low-energy tailing is more pronounced. There is some possibility for
misidentication, but the most likely problem arises in any spectrum processing not setup
to take account of this peak. In cases such as silicon wafers, where the crystal planes are
highly oriented, the sample can be tilted or rotated to minimize the often sharp diraction
peaks that can be observed. Alternatively, eective use can be made of additional colli-
mation to constrain more closely the angles that can satisfy the Bragg condition or of
alternate excitation conditions, which change or reduce the incident energies available for
diraction. Figure 17 shows a comparison of the spectra taken from the same two samples
under dierent excitation conditions.
The increased kV and use of an absorption lter completely change the energy
distribution of the excitation spectrum and eliminate the diraction peak. This second set
of conditions is particularly eective for lines in this part of the spectrum and eectively
removes problems from diraction peaks. Once a diraction peak is identied as causing a
problem, a change of excitation conditions is often the best measure to adopt. The use of
monochromatic or polarized excitation or of secondary targets in Cartesian geometry will
also eliminate diraction peaks.
4. SystemContamination Peaks
Careful screening of all the components in an EDXRF spectrometer is required in order to
eliminate spurious element peaks in measured spectra. Each spectrometer designer will
adopt dierent approaches and materials for this purpose and, thus, there is a variety of
Copyright 2002 Marcel Dekker, Inc.
Figure 16 Spectrum from boric acid, overlaid with one from a sulfur pellet, showing the
diffraction peak. Conditions: Ag target x-ray tube, 5 kV, no primary beam filter.
Figure 17 Spectrum from boric acid, overlaid with one from a sulfur pellet, showing removal
of the diffraction peak seen in Figure 16. Conditions: Ag target x-ray tube, 15 kV, thin Al primary
beam filter.
Copyright 2002 Marcel Dekker, Inc.
potential spectrum contaminant peaks. Careful design will eliminate them, although, in
some cases, the use of a simple blank correction may be required in some trace analysis.
Detector components can be a cause of some system peaks. Beryllium entrance
windows usually contain some trace elements that, in extreme cases such as TXRF, can be
seen in blank spectra. The use of thick beryllium windows exacerbates this potential
problem, but window thicknesses commonly used (850 mm) should pose no practical
problem. The Si internal uorescence peak and absorption edge caused by the dead layer
are eects not seen in practice and need very careful characterization with specialized
conditions if they are to be seen at all.
The most likely source of spurious spectrum peaks is sample handling and pre-
sentation, both of which are largely in the control of the user.
F. Detection Efficiency and EntranceWindows
One of the more important advantages of semiconductor spectrometers is the absolute
eciency with which uorescence x-rays are detected and their energies measured. This is
the result of the intrinsically high photoelectric absorption eciencies for semiconductor
materials in the x-ray energy range and from the large solid angles achieved in typical
EDXRF geometries. The solid angle is determined by the area of the detector and the
sampledetector distance and varies with the design of the system. Typical areas are
1080 mm
2
for Si(Li) detectors. Although the additional solid angle is advantageous for
many applications, one must realize that the added capacitance associated with increased
active area results in an increased contribution in the electronic noise of the system. This,
in turn, results in an increased energy resolution of the detection system.
The intrinsic eciency of the semiconductor device can be approximated by a simple
model in which the probability of detecting an x-ray incident on the detector is assumed to
be the probability of photoelectric absorption within the sensitive volume. This can be
expressed as
eE e
mt
1 e
sd
where e(E) is the energy-dependent intrinsic eciency of the detector, t is the thickness of
any absorptive layer between the sample and the detector, mm(E) is the mass absorption
coecient of the absorptive layer, d is the detector thickness, and ss(E) is the photo-
electric mass-absorption coecient for the detector material. Figure 18 shows plots of the
above equation for the case of a 3-mm-thick Si(Li) and a 5-mm-thick HPGe detector.
The poor eciency at low energies is assumed to be determined by the combined
absorption of the 25-mm beryllium entrance window and a nominal 2-cm air pair between
sample and detector. The energies of the characteristic K emission lines for several
elements are also shown. The plots show the near-unity intrinsic eciencies for both
detectors over a wide range of useful x-ray energies. The Ge detector is ecient at much
higher energies than Si because of the higher atomic number and subsequent larger
photoelectric cross section.
The simple model gives a semi-quantitative picture of the eciency behavior of
semiconductor spectrometers. However, several other factors must be considered for
quantitative calibration of a uorescence spectrometer. The concept of a thin entrance
window that either absorbs or transmits an incident x-ray does not describe cases in which
the secondary electrons, either photoelectric or Auger, are emitted into the active volume
from the window layer. Similarly, detailed studies of the low-energy eciency have in-
dicated the presence of an absorbing layer on the surface of both Si and Ge devices
Copyright 2002 Marcel Dekker, Inc.
associated with the surface layer of the semiconductor material itself. This layer is com-
prised of an evaporated metal contact used to form the rectifying Schottky barrier and a
thin surface layer (the dead layer) of inactive semiconductor material from which charge
cannot be collected. The thickness of evaporated metal, normally Au, can be determined by
direct measurement during the manufacturing process and is typically in the range of 10 nm.
The eective dead layer of semiconductor material is a more complicated parameter to
determine. Empirical studies have attempted to measure the absorption inherent in the
semiconductor surface layer. Such studies established that the eective thickness is de-
termined largely by the absorption length of the low-energy photons and the charge-
transport characteristics of the associated ionization products. In the simplest model, there
is competition from the rate of diusion of the electronhole distribution, against the
gradient of the applied eld. For distances near the entry contact, a part of the charge can
diuse into the contact and be lost to the signal before the electrical eld can sweep it to the
opposing electrode. This loss of charge can be interpreted in terms of an eective window
thickness of typically 0.2 mm silicon equivalent. In addition, thin evaporated contacts are
deposited on the entry surface that can absorb incident low-energy photons.
Detailed experiments have indicated that the charge-collection eciency for events
near the surface depends in a complex way on low-energy x-ray and charge-transport
properties (Llacer et al., 1977; Goulding, 1977). As a consequence of these entry-window-
related eects, the eciency for low-energy photons can be reduced. The events lost from
the photopeak can then appear in a continuum background below the full-energy peak,
where they reduce detectability and interfere with spectral analysis. Similarly, secondary
electrons or photons originating from the photoelectric absorption events that occur in-
itially in the active volume can escape. This results in a collected charge pulse of reduced
Figure 18 Calculated intrinsic photopeak efficiency for Si and Ge detectors as a function of
incident photon energy. The high-energy limits are established by the photoelectric cross section of
the detector material and the diode thickness. The low-energy cutoff is determined either by
absorption in the thin Be window or in a nominal 2-cm air path from sample to detector or
fluorescence source.
Copyright 2002 Marcel Dekker, Inc.
amplitude. An easily observed manifestation involves the observation of discrete peaks in
the spectrum. Continuous-loss processes involving electron escape can also occur, al-
though the probability is small. These and related mechanisms that reduce the amplitude
for a given event can lower the photopeak eciency relative to the simple model described.
There have been studies in which the eciencies of Si(Li) and Ge have been carefully
measured using calibrated sources of x-rays spanning the energy region of interest (Cohen,
1980; Szoghy et al., 1981; Campbell and McGhee, 1986). These indicate that the maximum
intrinsic photopeak eciency is reduced by a few percent relative to the curves shown in
Figure 18 and is slightly higher at the upper energy cut o than calculated. These results
substantiate the overall validity of the simple photoelectric absorption model but de-
monstrate the limitations if precise results are required. For laboratory applications of
EDXRF, it is not necessary to explicitly determine the photoelectric eciency function
because it is included in the overall calibration factor of the instrument. In any event, the
internal eciency of the detector itself is dened by the design and manufacture of the
detector and this is out of the control of most users.
The eciency curve shown in Figure 18 illustrates that for conventional EDXRF
measurements, the absorption of uorescence x-rays in air and in the Be entry window
limits the accessible energy range to photons greater than approximately 2 keV. The ab-
sorption losses in the air path can be signicantly reduced by the common practice of
employing a vacuum or He atmosphere. A vacuum path is preferable due to higher x-ray
transmission, and if helium is used, the beryllium entrance window must be able to
eliminate leakage into the cryostat with subsequent loss of its vacuum. Coated beryllium
windows are commonly used nowadays to minimize helium leakage into the cryostat.
The necessity of a beryllium window to maintain vacuum integrity between the
cryostat enclosure and atmospheric pressure does not constitute a serious limitation for
most analyses. However, there are applications in which the detection of x-rays with en-
ergies below 1 keV (e.g., for F) becomes necessary. The relatively high detection limits
achieved by EDXRF systems because of the poor energy resolution (compared with
WDXRF) combined with absorption by the beryllium window generally make EDXRF
the second choice for these very low-energy x-ray lines. A beryllium window thickness of
8 mm is typically used for very light-element analysis, although windows as thin as 5 mm
have been used. The use of extremely thin windows may lead to increasing concerns for
helium integrity and equipment cost.
High-strength thin windows made from low-atomic-number elements that are cap-
able of withstanding a full 1-atm-pressure dierential have become commercially available
in the last decade. These include self-supporting 0.5-mm diamond polycrystalline lms
and 0.25-mm windows composed of a vapor-deposited amorphous material consisting of
90% boron by weight with nitrogen and oxygen for the remainder. These windows
exhibit signicant x-ray transmission for photons well below 1 keV and have found
widespread use in x-ray microanalysis.
G. Detector Background
Eects associated with the partial collection of the photoelectric signal by the detector
have a small eect on its eciency. However, these incomplete charge-collection processes
can have consequences that are far more serious for analytical performance through their
eect on spectral background. The eect of incomplete charge collection of these events is
to produce tailing on the low-energy side of a peak. In addition, it produces a continuum
of events that appear as a shelf on the low-energy side of major peaks in the spectrum. This
Copyright 2002 Marcel Dekker, Inc.
background tailing and shelf interfere with the measurement of lower-energy uorescence
x-rays and is a signicant factor in obtaining the lowest detection limits for the determi-
nation of trace quantities on the low-energy side of major peaks.
Studies designed to reduce this background have indicated that in addition to
fundamental x-ray and electron-energy-loss processes, a more signicant background
resulting from incomplete charge collection from the detector active volume is normally
the dominant contribution (Goulding et al., 1972; Jaklevic and Goulding, 1972). This
process is an artifact of the detector operation in which the collection of the free charge
from the depleted volume is inhibited as a result of nonuniformities in the applied
electrical eld. These nonuniformities are typically associated with edge eects at the
periphery of the cylindrical detector. The incomplete charge-collection background can
be reduced either by external collimators that prevent the incident radiation from in-
teracting in the periphery of the detector or by internal electronic collimation brought
about by the use of a guard-ring structure (Goulding et al., 1972). The tailing can
sometimes be improved by increasing the detector bias, but this is not an option
available to the majority of users. There has been much work on characterizing the
detector line shape and the eects on background and tailing of incomplete charge
collection (Campbell and Wang, 1991; Campbell et al., 1997; Heckel and Scholz, 1987;
Lepy et al., 1997). Such studies have led to a greater understanding of the processes
involved and to peak-tting procedures that can take account of them. However, im-
provements in this form of background can only be achieved through improvements in
detector design and manufacturing processes.
The P=B ratio of a detector is a standard measurement of detector quality. The
measurement is made using a
55
Fe radioisotope source to ensure that there is no other
contribution to background. The measurement cannot be made using x-ray tube excita-
tion, as it is extremely dicult to eliminate the residual bremsstrahlung continuum. The
peak is taken from the most intense channel in the MnKa peak and the background is
taken as the average intensity in the channels from 0.9 to 1.1 keV. A high-quality Si(Li)
detector will deliver a P=B ratio in excess of 10,000:1 and have minimal low-energy tailing.
IV. SEMICONDUCTOR DETECTOR ELECTRONICS
A. Sources of Electronic Noise
The contribution to resolution associated with electronic noise (DE
Elec
) is the result of
random uctuations in thermally generated leakage currents within the detector itself and
in the early stages of the amplier components. Although these processes are intrinsic to
the overall measurement process, there are methods for limiting the impact on the nal
system resolution. Figure 19 is a schematic diagram of a typical pulse processing system
employed in a semiconductor detector x-ray spectrometer. The pulse processing can be
divided between the charge integration, which takes place in the preamplier, and the
voltage amplication and pulse shaping, which occur in the main amplier (pulse
processor).
The function of the charge-sensitive preamplier and subsequent amplication stages
is to convert the integrated charge pulse, produced by collection of the photoelectrically
induced ionization, to a voltage pulse that can be measured and stored in the multichannel
pulseheight analyzer (MCA). The rst stage of the process occurs in the FET, which is
attached to the rear of the detector crystal. The method of charge restoration used for the
FET has a fundamental eect on the electronic noise of the overall detection system and is
Copyright 2002 Marcel Dekker, Inc.
particularly important. Early detectors used pulsed optical charge restoration, but newer
systems make use of integrated solid-state devices incorporating low-noise-junction FETs
to minimize electronic noise (Statham and Nashashibi, 1988; Lund et al., 1995, Lund et al.,
1996). However, the pulse processing must also achieve a very important goal of ampli-
fying the weak charge signal to a measurable level while suppressing random uctuations
in the signal amplitude produced by thermal noise. This is achieved by generating a
carefully dened pulse shape in the main amplier, which restricts the Fourier frequency
components in the nal signal in such a way that the signal contributions are emphasized
relative to the noise uctuations.
The most common pulse shapers employed in modern semiconductor spectro-
meters generate output pulses that are Gaussian, triangular, or cusp shaped (Fairstein
and Hahn, 1965; Kandiah et al., 1972; Landis et al., 1982). Each is capable of achieving
adequate energy resolution for EDXRF analysis; the dierences are due mainly to the
eective time interval required for processing a pulse. Because the relative amplitude of
the noise contribution is a strong function of the characteristic time constant associated
with the pulse shaper, the dierence between pulse shapes becomes important for ap-
plications in which high counting rates are important. A detailed analysis of the eects
of various pulse-shaping options on spectrometer performance is presented by Goulding
and Landis (1982). More recently, suciently fast, sampling analog-to-digital converters
(ADCs) have become available and these have resulted in the availability of digital pulse
processors (Warburton et al., 1988). These devices can implement the optimum noise-
reduction lters in conjunction with much higher count rates than conventional analog
pulse processors.
As pointed out below in the following subsection, however, it is not always desirable
to operate at the minimum-noise shaping time because an unacceptably large dead time
may result. Although EDXRF instruments are designed with a suitable compromise be-
tween energy resolution and count rate capabilities, an understanding of these trade-os
on the part of the user is important for optimum usage of the instrument.
Figure19 Typical pulse processing system used for energy-dispersive detectors.
Copyright 2002 Marcel Dekker, Inc.
B. Resolution and Count Rate
There exists a great deal of literature on the nature of the detection and pulse processing
systems and the great strides which have been made over the past 20 years to improve
resolution and increase count rate. For many years, the complete integration of the de-
tection and electronics systems within the instrumentation has meant that the individual
choice of and matching by the EDXRF user of detector, main amplier, ADC unit, and
MCA is no longer required. Once the application is determined, the choice of system will
typically revolve around achievable analytical performance.
The user will typically be left with a relatively straightforward selection of the single
signal processing setting providing the required resolution and count rate
(see Fig. 20). There is only a single setting and, unfortunately, resolution and rate go in
opposite directions when this setting is changed. In one direction, the amplier time
constant will increase, which gives better resolution (lower FWHM) at the expense of a
lower count rate into the MCA memory. In the opposite direction, resolution degrades but
count rate increases. Once the EDXRF system is chosen, this control is really the only way
in which the user can inuence resolution and count rate.
It is a little like choosing a car where you select the type of vehicle most suited to
your use. That vehicle then has a gearing selection system that allows you some degree of
trade-o in control. Low gears give highest accuracy in maneuverability (i.e., resolution)
but lowest speed (i.e., count rate). Conversely, the highest gears give a high speed (rate)
at the expense of accurate maneuverability!
At one end of the range, best resolution is obtained with lowest total count rate,
whereas at the other, highest count rate is delivered with the worst energy resolution.
When severe peak overlap is the dominant limitation to analytical performance, the best
resolution should be selected. In this case, which is typical for the light elements, the
Figure 20 Plots of energy resolution (at 5.9 keV) and count rate for a Si(Li) detector.
Copyright 2002 Marcel Dekker, Inc.
excitation conditions will need adjusting to make best use of the limited count rate
available. Where peak separation is good, such as above 15 keV, or where there are no
severe overlaps, select the fastest count rate. In both cases, the current is used to adjust for
optimum dead time (50% typically) and the measurement time is used to control the
precision.
C. Pulse Pileup
The count-rate limitations associated with a semiconductor spectrometer are an inherent
property associated with the nite pulse processing time required by the electronic shaping
network. When a random sequence of pulses is incident on the detection system, some of
the events cannot be processed without ambiguity. To appreciate the fundamental nature
of this limitation and its relationship to system performance, some elementary concepts of
electronic pulse processing must be considered.
Figure 21 illustrates the time sequence of pulses that occur at various stages in the
pulse processing chain. Trace A shows the output of the charge-integrating preamplier.
The steps at times 1, 2, and 3 represent the charge integrals of discrete events. Traces B and
C are the outputs of the fast (short shaping time) amplier and an associated discriminator
that serves as a timing marker for the individual events. The main shaping amplier output
is shown as the Gaussian pulse shapes (in trace D). For each event, a total pulse processing
time (neglecting analog-to-digital conversion in the pulseheight analyzer) of t
d
is re-
quired after the arrival of the pulse and before the system is ready to accept the next event.
Although the average counting rate detected by the system can be well below the
frequency dened by the reciprocal of the pulse width t
d
, that the events are statistically
uncorrelated implies that the events are not uniformly distributed in time. There then
exists a probability that two pulses will occur within the same processing time interval.
This is illustrated by the overlap of pulses 2 and 3 in the trace.
At a low average counting rate, this pulse overlap is not a limiting factor. As the
average counting rate increases, however, a point is reached at which there is a signicant
probability that a second event will occur before the rst event has been fully analyzed.
If the two events occur within a time less than the shaping time of the amplier, the charge
signals are indistinguishable and an erroneous pileup energy signal results.
Modern spectrometers employ some form of rejection circuitry to eliminate these
pileup events from the pulseheight spectrum. Typical systems rely on the inspection of
the fast discriminator output to determine if two pulses have occurred in rapid succes-
sion. Appropriate logic is then employed to gate the output of the processor to eliminate
the resultant ambiguous energy signals. This is shown in traces E and F of Figure 21, in
which the logic signal causes the output to be inhibited when the pulses overlapped to
produce an ambiguous pileup energy output. Because the fast shaper that generates the
discriminator output has inherently poorer energy resolution relative to the slow channel,
the pileup rejection circuit is limited in the minimum energy of pulses it can detect.
However, in typical tube-excited EDXRF, the major fraction of pulses occur from
scattered high-energy radiation and a pileup rejector system is an important feature in
the system design.
The pileup probability is obviously a function of the characteristic shaping time t
p
,
which, in turn, establishes the eective dead time, t
d
, of the system. This probability is
independent of the details of the specic type of pileup rejection circuit used to eliminate
the ambiguous events. The number of events that experience pileup and are consequently
eliminated from the spectrum can be estimated for the case of a totally random arrival
Copyright 2002 Marcel Dekker, Inc.
distribution. For a series of events randomly distributed in time with an average frequency
N
0
, the probability P that no pileup events occur within a characteristic time t
d
after a
given event can be expressed as
P N
0
e
N
0
t
d
From this expression, we can calculate the fraction of events transmitted through the
system. Figure 22 is a plot of nonpileup output rate versus input rate expressed in terms of
a characteristic pulse processing time.
The input rate for which the output rate is a maximum is seen to be the reciprocal of
the shaping time. The output rate at this point is reduced by a factor of 1=e. It should be
emphasized that this behavior is a fundamental consequence of random arrival statistics
and a nite measurement time. However, one should be aware that it is always possible to
reduce the characteristic processing time to achieve an increase in counting rate at some
sacrice in energy resolution. It is also worth noting that, by working at too high a dead
Figure21 Time sequence of pulses during the processing of a series of x-ray events; (A) the output
of the preamplifier is represented as a series of voltage steps reflecting the integrated charge from the
individual events; (B) the output from a fast shaping amplifier that operates with a shorter time
constant than the main shaping system; (C) a fast discriminator timing pulse derived from B; (D) the
output of the main shaper, with the pileup of pulses 2 and 3 indicated; (E) the dead-time gates; and
(F) the final output with the ambiguous 2 and 3 pileup pulses rejected.
Copyright 2002 Marcel Dekker, Inc.
time, the throughput curve inverts and the output rate begins to fall. Further increase
of the dead time will eventually lead to paralysis of the system and negligible output
countrate.
Although dierent algorithms can be used to correct for events lost to pileup, there
is no way to eliminate the eect through passive pulse processing (Gedcke, 1972; Hayes
et al., 1978; Statham, 1977). Because dierent manufacturers of x-ray equipment vary in
their approaches to pileup rejection, the only way to evaluate the throughput of the system
is using a variable intensity source of radiation. The input rate can be determined by
scaling the fast discriminator output, which is sometimes available from the instrument
data system. The output rate can be simultaneously measured in the pulseheight ana-
lyser. A plot of nonpileup output rate versus input rate can then be generated and com-
pared to the ideal case shown in Figure 22.
When the excitation source can be switched o in a time interval that is short com-
pared to the characteristic pulse processing time, it is possible to increase the average output
counting rate by eliminating the eects of pileup. Such pulsed excitation systems rely on the
ability to detect an event in the fast channel and to shut o the excitation before a second
pileup event can occur. In this mode of operation, the output rate can equal the input rate
up to the point at which the system is continuously busy. This method of pileup control has
been implemented using pulsed x-ray tubes (Jaklevic et al., 1976; Stewart et al., 1976),
although the increased complexity of the electron optics tends to lead to increased cost.
D. Dead-Time Correction
The presence of pulse pileup must be considered in any system designed for quantitative
measurements because it causes the eciency with which pulses are processed to be rate
dependent. EDXRF analytical systems are designed to correct for this discrepancy
through a variety of approaches. The most straightforward involve the use of a live-time
Figure 22 Nonpileup output rate (N
0
) as a function of input rate (N
1
) scaled as a function of
a characteristic shaping time. Maximum output is 0.37 of the input rate that is equal to the reciprocal
of the time constant.
Copyright 2002 Marcel Dekker, Inc.
clock consisting of a gated oscillator and scaler. The oscillator clock is turned o when the
system is busy processing pulses. In this way, the duration of the measurement in terms of
a live-time interval is extended to correct for those periods when the system is incapable of
processing a pulse. Alternative methods rely on the direct measurement of the fast dis-
criminator pulses to keep track of those events that were missed. An additional correction
is added to compensate for those intervals in which pileup occurs (Bloomeld et al., 1983;
Hayes et al., 1978).
A simple empirical method used to correct for dead-time losses is a measurement of
the ratio of input to output counts as a function of input rate over the range of values
normally encountered. Subsequent analyses require that the input rate be measured for
each experiment. The live-time correction can then be applied using the previously de-
termined response function.
Because all live-time correction methods have some limitation on the range of
counting rates over which they can be used, it is important that methods be devised to
evaluate the precision with which such corrections are generated. A carefully prepared
series of standards of varying concentrations represents a direct approach. A potentially
more precise method involves the use of a single thin-lm standard. Variable-mass targets
are placed behind the standard to vary the total counting rate over the range of interest.
If the variable mass target is chosen so that the variable intensity of scattered or uor-
escence radiation does not induce uorescence in the thin-lm standard, the measured
intensity of uorescence from the standard should be independent of total counting rate.
The ability of the dead-time correction system to compensate for pileup can then be
empirically evaluated. All modern EDXRF systems incorporate eective dead-time cor-
rection circuitry, which is likely to be at least as accurate as experiments designed to test
this standard feature.
V. SUMMARY
The technology that makes chemical analysis with EDXRF practical is based on the use of
semiconductor x-ray detectors and associated pulse processing and data acquisition sys-
tems. The present chapter attempted to explain to the analyst the basic concepts behind
the operation of these components and the manner in which they inuence overall system
performance. The trade-os one must make between such parameters as excitation con-
ditions, detector resolution, count rate, and other design variables determine how eec-
tively one can tailor a given instrument or experimental apparatus to a specic application.
Furthermore, a thorough understanding of the factors that limit performance should
enable one to implement experimental tests to determine the eectiveness of a particular
approach and evaluate various commercial options.
REFERENCES
Aitken DW. IEEE Trans Nucl Sci NS-15:10, 1968.
Bertolini G, Cappelani F, Restelli G. Nucl Instrum Methods 32:86, 1965.
Bertuccio G, Fasoli L, Fiorini C, Gatti E, Longoni A, Sampietro M, Hau D, Kemmer J, Richter R.
Nucl Instrum Methods 377:352, 1996.
Bisgard KM, Laursen J, Nielsen BS. X-Ray Spectrom 10:17, 1981.
Bloomeld DJ, Love G, Scott VD. X-Ray Spectrom 12:2, 1983.
Copyright 2002 Marcel Dekker, Inc.
Bowman HR, Hyde EK, Thompson DG, Jared RC. Science 151:562, 1966.
Campbell JL, McGhee PL. Nucl Instrum Methods Phys Res A248:393, 1986.
Campbell JL, Wang JX. X-Ray Spectrom 20:191, 1991.
Campbell JL, Maxwell JA, Papp T, White G. X-Ray Spectrom 26:223, 1997.
Cesareo R, Gigante GE, Canegallo P, Castellano A, Iwanczyk JS, Dabrowski A. Nuclear Instrum
Methods 380:440, 1996.
Christensen LH, Rasmussen SE, Pind N, Henriksen K. Anal Chim Acta 116:7, 1980.
Cohen DD. Nucl Instrum Methods 178:481, 1980.
Coolidge WD. Phys Rev 2:409, 1913.
Dabrowski AJ, Singh M, Huth GC, Iwanczyk JS. In: Heinrich KFJ, Newbury DE, Myklebust RL,
Fiori CE, eds. Energy Dispersive X-ray Spectrometry Special Publication 304. Gaithersburg,
MD: National Bureau of Standards, 1981, p 45.
Dyson NA. Nucl Instrum Methods 114:131, 1974.
Elad E, Nakamura M. Nucl Instrum Methods 41:161, 1966.
Fairstein E, Hahn J. Nucleonics 23:50, 1965.
Field KM, Anderson RJ. US Patent 08=277, 768, 1993.
Frankel RS, Aitken DW. Appl Spectrosc 24:557, 1970.
Gedcke DA. X-ray Spectrom 2:129, 1972.
Gedcke DA, Elad E, Denee PB. X-Ray Spectrom 6:21, 1977.
Goulding FS, Jaklevic JM, Jarrett BV, Landis DA. Adv X-ray Anal 15:470, 1972.
Goulding FS. Nucl Instrum Methods 142:213, 1977.
Goulding FS, Landis DA. IEEE Trans Nucl Sci NS-29:1125, 1982.
Haselberger N, Markowicz A, Valkovic V. Appl Radiat Isot 47:455, 1996.
Hayes JM, Matthews DE, Schoeller DA. Anal Chem 50:25, 1978.
Heath RL. Adv X-ray Anal 15:1, 1972.
Heckel J, Scholz W. X-Ray Spectrom 16:181, 1987.
Hershyn W. Research=Development 26:32, 1975a.
Hershyn W. Watkins Johnson Tech-Notes 2:1, 1975b.
Iwanczyk JS, Patt BE, Wang YJ, Croft M, Kalman Z, Mayo W. IEEE Trans Nucl Sci
NS-42:606, 1995.
Iwanczyk JS, Patt BE, Wang YJ, Khusainov AK. Nucl Instrum Methods 380:186, 1996.
Jaklevic JM, Giauque RD, Malone DF, Searles WL. Adv X-ray Anal 15:266, 1972.
Jaklevic JM, Goulding FS. IEEE Trans Nucl Sci NS-19:384, 1972.
Jaklevic JM, Landis DA, Goulding FS. Adv X-ray Anal 19:253, 1976.
Kandiah K, Smith AJ, White G. IEEE Trans Nucl Sci NS-22:2058, 1972.
Kirkpatrick P. Rev Sci Instrum 10:186, 1939.
Kirkpatrick P. Rev Sci Instrum 15:223, 1944.
Kis-Varga M. Isotopenpraxis 24:435, 1988.
Knoll GF. Radiation Detectors and Measurements. Wiley, New York: 1979.
Landis DA, Goulding FS, Pehl RH, Walton JT. IEEE Trans Nucl Sci NS-18:115, 1971.
Landis DA, Cork CP, Madden NW, Goulding FS. IEEE Trans Nucl Sci NS-29:619, 1982.
Lechner P, Eckbauer S, Hartman R, Krisch S, Hau D, Richter R, Soltau H, Strueder L, Fiorini C.
Nucl Instrum Methods 377:346, 1996.
Lepy MC, Plagnard J, Stemmler P, Ban G, Beck L, Dhez P. X-Ray Spectrom 26:195, 1997.
Llacer J, Haller EE, Cordi RC. IEEE Trans Nucl Sci NS-24:53, 1977.
Lund JC, Olschner F, Bennett P, Rehn L. IEEE Trans Nucl Sci NS-42:820, 1995.
Lund MW, Decker KW, Perkins RT, Phillips JD. Nucl Instrum Methods Phys Res A380:318, 1996.
Michette A, Pfauntsch S. X-rays: The First Hundred Years. Chichester: Wiley, 1996.
Patt BE, Iwanczyk JS, Szczebiot R, Maculewicz G, Wang M, Wang YJ, Hedman B, Hodgson KO,
Cox AD. IEEE Trans Nucl Sci NS-42:558, 1995.
Ponpon JP, Sieskind M. Nucl Instrum Methods 380:173, 1996.
Porter DE. X-Ray Spectrom 2:85, 1973.
Potts PJ, Webb PC, Watson JS. J Anal Atomic Spectrom 1:467, 1986.
Copyright 2002 Marcel Dekker, Inc.
Ro ntgen WC. Science 3:227, 1896a.
Ro ntgen WC. Science 3:726, 1896b.
Ross PA. J Opt Soc Am 16:433, 1928.
Sandborg A, Shen R. EDAX Editor 14:8, 1984.
Skillicorn B. Adv X-ray Anal 25:49, 1982.
Skillicorn B. Kevex Analyst 5:2, 1983.
Slapa M, Huth GC, Seibt M, Randtke PT. IEEE Trans Nucl Sci NS-23:101, 1976.
Spatz R, Lieser KH. X-Ray Spectrom 8:110, 1979.
Standzenieks P, Selin E. Nucl Instrum Methods 165:63, 1979.
Statham PJ. J Phys E 9:1023, 1976.
Statham PJ. X-Ray Spectrom 6:95, 1977.
Statham PJ, Nashashibi T. Microbeam Anal 50, 1988.
Stewart JE, Zulliger HR, Drummond WE. Adv X-ray Anal 19:153, 1976.
Swierkowski SP, Armantrout GA, Wichner R. IEEE Trans Nucl Sci NS-21:302, 1974.
Szoghy IM, Simon J, Kish L. X-Ray Spectrom 10:168, 1981.
Zulliger HR, Stewart JE. Adv X-ray Anal 18:278, 1975.
Vane RA, Stewart WD. Adv X-ray Anal 23:231, 1980.
Warburton WK, Darknell DA, Hubbard-Nelson B. Mater Res Soc Symp Proc 487:559, 1998.
Yokhin B. Adv X-ray Anal 42:11, 2000.
Copyright 2002 Marcel Dekker, Inc.
4
Spectrum Evaluation
Piet Van Espen
University of Antwerp, Antwerp, Belgium
I. INTRODUCTION
This chapter deals with (mathematical) procedures to extract relevant information from
acquired x-ray spectra. Smoothing of the spectrum results in a graph that can be easier
interpreted by the human observer. To determine which elements are present in a speci-
men, peak search methods are used. To obtain the analytically important net peak areas of
the uorescence lines, a variety of methods, ranging from simple summation to sophisti-
cated least-squares-tting procedures, are to the disposal of the spectroscopist.
Spectrum evaluation is a crucial step in x-ray spectrometry, as much as sample pre-
paration and quantication. As with any analytical technique, the performance of x-ray
uorescence analysis is determined by the weakest step in the process. Spectrum evaluation
in energy-dispersive x-ray uorescence analysis (EDXRF) is certainly more critical than in
wavelength-dispersive spectrometry (WDXRF) because of the relatively low resolution of
the solid-state detectors employed. The often-quoted inferior accuracy of EDXRF can, to a
large part, be attributed to errors associated with the evaluation of these spectra. As a
consequence of this, most of the published work in this eld deals with ED spectrometry.
Although rate meters and=or strip-chart recorders have been employed in WD
spectrometry for a long time, the processing of ED spectra by means of computers has
always been more evident because of their inherent digital nature. Some of the techniques
to be discussed have their roots in g-ray spectrometry developed mainly in the sixties; for
others (notably the spectrum-tting procedures), EDXRF has developed its own specia-
lized data processing methodology. The availability of inexpensive and fast personal
computers together with the implementation of mature spectrum evaluation packages on
these machines has brought sophisticated spectrum evaluation within the reach of each
x-ray spectrometry laboratory.
In this chapter, various methods for spectrumevaluation are discussed, with emphasis
on energy-dispersive x-ray spectra. Most of the methods are relevant for x-ray uorescence,
particle-induced x-ray emission (PIXE), and electron beamx-ray analysis [electron probe x-
ray microanalysis (EPXMA), scanning electron microscopy energy dispersive x-ray
analysis (SEMEDX), and analytical electron microscopy (AEM)]. The principles of the
methods and their practical use are discussed. Least-squares tting, which is of importance
not only for spectrum evaluation but also for qualication procedures, is discussed in detail
in Sec. IX. Section X presents computer implementations of the main algorithms.
Copyright 2002 Marcel Dekker, Inc.
II. FUNDAMENTAL ASPECTS
The aims of spectrum evaluation is to extract analytically relevant information from ex-
perimental spectra. Obtaining this information is not straightforward because the spectral
data are always corrupted with measurement noise.
A. Amplitude and Energy Noise
In x-ray spectra, we can distinguish between amplitude and energy noise. Amplitude noise
is due to the statistical nature of the counting process, in which random events (the arrival
of x-ray photons at the detector) are observed during a nite time interval. For such a
process, the probability of observing N counts when the true number of counts is N
0
is
given by the Poisson distribution (Bevington and Robinson, 1992):
P(N; N
0
) =
N
N
0
N!
e
N
0
(1)
The number of counts in each channel of an x-ray spectrum as well as the sum over a
number of channels obey this Poisson distribution. For a Poisson random variable, the
population standard deviation is equal to the square root of the true number of counts:
s
N
0
=
N
0
_
(2)
The sample standard deviation, which is an estimate of the true standard deviation,
therefore can be calculated as the square root of the observed number of counts:
s
N
=
N
_
- s
N
0
(3)
The statistical nature of the counting process (Poisson statistics or counting statistics)
causes the typical channel to channel uctuations observed in x-ray spectra. That the
uncertainty of the data can be calculated from the data itself [Eq. (3)] is of great im-
portance for the spectrum evaluation methods.
Energy noise, on the other hand, causes the characteristic x-ray lines of ED spectra
to appear much wider than their natural linewidth of about 510 eV. Part of this line
broadening is due to the nature of photon-to-charge conversion process in the detector,
and part of it is associated with the electronic noise in the pulse amplication and pro-
cessing circuit, as discussed in Chapter 3. As a result, x-ray photons with energy E, which
on average correspond to a pulse height stored in channel i, from time to time gives rise to
slightly higher or lower pulses, causing the x-ray events to be stored in channels above and
below i, respectively. Accordingly, characteristic x-ray lines appear as relatively broad
(140250 eV), nearly Gaussian-shaped peaks in the spectrum. The peaks observed in
wavelength-dispersive spectra are also wider than the natural linewidth because of im-
perfections in the diraction crystal and the nite size of the collimators.
B. Information Content of a Spectrum
In the absence of these two noise contributions, spectrum evaluation would be trivial. A
spectrum would consist of a well-dened continuum on which sharp characteristic lines
were superimposed. The intensity of the continuum and the net x-ray line intensities could
be determined without error. Any remaining peak overlap (e.g., between AsKa and PbLa,
where the separation of 8 eV is less than the natural width of the AsKa line) could be dealt
with in an exact manner.
Copyright 2002 Marcel Dekker, Inc.
Unfortunately, we cannot eliminate the noise in the measurements completely. It is
possible, however, to reduce the noise in various ways. Amplitude noise (i.e., counting
statistics) can be reduced by acquiring the spectrumfor a longer period of time or by using a
more intense primary beam. The eect of energy noise can be lowered by using a detector
and associated electronics of good quality and by shielding the systemfromexternal sources
of electronic noise. Although these suggestions may sound straightforward and not ap-
propriate in the context of spectrum evaluation, it is important to realize that once a
spectrum has been acquired, the information content remains constant. No spectrum pro-
cessing procedure, no matter howsophisticated, can produce more information than present
originally. It is therefore much more ecient to employ optimal experimental conditions
when acquiring the data rather than to rely on mathematical techniques in an attempt to
obtain information which is not present in the rst place (Statham and Nashashibi, 1988).
From this point of view, spectrum processing can be seen as any (mathematical)
procedure that transforms the information content of a measured spectruminto a formthat
is more useful for our purposes (i.e., more accessible). As is indicated in Figure 1, most of
the procedures that calculate this useful information require some form of additional
input. Sometimes, this extra information is intuitive and not clearly dened; in other cases,
additional information is used in the form of mathematical model. In this respect, not the
complexity of model but rather the ability to accurately describe the physical reality is of
relevance. When we use a procedure to estimate the net peak area of a characteristic line by
summing the appropriate channels and interpolating the continuum left and right of the
peak, explicitly some additional information is given to this spectrum processing
method in the form of peak and continuum boundaries. In addition (implicitly), a certain
mathematical model is assumed regarding the shape of the peak and the continuum.
Provided this model is correct, this relatively simple procedure returns a correct estimate of
the net peak area as good as one could obtain by a complicated tting procedure.
The important distinction between simple and more sophisticated spectrum eva-
luation procedures lies in the incorporated exibility and applicability. A simple area es-
timation based on the integration of peak and continuum regions is not generally
applicable in real-world situations in which peak overlap, curvature in the continuum, and
peak tailing occur. Procedures employing more complex models are adaptable to each
specic situation, yielding reliable peak estimates. Spectrum evaluation procedures should
Figure 1 Spectrum evaluation seen as an information process: spectrum processing requires
additional information to extract useful information from measured spectral data.
Copyright 2002 Marcel Dekker, Inc.
therefore be compared on the basis of the explicit and implicit assumptions that are made
in the model(s) they employ.
C. Components of an X-ray Spectrum
To evaluate an x-ray spectrum correctly, it is necessary to understand all the phenomena
that contribute to the nal appearance of the spectrum. This includes the two main fea-
tures, characteristic lines and continuum, and also a number of spectral artifacts, which
become important especially in trace analysis work (Van Espen et al., 1980).
1. Characteristic Lines
The characteristic radiation of a particular x-ray line has a Lorentz distribution. Peak
proles observed with a semiconductor detector are the convolution of this Lorentz dis-
tribution with nearly Gaussian detector response function, giving rise to what is known as
Voigt prole (Wilkinson, 1971). Because the Lorentz width is of the order of only 10 eV
for elements with atomic number below 50, whereas the width of the detector response
function is of the order of 160 eV, a Gauss function is an adequate approximation of the
line prole. Only for K lines of elements such as U and Th does the Lorentz contribution
become signicant and need to be taken into account (Gunnink, 1977).
A more close inspection of the peak shape reveals some distinct tailing at the low
energy side of the peak and a shelf extending to zero energy. This is mainly due to
incomplete charge collection caused by detector imperfections (dead layer and regions of
low electric eld) as discussed in Chapter 3. The eect is most pronounced for low
energy x-rays. For photons above 15 keV Compton scatter in the detector also con-
tributes to deviation from the Gaussian shape. The distortion caused by incomplete
charge collection has been described theoretically (Joy, 1985; Heckel and Scholz, 1987).
Various functions have been proposed to model the real peak shape more accurately
(Campbell et al., 1987).
The observed emission spectrum of an element is the result of many transitions, as
explained in Chapter 1. The resulting x-ray lines, including possible satellite lines, need to
be considered in the evaluation of an x-ray spectrum. A more detailed discussion on the
representation of the K and L spectra and the peak shape is given in Sec. VII.
2. Continuum
The continuum observed in x-ray spectra results from a variety of processes. The con-
tinuum in electron-induced x-ray spectra is almost completely due to the retardation of the
primary electrons (bremsstrahlung). The intensity distribution of the continuum radiation
emitted by the sample is in rst approximation given by Kramers formula (Chapter 1).
Absorption in the detector windows and in the sample causes this continuous decreasing
function to fall o at low energies, giving rise to the typical continuum shape observed.
The attenuation of this bremsstrahlung by major elements in the sample also causes ab-
sorption edges to appear in these spectra. Continuum modeling for electron-induced x-ray
spectra has been studied in detail by a number of authors (Statham, 1976a, 1976b,
Smith et al., 1975; Sherry and Vander Sande, 1977; Bloomeld and Love, 1985).
For particle-induced x-ray emission, a similar continuum is observed also mainly due
to secondary electron bremsstrahlung. Other (nuclear) processes contribute here, making
it virtually impossible to derive a real physical model for the continuum. Special absorbers
placed between the sample and detector further alter the shape of the continuum.
Copyright 2002 Marcel Dekker, Inc.
In x-ray uorescence, the main source of the continuumis the coherent and incoherent
scattering of the excitation radiation by the sample. The shape can therefore become very
complex and depends both on the initial shape of the excitation spectrumand on the sample
composition. When white radiation is used for the excitation, the continuum is mainly ra-
diative and absorption edges can also be observed. With quasimonoenergetic excitation
(secondary target, radioisotope), the incomplete charge collection of the intense coherently
and incoherently scattered peaks is responsible for most of the continuum (see Chapter 3).
Also here, realistic physical models for the description of the continuum are not used.
The incomplete charge collection of intense uorescence lines in the spectrum further
complicates the continuum. The cumulative eect of the incomplete charge collection of all
lines causes the apparent continuum at lower energies to be signicantly higher that ex-
pected on the basis of the primary continuum generating processes.
3. Escape Peaks
Escape peaks result from the escape of SiK or GeK photons from the detector after
photoelectric absorption of the impinging x-ray photon near the edge regions of the de-
tector. The energy deposited in the detector by the incoming x-ray is diminished with the
energy of the escaping SiK or GeK photon. Typical examples of the interference due to Si
escape peaks are the interference of TiKa (4.51 keV) by the FeKa escape at 4.65 keV and
the interference of FeKa by the CuKa escape.
For a Si(Li) detector, the escape peaks is expected 1.742 keV (SiKa) below the parent
peak. Experimentally, it is observed that the energy dierence is slightly but signicantly
higher, 1.750 keV (Van Espen et al., 1980). Ge escape peaks are observed 9.876 (GeKa)
and 10.984 keV (GeKb) below the parent peak. The width of the escape peaks is smaller
than the width of the parent peak and corresponds to the spectrometer resolution at the
energy of the escape peak.
The escape fraction f is dened as the number of counts in the escape peak N
e
di-
vided by the number of detected counts (escape parent). Assuming normal incidence to
the detector and escape only from the front surface, the following formula can be derived
for the escape fraction (Reed and Ware, 1972):
f =
N
e
N
p
N
e
=
1
2
o
K
1
1
r
_ _
1
m
K
m
I
ln 1
m
I
m
K
_ _ _ _
(4)
where m
I
and m
K
are the mass-attenuation coecient of silicon for the impinging and the
SiK radiation, respectively, o
K
is the K uorescence yield, and r is the K jump ratio of
silicon. Using 0.047 for the uorescence yield and 10.8 for the jump ratio, the calculated
escape fraction is in very good agreement with the experimentally determined values for
impinging photons up to 15 keV (Van Espen et al., 1980). Equation (4) is also applicable
for estimating the escape fraction in Ge detectors, provided that the parameters related to
Si are substituted with these of Ge. Knowing the energy, width, and intensity of the escape
peak, corrections for its presence can be made in a straightforward manner.
4. Pileupand SumPeaks
With modern pulse processing electronics, the pileup eects are suppressed to a large
extend. With a pulse-pair resolution time of a few 100 ns or less, only true sum peaks are
observed. These sum peaks are, within a few electron volts, located at their expected
position, and they are slightly wider (5%) than normal peaks located at the same energy
in the spectrum (Van Espen et al., 1980). The count rate of a sum peak is given by
Copyright 2002 Marcel Dekker, Inc.
_
N
11
= t
_
N
1
_
N
2
(5)
and
_
N
12
= 2t
_
N
1
_
N
1
(6)
with
_
N
11
the count rate (counts=s) in a sum peak due to the coincidence of two x-rays with
the same energy,
_
N
12
the count rate of a sum peak resulting from two x-rays with dierent
energy, and t the pulse-pair resolution time in seconds. Sum peaks are often found when a
few large peaks at lower energy dominate the spectrum. Typical examples are PIXE
spectra of biological and geological samples. The high count rate of the K and CaK lines
produces sum peaks that are easily observed in the high-energy region of the spectrum
where the continuum is low. It is important to note that the intensity of sum peaks is
count-rate dependent, they can be reduced and virtually eliminated by performing the
measurement with a lower primary beam intensity. A method for correcting for the
contribution of sum peaks in least-squares tting has been proposed by Johansson (1982)
and is discussed further in Sec. VII.
5. Discrete Character of a PulseHeight Spectrum
Another aspect of spectral data that should be mentioned is that a pulseheight spectrum
is a discrete histogram representing a continuous function. Digitization of this continuous
function, especially the Gaussian peaks, into too few channels causes considerable sys-
tematic errors. If the peak contains less than 2.5 channels at the FWHM, the peak area
estimate, for example, obtained by summing the channel contents is largely overestimated.
This lower limit of 2.5 channels at FWHM corresponds to a spectrometer gain of
60 eV=channel for a peak width of 150 eV. In practice, 40 eV=channel or lower is re-
commended, otherwise peak position and width determinations and the results of spec-
trum tting become unreliable.
6. Other Artifacts
A number of other features might appear in an x-ray spectrum and can cause problems
during the spectrum evaluation. In the K x-ray spectra of elements with atomic number
between 20 and 40, one can detect a peaklike structure with a rather poorly dened
maximum and a slowly declining tail (Van Espen et al., 1979a). This structure is due to the
KLL radiative Auger transition, which is an alternative decay mode of the K vacancy.
The maximum is observed at the energy of the KLL Auger electron transition energy.
The intensity of the radiative Auger structure varies from approximately 1% of the Ka line
for elements below Ca to 0.1% for elements above Zn. For chlorine and lower atomic
number elements, the radiative Auger band overlaps with the Ka peak. In most analytical
applications, this eect will not cause serious problems. The structure can be considered as
part of non-Gaussian peak tail.
The scattering of the excitation radiation in x-ray uorescence is responsible for most
of the continuum observed in the spectrum. When characteristic lines are present in the
excitation spectrum, two peaks might be observed. The Rayleigh (coherently)-scattered
peak has a position and width as expected for a normal uorescence line. The Compton
(incoherently scattered) peak is shifted to lower energies according to the well-known
Compton formula and is much broader than a normal characteristic line at that energy.
This broader structure, resulting from scattering over a range of angles and Doppler eect,
Copyright 2002 Marcel Dekker, Inc.
is dicult to model analytically (Van Dyck and Van Grieken, 1983). The structure is
further complicated by multiple scattering phenomena in the sample (Vincze et al., 1999).
Apart fromthese commonly encountered scattering processes, it is possible to detect x-
ray Raman scattering (Van Espen et al., 1979b). Again, a bandlike structure is observed with
an energy maximumgiven by the incident photon energy minus the electron-binding energy.
The Raman eect is most prominently present when exciting elements with atomic number
Z72 to Z7 with the Kradiation of element Z. In this case, Raman scattering occurs on L
electrons. For x-ray excitation energies between 15 and 25 keV, the Raman scattering on the
K electrons of Al to Cl can also be observed. Because of its high-energy edge, the eect may
appear as a peak in the spectrumwith possible erroneous identication as a uorescence line.
The intensity of the Raman band increases as the incident photon energy comes closer to the
binding energy of the electron. The observed intensity can amount to as much as 10%of the
L uorescence intensity for the elements Rh to Cs when excitation with MoKx-rays is used.
When the excitation source is highly collimated diraction peaks can be observed in
the x-ray uorescence spectrum of crystalline materials. It is often dicult to deal with
these diraction patterns. They can interfere with the uorescence lines or even be mis-
interpreted as being uorescence lines, giving rise to false identication of elements.
III. SPECTRUMPROCESSINGMETHODS
Spectrum processing refers to mathematical techniques that alter the outlook of the
spectral data. This is often done, using some digital lter, to reduce the noise, to locate
peaks, or to suppress the continuum. In this section, various methods of ltering are
discussed. Because of its relation to the frequency domain, the concept of Fourier trans-
formation is introduced rst.
A. Fourier Transformation, Convolution, and Deconvolution
One can think of an x-ray spectrum as consisting of a number of components with dif-
ferent frequencies. In the spectrum shown in Figure 2, one recognizes a nearly constant
component (the continuum) as well as a component that uctuates from channel to
channel ( =fast). The latter is obviously the noise due to counting statistics. Peaks, then,
must have frequency components intermediate between these two. The frequency char-
acteristics of a spectrum can be studied in the Fourier space.
For any discrete function f(x), x=0, . . . , n 71 (e.g., a pulseheight spectrum), the
discrete Fourier transform is dened as
F(u) =
1
n
n1
x=0
f(x) exp
j2pux
n
_ _
=
1
n
n1
x=0
f(x) cos 2p
ux
n
j sin 2p
ux
n
_ _
(7)
with j =
1
_
and u =0, . . . , n 71. F(u) is a complex number. The real part, R(u), and the
imaginary part, I(u), represent respectively the amplitude of the cosine and the sine
functions that are necessary to describe the original data. The square of F(u) is called the
power spectrum,
[F(u)[
2
= R
2
(u) I
2
(u) (8)
Copyright 2002 Marcel Dekker, Inc.
and gives an idea about the dominant frequencies in the spectrum. Because there are n
dierent nonzero real and imaginary coecients, no information is lost by the Fourier
transform and the inverse transformation is always possible:
f(x) =
n1
u=0
F(u) exp
j2pux
n
_ _
(9)
Figure 3 shows that power spectrum of the pulseheight distribution in Figure 2 (a single
Gaussian on a constant continuum). The frequency (inverse channel number) is dened as
u=n, with n =256 and u =0, . . . , n=2. The amplitude of the zero frequency [F(0)[
2
, which is
equal to the average of the spectrum, is not shown. The dominating low frequencies
originate from the continuum and from the Gaussian peak, whereas the higher frequencies
are caused mainly by the counting statistics. It is clear that if we eliminate those high
frequencies, we are reducing this noise. This can be done by multiplying the Fourier
transform with a suitable function:
G(u) = F(u)H(u) (10)
An example of such a function is a high-frequency cuto lter:
H(u) =
1; u _ u
crit
0; u > u
crit
_
(11)
which sets the real and imaginary coecients above a frequency u
crit
to zero. If we apply
this lter to the Fourier transform of Figure 3 using u
crit
=0.05 and then apply the inverse
Fourier transformation [Eq. (9)], the result as shown by the solid line in Figure 2 is ob-
tained. The peak shape is preserved, but most of the statistical uctuations are eliminated.
Figure 2 A 256-channel pulseheight spectrum (single Gaussian on a constant continuum) and
the Fourier-filtered spectrum.
Copyright 2002 Marcel Dekker, Inc.
If we would cut o at even lower frequencies, peak distortions at the top and at the base of
the peak would become more pronounced.
This Fourier ltering can also be done directly in the original data space. Indeed, the
convolution theorem says that multiplication in the Fourier space is equivalent to con-
volution in the original space:
G(u) = F(u)H(u) =f(x) h(x) = g(x) (12)
The convolute at data point x is dened as the sum of the products of the original data and
the lter centered around point x:
g(x) = f(x) h(x) =
x
/
f(x x
/
)h(x
/
) (13)
h(x) is called a digital lter and is the inverse Fourier transformation of H(u). In
general, this convolution or ltering of a spectrum y
i
with some weighing function is
expressed as
y
+
i
=
1
N
j=m
j=m
h
j
y
ij
(14)
where h
j
are the convolution integers and N is a suitable normalization factor. The lter
width is given by 2m1. Fourier ltering with the purpose to reduce or eliminate some
(high or low) frequency components in the spectrum can thus be implemented as a con-
volution of the original data with a digital lter. This convolution also alters the variance
of the original data. Applying the concept of error propagation, one nds that the variance
of the convoluted data is given by
Figure 3 Fourier power spectrum of the pulseheight distribution shown in Figure 2.
Copyright 2002 Marcel Dekker, Inc.
s
2
y
+
i
=
1
N
2
m
j=m
h
2
i
y
ij
(15)
when the original data follows a Poisson distribution (s
2
y
= y).
Because the measured spectrum y(x) is itself a convolution of the original (x-ray
emission) signal f(x), with the instrument (or detector) response function h(x), it is, in
principle, possible to restore the measured signal if this response function is know. This
can be accomplished by dividing the Fourier transform (FT) of the measured spectrum by
Fourier transform of the (nearly Gaussian) response function, followed by the inverse
Fourier transform (IFT) of the resulting quotient:
y(x)
FT
Y(u)
h(x)
FT
H(u)
_
Y(u)
H(u)
= F(u)
IFT
f(x) (16)
The fact that the detector response function changes with energy (becomes broader) and,
more importantly, the presence of noise prohibits the straightforward application of this
Fourier deconvolution technique. Indeed, in the presence of noise, the measured signal
must be presented by
y(x) = f(x) h(x) n(x) (17)
and its Fourier transform
Y(u) = F(u)H(u) N(u) (18)
or
Y(u)
H(u)
= F(u)
N(u)
H(u)
(19)
At high frequencies, the response, H(u), goes to zero while N(u) is still signicant, so that
the noise is emphasized in the inverse transformation. This clearly shows that the noise
(counting statistics) is the ultimate limitation for any spectrum processing and analysis
method.
A clear introduction to Fourier transformations related to signal processing can be
found in the work of Massart et al. (1998). Algorithms for Fourier transformation and
related topics are given in the work of Press et al. (1998). Detailed discussions on Fourier
deconvolution can be found in many textbooks (Jansson, 1984; Brook and Wynne, 1988).
Fourier deconvolution in x-ray spectrometry based on maximum a posteriori or maximum
entropy principles is discussed by several authors (Schwalbe and Trussell, 1981; Nunez
et al., 1988; Gertner et al., 1989). Gertner implemented this method for the analysis of real
x-ray spectra and compared the results with those obtained by simple peak tting. The
problem that the deconvolution algorithms are limited to systems exhibiting translational
invariance was overcome by a transformation of the spectrum, so that the resolution
becomes independent of the energy.
B. Smoothing
Because of the uncertainty
y
_
on each channel content y
i
, ctitious maxima can occur
both on the continuum and on the slope of the characteristic peaks. Removal or sup-
pression of these uctuations can be useful during the visual inspection of spectra (e.g., for
locating small peaks on a noisy continuum) and is used in most automatic peak search and
Copyright 2002 Marcel Dekker, Inc.
continuum estimation procedures. Although smoothing can be useful in qualitative ana-
lysis, its use is not recommended as part of any quantitative spectrum evaluation.
Smoothing, although reducing the uncertainty in the data locally, redistributes the original
channel content over the neighboring channel, thus introducing distortion in the spectrum.
Accordingly, smoothing can provide a (small) improvement in the statistical precision
obtainable with simple peak integration but is of no advantage when used with least-
squares-tting procedures in which assumptions about the peak shapes are made.
1. Moving Average
The most straightforward way of smoothing (any) uctuating signal is to employ the
box-car or moving-average technique. Starting from a measured spectrum y, a
smoothed spectrum y* is obtained by calculating the mean channel content around each
channel i:
y
+
i
= " y
i
=
1
2m1
m
j=m
y
ij
(20)
This can be seen as a convolution [Eq. (14)] with all coecients h
j
= 1. The smoothing
eect obviously depends on the width of the lter, 2m1. The operation being a simple
averaging, the standard deviation of the smoothed data is reduced by a factor
2m1
_
in
regions where y
i
is nearly constant. On the other hand, such a lter introduces a con-
siderable amount of peak distortion. This distortion depends on the lter width-to-peak
width ratio. Figure 4 shows the peak distortion eects when a moving-average lter of
widths 9, 17, and 25 is applied to a peak with full width at half-maximum (FWHM) equal
to nine channels. Being a unit area lter
h
j
=N = 1 with N=2m1, the total counts in
the peak is not aected in an appreciable way other than by rounding errors.
Figure 4 Distortion introduced by smoothing of a peak with a moving-average filter 9, 17, and 25
points wide. The FWHM of the original peak is nine channels.
Copyright 2002 Marcel Dekker, Inc.
Figure 5 shows the eect on the peak height and width when applying this type of
lter with dierent sizes. The peak distortion is caused by the fact that in the calculation of
y
+
i
, the content of all neighboring channels is used with equal weight. Consequently, by
employing a nonuniform lter h, which places more weight on the central channels and
less on the channels near the edge of the lter, smoothing can be achieved with less
broadening eects.
2. Savitskyand Golay Polynomial Filters
Another way of dealing with statistical uctuations in experimental data is by drawing a
best-tting curve through the data points. This idea resulted in the development by
Savitsky and Golay (1964) of a general type of smoothing lters with very interesting
features. The method is based on the fact that nearly all experimental data can be modeled
by a polynomial of order r, a
0
a
1
x a
2
x
2
a
r
x
r
, when the data are conned to a
suciently small interval. If we consider a number of data points around a central channel
i
o
, such as y
i
0
2
; y
i
0
1
; y
i
0
; y
i
0
1
; y
i
0
2
, a least-squares t with the function.
y(i) = a
0
a
1
(i i
0
) a
2
(i i
0
)
2
(21)
can be made. Once we have determined the coecients a
j
, the value of the polynomial at
the central channel i
0
can be used as the smoothed value:
y
+
i
= y(i
0
) = a
0
(22)
This concept is schematically illustrated in Figure 6. By moving the central channel to the
right (from i
0
to i
0 1
), the next smoothed channel content can be calculated by repeating
the entire procedure.
Figure 5 Percent change in peak height and width introduced by filtering with a moving-average
and a Savitsky and Golay polynomial filter as a function of the filter width-to-peak FWHM ratio,
(2m1)/FWHM.
Copyright 2002 Marcel Dekker, Inc.
At rst sight, this smoothing method would require a least-squares t for each
channel in the spectrum. However, the fact that the x values are equidistant allows us to
formulate the problem in such a way that the polynomial coecients can be expressed as
simple linear combinations involving only y
i
values:
a
k
=
1
N
k
j=m
j=m
C
k; j
y
ik
(23)
This means that it is possible to implement the least-squares-tting procedure more e-
ciently as a convolution of the spectrum with a lter having appropriate weights. For this
second-order polynomial, the coecients are given by
C
0; j
N
0
=
3(3m
2
3m1 5j
2
)
(2m1)(2m1)(2m3)
(24)
Smoothing with a ve-point second-degree polynomial (2m1 = 5) thus becomes
y
+
i
= a
0
=
1
35
(3y
i2
12y
i1
17y
i
12y
i1
3y
i2
) (25)
In general, for a polynomial of degree r tted to 2m1 points, this can be written as
y
+
i
=
1
N
rm
j=m
j=m
C
rm; j
y
ij
(26)
where the convolution integers C
rm;j
and the normalization factors N
rm
do not depend on
the data to be smoothed but are a function only of the polynomial degree r and on the lter
Figure 6 Concept of polynomial smoothing. A parabola is fitted trough the points i
0
3 to i
0
3.
The value of the parabola at i
0
is the smoothed value (
v
).
Copyright 2002 Marcel Dekker, Inc.
half-width m. Table 1 lists the coecients of polynomial smoothing lters with widths
between 5 and 25 points. The coecients for a second- and a third-degree polynomial are
identical. In comparison with the moving-average lter of the same width, polynomial
smoothing lters are less eective in removing noise but have the advantage of causing less
peak distortion. The distortion eect as a function of the lter width-to-peak width ratio is
given in Figure 5. When the lter becomes much wider than the peak, the smoothed
spectrum features oscillations near the peak boundaries, as illustrated in Figure 7.
An interesting feature of this type of lters is that they can produce not only a
smoothed spectrum but also a smoothed rst and second derivative of the spectrum. If we
dierentiate Eq. (21) and take the value at the center position,
y
+/
=
dy(i)
di
i=i
0
= a
1
(27)
y
+//
=
d
2
y(i)
di
2
i=i
0
= 2a
2
(28)
it follows from Eq. (23) that the smoothed rst and second derivative of the spectrum can
also be calculated by means of suitable convolution coecients. For instance, for the rst
derivative of a second-order polynomial, using ve data points this becomes
y
+/
i
=
1
10
(2y
i2
y
i1
y
i1
2y
i2
) (29)
The corresponding convolution integers for the calculation of the smoothed rst and
second derivative are listed in Tables 24. The use of the derivative spectra is illustrated in
next section dealing with peak search methods. The FORTRAN implementation of the
Savitsky and Golay lters is given in Sec. X.
Variations on these smoothing strategies, such as the use of variable-width lters are
reviewed by Yule (1967). The eect of repeated smoothing on the accuracy and precision
of peak area determination is discussed by Nielson (1978). A more comprehensive treat-
ment on polynomial smoothing can be found in Enke and Nieman (1976) and its
references.
Table 1 Savitsky and Golay Coefcients for Second- and Third-Degree (r =2 and r =3)
Polynomial Smoothing Filter; y
+
i
= (1=N
rm
)
j=m
j=m
C
rm; j
y
ij
, Filter Width =2m1
j (C
rm; j
= C
rm;j
)
m N
rm
0 1 2 3 4 5 6 7 8 9 10 11 12
2 35 17 12 3
3 21 7 6 3 2
4 231 59 54 39 14 21
5 429 89 84 69 44 9 36
6 143 25 24 21 16 9 0 11
7 1105 167 162 147 122 87 42 13 78
8 323 43 42 39 34 27 18 7 6 21
9 2261 269 264 249 224 189 144 89 24 51 136
10 3059 329 324 309 284 249 204 149 84 9 76 171
11 805 79 78 75 70 63 54 43 30 15 2 21 42
12 5175 467 462 447 422 387 342 287 222 147 62 33 138 253
Copyright 2002 Marcel Dekker, Inc.
3. LowStatistics Digital Filter
Most smoothing techniques originate from signal processing and were initially introduced
in the eld of g-ray spectroscopy. Also in the PIXE community, considerable attention has
been devoted to a number of aspects of spectrum processing and evaluation. Within the
framework of continuum estimation (see later), a smoothing algorithm was developed that
removes noise from a spectrum on a selective basis (Ryan et al., 1988). The method
provides an n-point means smoothing in regions of low statistics (few counts) while
avoiding spreading of the base of peaks and degradation of valleys between peaks.
Figure 7 Effect of the smoothing of a peak with a Savitsky and Golay polynomial filter. The
FWHM of the original peak is nine channels.
Table 2 Savitsky and Golay Coefcients for Second-Degree (r =2) Polynomial, First-Derivative
Filter; y
+
i
= (1=N
rm
)
j=m
j=m
C
rm; j
y
ij
, Filter Width =2m1
j (C
rm; j
= C
rm;j
)
m N
rm
0 1 2 3 4 5 6 7 8 9 10 11 12
2 10 0 1 2
3 28 0 1 2 3
4 60 0 1 2 3 4
5 110 0 1 2 3 4 5
6 182 0 1 2 3 4 5 6
7 280 0 1 2 3 4 5 6 7
8 408 0 1 2 3 4 5 6 7 8
9 570 0 1 2 3 4 5 6 7 8 9
10 770 0 1 2 3 4 5 6 7 8 9 10
11 1012 0 1 2 3 4 5 6 7 8 9 10 11
12 1300 0 1 2 3 4 5 6 7 8 9 10 11 12
Copyright 2002 Marcel Dekker, Inc.
Table 3 Savitsky and Golay Coefcients for Third-Degree (r =3) Polynomial, First-Derivative Filter; y
+
i
= (1=N
rm
)
j=m
j=m
C
rm; j
y
ij
, Filter
Width =2m1
j (C
rm; j
= C
rm;j
)
m N
rm
0 1 2 3 4 5 6 7 8 9 10 11 12
2 12 0 8 1
3 252 0 58 67 22
4 1,188 0 126 193 142 86
5 5,148 0 296 503 532 294 300
6 24,024 0 832 1,489 1,796 1,578 660 1,133
7 334,152 0 7,506 13,843 17,842 18,334 14,150 4,121 12,922
8 23,256 0 358 673 902 1,002 930 643 98 748
9 255,816 0 2816 5,363 7,372 8,574 8,700 7,481 4,648 68 6,936
10 3,634,092 0 29,592 56,881 79,564 95,338 101,900 96,947 78,176 43,284 10,032 84,075
11 197,340 0 1,222 2,365 3,350 4,098 4,530 4,567 4,130 3,140 1,518 815 3,938
12 1,776,060 0 8,558 16,649 23,806 29,562 33,450 35,003 33,754 29,236 20,982 8,525 8,602 30,866
Copyright 2002 Marcel Dekker, Inc.
The rather heuristic algorithm is discussed in Section X, together with the computer im-
plementation. The eect of the lter is illustrated in Figure 8, where it is compared with the
other smoothing methods discussed.
C. Peak Search Methods
Several methods have been developed for the automatic localization of peaks in a spec-
trum. Nearly all methods follow a strategy where the original spectrum is transformed into
a form that emphasizes the peaklike structures and reduces the continuum, followed by a
decision whether these peaklike structures are statistically signicant. The latter involves
some adjustable parameter(s) controlling the sensitivity of the peak search.
Although visual inspection of the spectrum still appears to be the best method, peak
search algorithms, which are heavily used in g-ray spectrometry, may have some value in
x-ray spectrometry. Their use in energy-dispersive x-ray analysis as part of an automated
qualitative analysis procedure is hampered by the extreme peak overlap in these spectra.
More elaborated procedures involving articial intelligence techniques have been con-
sidered for this (Janssens and Van Espen, 1986; Janssens et al., 1988).
Peak search procedures usually involve three steps: (1) transformation of the original
spectrum so that continuum contribution is eliminated, peaks are readily locatable, and
overlapping peaks are (partially) resolved; (2) signicance test and approximate location of
the peak maximum; and (3) more accurate peak position estimate in the original spectrum.
The various peak search algorithms mainly dier in the choice of the transformation.
Some methods use the rst and second smoothed derivative of the spectrum. This method is
illustrated in Figure 9. The sign change (crossing of the x axis) of the rst derivative and the
minimumof the secondderivative are quire suitable todetect the peaks inthe original spectrum.
Other methods employ some form of correlation technique, which is basically the
convolution of the original spectrum with a lter that approximates the shape of the peak
and, therefore, emphasizes the peak. If a zero-area correlator (lter) is used, the continuum
is at the same time eectively suppressed. The simplest and most eective correlators belong
to the group of zero-area rectangular lters. These lters have a central window with
Table 4 Savitsky and Golay Coefcients for Second- and Third-Degree (r =2 and r =3)
Polynomial Second-Derivative Filter; y
+
i
= (1=N
rm
)
j=m
j=m
C
rm; j
y
ij
, Filter Width =2m1
j (C
rm; j
= C
rm;j
)
m N
rm
0 1 2 3 4 5 6 7 8 9 10 11 12
2 7 2 1 2
3 42 4 3 0 5
4 462 20 17 8 7 28
5 429 10 9 6 1 6 15
6 1,001 14 13 10 5 2 11 22
7 6,188 56 53 44 29 8 19 52 91
8 3,876 24 23 20 15 8 1 12 25 40
9 6,783 30 29 26 21 14 5 6 19 34 51
10 33,649 110 107 98 83 62 35 2 37 82 133 190
11 17,710 44 43 40 35 28 19 8 5 20 37 56 77
12 26,910 52 51 48 43 36 27 16 3 12 29 48 69 92
Copyright 2002 Marcel Dekker, Inc.
constant and positive coecients and two side lobes with constant and negative coecients.
Convoluting an x-ray spectrumwith this kind of lter yields spectra in which the continuum
is removed and peaks are easily locatable. They are similar to inverted second-derivative
spectra. An important representative of this group of lters is the top-hat lter, which has
a central window with an odd number of channels w and two side windows each n channels
wide. The value of the lter coecients follows from the zero-area constraint:
h
k
=
1
2n
; n
w
2
_ k <
w
2
1
w
;
w
2
_ k _
w
2
1
2n
;
w
2
< k _
w
2
n
_
_
_
(30)
The ltered spectrumis obtained by the convolution of the original spectrumwith this lter:
y
+
i
=
k=nw=2
k=nw=2
h
k
y
ik
(31)
The eect of this lter on a typical spectrum is shown in Figure 10. The variance of the
ltered spectrum is obtained by simple error propagation:
s
2
y
+
i
=
k=nw=2
k=nw=2
h
2
k
y
ik
(32)
Figure 8 Effect of various smoothing methods applied to part of a x-ray spectrum: (A) original
spectrum, (B) nine-point Savitsky and Golay filter, (C) nine-point moving average, (D) low-noise
statistical filter.
Copyright 2002 Marcel Dekker, Inc.
If y
+
i
is signicantly dierent from zero, a peak structure is found and the top of the peak
can approximately be located by searching for the maximum. Thus, i corresponds to the
position of a peak maximum in the original spectrum if
y
+
i
> rs
y
i
(33)
Figure9 Peak search using first and second derivatives: (A) doublet, peak width (s) four channels,
separation eight channels, on a constant continuum. (B) First and (C) second smoothed derivative
using a five-point Savitsky and Golay filter.
Copyright 2002 Marcel Dekker, Inc.
and
y
+
i1
_ y
+
i
> y
+
i1
(34)
In Figure 11, the positive part of the ltered spectrum (w=9 and n = 5) and the decision
line (rs
+
y
) for r =1 and 4 are displayed.
If required, other peak features can be obtained from the ltered spectrum: The
distance between the two local minima is a measure of the width of the peak and the height
at the maximum is related to the net peak area.
Because the width and heights of the peaks in the ltered spectrum strongly depend
on the dimensions of the lter, it is important that its dimensions are matched to the width
of the peaks in the original spectrum. From considerations of peak detectability (signal-to-
noise ratio) and resolution (peak broadening), it follows that the optimum width of the
positive window w is equal to the FWHM of the peaks (Robertson et al., 1972). The width
of the negative side windows should be chosen as large as the curvature of the continuum
allows. A reasonable compromise between sensitivity to peak shapes and rejection of
continuum is reached when n equals FWHM=2 to FWHM=3. Typical values for the
sensitivity factor r are between 2 and 4. Higher values result in the loss of small peaks;
lower values will cause continuum uctuations to be interpreted as peaks.
Other zero-area rectangular lters, variations to the top-hat lter, are also in use,
such as the square-wave lter with typical coecient sequence 1, 1, 2, 2, 1, 1 (Philips
and Marlow 1976; McCullagh, 1982) and the symmetric square-wave lter with coe-
cients 1, 1, 1, 1 (Op De Beeck and Hoste 1975). A detailed account of the performance
of this lter is given by Op De Beeck and Hoste (1975). A method using a Gaussian
correlator function is discussed by Black (1969).
Figure 10 Result of applying a top-hat filter. Dotted line: typical X-ray spectrum; solid line:
filtered spectrum.
Copyright 2002 Marcel Dekker, Inc.
Once the peak top is approximately located in the ltered spectrum, a more precise
maximum can be found by tting a parabola over a few channels around the peak. For a
well-dened peak on a low continuum (or after continuum subtraction), the channel
content near the top of the peak can be approximated by a Gaussian:
y
i
- h exp
(x m)
2
2s
2
_ _
(35)
The logarithm of the data then is a simple polynomial
ln y
i
= ln h
m
2
2s
2
_ _
m
s
2
x
m
2
2s
2
x
2
(36)
If we t ln y
i
with a polynomial a
0
a
1
x a
2
x
2
, where x represents the channel number,
the position of the peak m is obtained from
m =
a
1
2a
2
(37)
Figure 11 Peak search using the positive part of the top-hat filtered spectrum and the decision
level for one and four times the standard deviation. The original spectrum is shown at the bottom.
Copyright 2002 Marcel Dekker, Inc.
with an accuracy of 0.1 channel or better if the continuum contribution is small and if the
peak is interference free [i.e., if Eq. (35) accurately describes the data]. An estimate of the
peaks width and height is obtained at the same time:
FWHM = 2
2 ln 2
_
s = 2:3548
1
2a
2
_
(38)
h = exp a
0
a
2
1
4a
2
_ _
(39)
To obtain a reliable estimate of the parameters, only the channels in the FWHM or, at
most, within the FWTM region of the peak must be included in the t.
As a somewhat simpler and faster alternative, one can nd an estimate of the peak
maximum by tting the parabola over the three top channels of the peak. If i is the peak
maximum found in the ltered spectrum, a better estimate of the maximum in the original
spectrum is found by
m = i
1
2
y
i1
y
i1
y
i1
y
i1
2y
i
(40)
This method might be preferred for small peaks when the continuum cannot be
disregarded.
A FORTRAN implementation of a peak search algorithm is given in Section X.
IV. CONTINUUMESTIMATIONMETHODS
Except for some special quantication procedures (e.g., the peak-to-background method
in electron microscopy), the relevant analytical information is found in the net peak areas
and continuum is considered a nuisance. There are, in principle, three ways to deal with
the continuum: (1) the continuum can be suppressed or eliminated by a suitable lter;
(2) the continuum can be estimated and subtracted from the spectrum prior to the esti-
mation of the net peak areas; and (3) the continuum can be estimated simultaneously with
the other features in the spectrum. The rst approach is discussed in Section VI, where the
continuum is removed from spectra by applying a top-hat lter followed by linear least-
squares t of the spectrum with a number of (also ltered) reference spectra. Least-squares
t (linear or nonlinear) with analytical functions (Sec. VII) allows the simultaneous esti-
mation of continuum and peaks, providing a suitable mathematical function can be found
for the continuum. In this section, we discuss a number of procedures that aim to estimate
the continuum independently of the other features in the spectrum. Once estimated, this
continuum can be subtracted from the original spectrum and all methods for further
processing, ranging from simple peak integration to least-squares tting can be applied.
Any continuum estimation procedure must fulll two important requirements. First,
the method must be able to reliably estimate the continuum in all kinds of situations (e.g.,
small isolated peaks on a high continuum as well as in the proximity of a matrix line).
Second, to permit processing of a large number of spectra, the method needs to be nearly
free of user-adjustable parameters.
Although a number of useful continuum estimation procedures has been developed,
it must be realized that their accuracy in estimating the continuum is not optimal. In one
way or another, they rely on the dierence in frequency response of the continuum
compared to other structures such as peaks, the former mainly consisting of low
Copyright 2002 Marcel Dekker, Inc.
frequencies (slowly varying). Because the peaks also exhibit low frequencies at the peak
boundaries, it is dicult to control the method in such a way that it correctly discriminates
between peaks and continuum. This results in either a small underestimation or over-
estimation of the continuum, introducing potentially large relative errors for small peaks.
In this respect, the tting of the continuum with analytical functions may provide more
optimal results (Vekemans et al., 1994). A considerable advantage of the methods dis-
cussed here is that they do not assume an explicit mathematical model of the continuum.
Constructing a detailed and accurate analytical model for the continuum based in physical
theory is nearly impossible except for some simple geometry and particular excitation
conditions. Most often, some polynomial type of function must be chosen when tting a
portion of the spectrum with analytical functions.
A. Peak Stripping
These methods are based on the removal of rapidly varying structures in a spectrum by
comparing the channel content y
i
with the channel content of its neighbors. Clayton et al.
(1987) proposed a method which compares the content of channel i with the mean of its
two direct neighbors:
m
i
=
y
i1
y
i1
2
(41)
If y
i
is smaller than m
i
, the content of channel i is replaced by the mean m
i
. If this trans-
formation is executed once of all channels, one can observe a slight reduction in the height
of the peaks while the rest of the spectrum remains virtually unchanged. By repeating this
procedure, the peaks are gradually stripped from the spectrum. Because the method
tends to connect local minima, it is very sensitive to local uctuations in the continuum due
to counting statistics. This makes smoothing of the spectrum, as discussed in the previous
section, prior to the stripping process mandatory. Depending on the width of the peaks
after typically 1000 cycles, the stripping converges and a more or less smooth continuum
remains. To reduce the number of iterations, it might be advantageous to perform a log or
square root transformation to the data prior to the stripping: y
/
i
= log(y
i
1) or y
/
i
=
y
i
_
.
After the stripping, the continuum shape is obtained by applying the inverse transforma-
tion. A major disadvantage of this method is that after a number of cycles, the bases of
partially overlapping peaks are transformed into broad humps, which take much longer
to remove than isolated peaks. The method was originally applied to PIXE spectra but
proves to be generally applicable for pulseheight spectra.
In Figure 12, this method is applied to estimate the continuum of an x-ray spectrum
in the region between 1.6 and 13.0 keV. The spectrum results from a 200-mg=cm
2
pellet of
NIST SRM Bovine Liver sample excited with the white spectrum of an Rh-anode x-ray
tube ltered through a thin Rh lter (Tracor Spectrace 5000). Because of the white tube
spectrum, a considerable continuum intensity was observed, increasing quite steeply in the
region above 10 keV. To calculate the continuum, the following algorithm was used:
(1) the square root of the original spectrum was taken; (2) these data were smoothed with a
10-point Savitsky and Golay lter; (3) a number of iterations were performed applying
Eq. (41) over the region of interest; (4) the square of each data point was taken (back
transformation) to obtain the nal continuum shape. In Figure 12, the continuum after
10, 100, and, nally, 500 iterations is shown.
As a generalization of the above-discussed method, the average of two channels a
distance w away from i can be used:
Copyright 2002 Marcel Dekker, Inc.
m
i
=
y
iw
y
iw
2
(42)
Ryan et al. (1988) proposed using twice the FWHM of the spectrometer at channel i as the
value for w. They reported that only 24 passes were required to produce acceptable
continuum shapes in PIXE spectra. During the last eight cycles, w is progressively reduced
by the factor
2
_
to obtain a smooth continuum. To compress the dynamic range of the
spectrum, a double-log transformation of the spectrum, log[log( y
i
+1)+1], before the
actual stripping was proposed. In combination with the low statistics digital lter, this
procedure is called the SNIP algorithm (Statistical Nonlinear Iterative Peak clipping).
Avariant of this procedure is implemented in the procedure SNIPBG given in Sec. X.
Instead of the double logarithmic, we employed a square root transformation, and a Sa-
vitsky and Golay smoothing is performed on the square root data. The width w is kept
constant over the entire spectrum. The value of w is also used as the width of the smoothing
lter. Using this implementation, the continuum of the above-discussed spectrum is cal-
culated and represented in Figure 13. The width was set 11 to channels approximately
corresponding to the FWHM of the peaks in the center of the spectrum and 24 iterations
were done. Apart from delivering a smoother continuum with smaller humps, this
method executes much faster than the original method proposed by Clayton.
B. ContinuumEstimation Using Orthogonal Polynomials
Another interesting continuum estimation procedure was introduced by Steenstrup (1981),
who applied the method to energy-dispersive x-ray diraction spectra. The spectrum is
tted using orthogonal polynomials, and the weights of the least-squares t are iteratively
adjusted so that only channels belonging to the continuum are included in the t.
Figure 12 Continuum estimate after 10, 100, and 500 iterations obtained with simple iterative
peak stripping.
Copyright 2002 Marcel Dekker, Inc.
The method is generally applicable to pulseheight spectra and can be implemented as an
algorithm that needs little or no control parameters.
The continuum is described by a set of polynomials up to degree m:
y(i) =
m
j=0
c
j
P
j
(x
i
) (43)
where P
j
(x
i
) is an orthogonal polynomial of degree j. As an example of m=3, the function
becomes
y(i) = c
0
c
1
(x
i
a
0
) c
2
((x
1
a
1
)(x
i
a
0
) b
1
) (44)
The least-squares estimates of the parameters c
j
are given by
c
j
=
n
i=1
w
i
y
i
P
j
(x
i
)
g
j
(45)
where w
i
are the weights of the t. A detailed discussion on orthogonal polynomial tting
can be found in Sec. IX. Because the polynomial terms P
j
(x
i
) are orthogonal, no matrix
inversion is required to obtain the results. Orthogonal polynomials of a much higher
degree can be tted to the experimental data without running into problems with ill-
conditioned normal equations and oscillating terms.
The goal of this continuum estimation method is to t the continuum with the
above-described orthogonal polynomial of degree m and to interpolate under the peaks.
This can be achieved by careful manual selection of only those data pairs (x
i
; y
i
) that
belong to the continuum. The more elegant approach proposed by Steenstrup consisted of
using all channels and automatically adjusting the weights in such a way that the con-
tinuum contributions are emphasized. If y(i) is a polynomial approximation of the
continuum of degree m, then y
i
- y(i) if i is a continuum channel, otherwise y
i
> y(i).
Figure13 Continuum estimate using statistical nonlinear iterative peak clipping (SNIP algorithm)
with 24 iterations.
Copyright 2002 Marcel Dekker, Inc.
A better approximation to the continuum can then be found by choosing small weights for
the data points where y
i
> y(i) and repeating the t. The following weighting scheme is
proposed by Steenstrup:
w
i
=
1
y(i)
if y
i
_ y(i) r
y(i)
_
(46)
w
i
=
1
[ y(i) y
i
[
2
if y
i
> y(i) r
y(i)
_
(47)
where r is an adjustable parameter. Typically, r = 2 is used. If y
i
is normally distributed
(which is approximately the case for y
i
> 30 counts), y
i
_ y(i) 2
y(i)
_
holds for 97.7%
of the channels containing only continuum. A too high value of r will cause the inclusion
of the tails of the peaks into the t. With the new weights, a new polynomial t of degree m
is performed. The process is stopped when the new polynomial coecients c
j
are within
one standard deviation from the previous ones. We have also obtained good results by
setting the weights eectively to zero when the channel content is statistically above the
tted continuum [Eq. (47)].
The method can be made even more unsupervised by including a procedure to au-
tomatically select the best degree of the polynomial. This can be done by tting (as de-
scribed earlier) successive polynomials with higher degrees and testing the signicance of
each additional polynomial coecient. If the coecient C
m1
is statistically not signicant
dierent from zero,
[C
m1
[ < 2s
C
m1
(48)
a polynomial of degree m is retained.
Figure 14 shows the spectrum of bovine liver sample (same as Fig. 12) and the tted
continuum using a fourth-, fth-, and sixth-degree orthogonal polynomial. A value of 1.5
was chosen for r [Eqs. (46) and (47)] and the weights were set to zero for the channels not
belonging to the continuum; w
i
= 0 in Eq. (47). Table 5 gives the data on the ts. The nal
number of continuum channels retained, the number of weight adjustments performed,
and the value and standard deviation of the highest polynomial coecient are shown. The
total number of channels in the tting region was 571. Because the value of the seventh
degree polynomial coecient is insignicant [Eq. (48)], it is concluded that a sixth-degree
orthogonal polynomial is needed to model the continuum.
Section X lists the FORTRAN code that implements the complete procedure to esti-
mate the continuum of a given spectrum by an mth-degree orthogonal polynomial.
V. SIMPLE NET PEAK AREADETERMINATION
In both in WDXRF and EDXRF, the concentration of the analyte is proportional to the
number of counts under the characteristic x-ray peak corrected for the continuum. At
constant resolution, this proportionality also exits for the net peak height. In EDXRF,
preference is given to the peak area. In WDXRF, the acquisition of the entire peak prole
is very time-consuming and the count rate is usually measured only at the peak maximum.
A. Peak-Area Determination in EDXRF
The most straightforward method to obtain the net area of an isolated peak in a spectrum
consists of interpolating the continuum under the peak and summing the continuum-
corrected channel contents in a window over the peak, as shown in Figure 15.
Copyright 2002 Marcel Dekker, Inc.
The net peak area, N
p
, of an isolated peak on a continuum is given by
N
P
=
i
P1
i
P2
[ y
i
y
B
(i)[ =
i
y
i
i
y
B
(i) = N
T
N
B
(49)
where N
T
and N
B
are the total number of counts of the spectrum and the continuum in the
integration window i
P1
i
P2
. The uncertainty of the estimated net peak area due to
counting statistics is
s
N
P
=
N
T
N
B
_
(50)
The continuum y
B
(i) is calculated by interpolation, assuming a straight line (see Fig. 15):
y
B
(i) = Y
BL
(Y
BR
Y
BL
)
i i
BL
i
BR
i
BL
(51)
Figure 14 Continuum estimate obtained by fitting a fourth-, fifth-, and sixth-degree orthogonal
polynomial.
Table5 Data on the Fit of the Continuum of the Spectrum Shown in Figure 14 Using Orthogonal
Polynomials
Degree of
polynomial
No. of channels
used
No. of weight
adjustments
Highest degree coecient
and standard deviation
4 126 7 (3.980.20)10
6
5 199 16 (1.880.09)10
7
6 267 6 (7.660.28)10
10
7 266 6 (3.662.03)10
13
Copyright 2002 Marcel Dekker, Inc.
Y
BL
and Y
BR
are the values of the continuum at the channels i
BL
and i
BR
, left and right
from the peaks, respectively. These values are best estimated by averaging over a number
of channels:
Y
BL
=
1
n
BL
i
BL1
i=i
BL2
y
i
=
N
BL
n
BL
(52)
Y
BR
=
1
n
BR
i
BR1
i=i
BR2
y
i
=
N
BR
n
BR
(53)
The number of channels in the continuum windows are n
BL
= i
BL2
i
BL1
1 and
n
BR
= i
BR2
i
BR1
1. The center position of the continuum windows (not necessarily an
integer number!) used in Eq. (51) are i
BL
= (i
BL1
i
BL2
)=2 and i
BR
= (i
BR1
i
BR2
)=2.
If both continuum windows have equal width, n
BL
= n
BR
= n
B
=2, and are positioned
symmetrically with respect to the peak window (i
P
i
BL
= i
BR
i
P
) a much simpler ex-
pression is obtained for the net peak area:
N
P
= N
T
n
P
n
B
(N
BL
N
BR
) (54)
where N
BL
and N
BR
are the total counts in the left and right continuum windows, re-
spectively, each n
B
=2 channels wide, and n
P
equals the number of channels in the peak
window. Applying the principle of error propagation, the uncertainty in the net peak area
is then given by
s
N
P
=
N
T
n
P
n
B
_ _
2
(N
BL
N
BR
)
(55)
Figure15 Determination of the net peak area by interpolating the continuum.
Copyright 2002 Marcel Dekker, Inc.
From this equation, it can be seen that, in principle, the continuum should be estimated
using as many channels as possible (n
B
) to minimize random errors due to counting
statistics. In practice, the width will be limited by curvature of the continuum and by the
presence of other peaks. Most often, the total width of the continuum (n
B
= n
BL
n
BR
) is
taken equal to or slightly larger than the width of the peak window.
The optimum width of the peak window to minimize counting statistics depends on
the peak-to-continuum ratio (Jenkins et al., 1981). For low peak-to-continuum ratios, the
peak height being only a fraction of the continuum height, the optimum width of the
integration window is 1.17 times the FWHM of the peak. If the ratio of the peak height to
the continuum height is larger than 1, a slightly wider window is optimal, although, in
practice, the improvement in precision is negligible.
The method does not deliver the total net peak area, only a fraction of it. Integrating
over 1.176FWHM (from 1:378s to 1:378s) covers only 83% of the peak. To cover
99% of the peak, the window should be 2.196FWHM (2.579s). If position or width
changes are observed from one spectrum to the other, care should be taken to dene the
windows in such a way that cover the same part of the spectrum. The wider the peak
window, the less sensitive the method for peak shift.
Although such a peak-area-determination method seems naively simple, if correctly
used it provides results that are as accurate and precise as the most sophisticated proce-
dures. The premises for use are that the peak window should be known to be free from
interferences, that there should be no peaks in the continuum windows, and that the con-
tinuum should be linear over the extent of the windows. A peak search procedure can, in
principle, be used to setup the windows automatically. However, its practical use is limited
by complexity of the EDXRF spectra. Moreover, the use of such automated procedures is
hazardous because no measure can be given for the presence or absence of systematic errors.
Because of this restrictions, a simple peak integration method cannot be used as a
general tool for spectrum processing. In a limited number of applications, good results can
be obtained. An evaluation of various peak integration methods is given by Hertogen et al.
(1973).
B. Net Count Rate Determination inWDXRF
In WDXRF, the count rate at the 2Y angle of the peak maximum corrected for the
continuum is used as analytical signal. The continuum is estimated at a 2Y position on the
left or right side of the peak. Continuum interpolation as described in the previous section
is also possible.
If N
P
is the number of counts accumulated during a time interval t
P
at the top of the
peak and N
B
is the number of the continuum counts accumulated during time t
B
, then the
net count rate, I, is given by
I = I
P
I
B
=
N
P
t
P
N
B
t
B
(56)
and the uncertainty in the net count rate due to counting statistics is given by
s
I
=
N
P
t
2
P
N
B
t
2
B
(57)
Various counting strategies can be considered and the eect on the precision can be esti-
mated using Eq. (57) (Bertin, 1970). In a optimum xed time strategy, the minimum
Copyright 2002 Marcel Dekker, Inc.
uncertainty is obtained when, for a total measurement time t = t
P
t
B
; t
P
and t
B
are chosen
in such a way that their ratio is equal to the square root of the peak-to-continuum ratio:
t
P
t
B
=
I
P
I
B
_
(58)
Under, these conditions, the uncertainty in the net intensity is given by
s
I
=
I
P
_
I
B
_
t
P
t
B
_ (59)
VI. LEAST-SQUARES FITTINGUSINGREFERENCE SPECTRA
In this section, two techniques based on linear least squares are discussed. The lter-t
methods makes use of library spetra, measured or calculated spectra of pure compounds,
that are used to describe the spectra of complex samples. The other method is based on
partial least-squares (PLS) regression, a multivariate calibration technique. In this case, no
spectrum evaluation in the strict sense is performed, but, rather, relations between the
concentrations of the compounds in the samples and the entire spectrum are established.
In this way, quantitative analysis is possible without obtaining net peak areas of the
characteristic lines.
A. Filter-Fit Method
1. Theory
If a measured spectrum of an unknown sample can be described as a linear combination of
spectra of pure elements constituting the sample, then the following mathematical model
can be written:
y
mod
i
=
m
j=1
a
j
x
ji
(60)
with y
mod
i
the content of channel i in the model spectrum and x
ji
the content of channel i in
the jth reference spectrum. The coecients a
j
are a measure of the contribution of pure
reference spectra to the unknown spectrum and can be used for quantitative analysis. The
values of the coecients a
j
are obtained via multiple linear least-squares tting, mini-
mizing the sum of the weighted squared dierences between the measured spectrum and
the model:
w
2
=
n
2
i=n
1
1
s
2
i
[ y
i
y(i)[
2
=
n
2
i=n
1
1
s
2
i
y
i
m
j=1
a
j
x
ji
_ _
2
(61)
where y
i
and s
i
are the channel content and the uncertainty of the measured spectrum,
respectively, and n
1
and n
2
are the limits of the tting region. A detailed discussion of the
least-squares-tting method is given in Sec. IX.
The assumption of linear additivity [Eq. (60)] holds normally reasonable well for the
characteristic lines in the spectrum, but not for the continuum. To apply this technique,
the continuum can be removed from the unknown spectrum and from the reference
spectra using one of the procedures described in Sec. IV before the actual least-squares t.
Copyright 2002 Marcel Dekker, Inc.
Another, frequently used approach is to apply a digital lter to both unknown and
reference spectra. This variant is known as the lter-t method (Schamber, 1977; Statham,
1978; McCarthy and Schamber, 1981) and is discussed in some detail below.
By the discrete convolution of a spectrum with top-hat lter [Eqs. (30) and (31)], the
low-frequency component (i.e. the slowly varying continuum) is eectively suppressed as
discussed in Sec. III. Apart from removing the slowly varying continuum, a rather severe
distortion of the peaks is also introduced. If we apply this lter to both the unknown
spectrum and the reference spectra, the nonadditive continuum is removed and the same
type of peak distortion will be introduced in all spectra, allowing us to apply the method of
multiple linear least-squares tting to the ltered spectra. Equation (61) then becomes
w
2
=
n
2
i=n
1
1
s
/ 2
i
y
/
i
m
j=1
a
j
x
/
ji
_ _
2
(62)
Where y
/
i
and x
/
ji
are the ltered unknown and reference spectra, respectively, s
/ 2
i
is the
variance of y
/
i
given by
s
/2
i
=
k
h
2
k
y
ik
(63)
The least-squares estimates of the contribution of each reference spectrum is then given by
(see Sec. IX)
a
j
=
m
k=1
a
1
jk
b
k
; j = 1; . . . ; m (64)
with
b
j
=
n
2
i=n
1
1
s
/2
i
y
/
i
x
/
ji
(65)
a
jk
=
n
2
i=n
1
1
s
/ 2
i
x
/
ki
x
/
ji
(66)
The uncertainty in each coecient a
j
is directly estimated from the error matrix:
s
2
aj
= a
1
jj
(67)
Schamber (1977) suggested the following equation for the uncertainties, taking into ac-
count the eect of the lter:
s
2
aj
=
nw
n w
a
1
jj
(68)
Where w is the width of the central positive part of the lter and n is the width of the
negative wings. A measure of the goodness of t is available through the reduced w
2
value:
w
2
n
=
1
(n
2
n
1
1) m
w
2
(69)
which is the w
2
value of Eq. (62) divided by the number of points in the t minus the
number of reference spectra. A value close to 1 means a good t, indicating that the
reference spectra are capable of adequately describing the unknown spectrum.
Copyright 2002 Marcel Dekker, Inc.
Most of the merits and the disadvantages of the lter-t method can be deduced
directly from the mathematical derivation given in the preceding paragraphs. The most
interesting aspect of the lter-t method is that it does not require any mathematical
model for the continuum and that, at least in principle, the shape of the peaks in the
unknown spectrum are exactly represented by the reference spectra. Reference spectra
should be acquired with good counting statistics, at least better than the unknown spec-
trum, because the least-squares method assumes that there is no error in the independent
variables x
/
ji
. Reference spectra can be obtained from single-element standards. Only the
portion of the spectrum that contains peaks needs to be retained as reference in the t.
Multielement standards can be used if the peaks of each element are well separated.
The reference spectra must provide an accurate model of the peak structure present
in the unknown spectrum. This requires that reference and unknown spectra are acquired
under strictly identical spectrometer conditions. Changes in resolution and, especially,
energy shifts can cause large systematic errors. The magnitude of this error depends on the
degree of peak overlap. Peak shifts of more than a few electron volts should be avoided,
which is readily possible with modern detector electronics. If shifts are observed over long
periods of operations of the spectrometer, careful recalibration of the spectrometer is
required or, better, the reference spectra should be acquired again. Also, peak shift and
peak broadening due to dierences in count rate between standards and unknown must be
avoided. Dierential absorption is another problem that might inuence the accuracy of
the model. Because of the dierence in x-ray attenuation in the reference and the un-
known, the Kb to Ka ratios might be dierent in the two spectra. This becomes especially
problematic if the Kb line is above and the Ka line below an absorption edge of a major
element of the unknown sample. The magnitude of the error depends on the peak overlap.
Careful selection of the samples to produce the reference spectra is therefore required.
The procedure requires that a reference spectrum be included for each element
present in the unknown. The method allows no mechanism to deal with sum peaks. Apart
from removing the continuum, the lter also has some smoothing eect on the spectrum
and causes the peak structure to be spread out over more channels. This is equivalent to
tting a spectrum with a somewhat lower resolution than originally acquired. Therefore,
the precision and detection limits attainable with the lter-t method are slightly worse
than optimal. The width of the lter is important in this respect. Schamber (1977) suggests
taking the width of the top of the lter equal to the FWHM resolution of the spectrometer,
u =FWHM. The width of the wings can be taken as v = u=2.
2. Application
The calculation procedure is quite simple and requires the following steps. The top-hat lter
is applied to the unknown spectrumand the mreference spectra [Eqs. (30) and (31)], and the
modied uncertainties are calculated using Eq. (63). Next, the vector b of length m and the
mm square matrix a are formed using Eqs. (65) and (66), summing over the part of the
spectrum one wants to analyze (n
1
n
2
). Only the relevant part, such as the Ka or the Ka
plus the Kb, of the reference spectra needs to be retained; the rest of the ltered spectrum
can be set to zero. After calculating the inverse matrix a
1
, the contribution of each re-
ference to the unknown and its uncertainty are calculated using Eqs. (64) and (67) or (68).
A computer implementation of the lter-t method is given in Sec. X. The method
was used to analyze part of an x-ray spectrum of a polished NIST SRM 1103 brass sample
(Fig. 16A). The measurements were carried out using a Mo x-ray tube and a Zr secondary
target and lter assembly. Spectra of pure metals (Fe, Ni, Cu, Zn, and Pb) were acquired
Copyright 2002 Marcel Dekker, Inc.
under identical experimental conditions. A top-hat lter of width w = 5 was used. Table 6
shows how the spectra were divided in regions of interest to produce the reference spectra.
Because considerable x-ray attenuation is present in brass, separate references were created
for the Ka and Kb of Cu and Zn. This was not done for Fe, Ni, and Pb because these
elements are only present as minor constituents in the brass sample. Figure 16B shows the
ltered brass spectrum and the resulting t using the seven (ltered) reference spectra. The
region below CuKa is expanded 100 times, and the region above ZnKa is expanded 10
times. As can be seen, the agreement between the ltered brass spectrum and the t is very
good. The reduced w
2
value is 8.5. This high value is probably due to small peaks shifts in
the reference spectra compared to brass spectrum.
Table 7 compares the results of the lter t with the results obtained by nonlinear
least-squares tting using analytical functions (see Sec. VII). Although the w
2
value of the
nonlinear t is slightly better (2.7), one observes an excellent agreement between the two
methods for the analytical important data (i.e., intensity ratios). The uncertainties for
small peaks are slightly higher with lter-t method, as explained previously.
The lter-t method is fast and relatively easy to implement. It can produce reliable
results when the spectrometer calibration can be kept constant within a few electronvolts
and suitable standards for each element present in the sample are available. The method
Figure16 (A) Part of the x-ray spectrum of NIST SRM 1103 brass sample. (B) Top-hat-filtered
spectrum and result of fit using reference spectra.
Copyright 2002 Marcel Dekker, Inc.
performs well when one has to deal with a dicult to model continuum. If information on
trace elements and major elements is required (very large peaks next to very small ones),
the method might not be optimal. This lter-t method is frequently used to process x-ray
spectra obtained with electron microscopes (SEM-EDX), often in combination with a
ZAF or Phi-Rho-Z correction procedure.
B. Partial Least-Squares Regression
Spectrum evaluation as discussed in most of this chapter aims to obtain the net intensity of
the uorescence lines. These net peak intensities are then used to determine analyte con-
centrations using one of the many empirical, semiempirical, or fundamental approaches
detailed in various other chapters. In this, sub-section, we discuss an approach that is
relatively new for the XRF community and is based on multivariate calibration. We will
concentrate our discussion on partial least squares (PLS), a chemometrical technique that
is extensively used in infrared and relates spectrometric methods. The basic idea of the
method is to nd a (multivariate linear) model that directly relates the spectral data to the
analyte concentrations, thus avoiding the explicit evaluation of the spectrum in terms of
the net peak areas. The method involves a calibration step where a large number of
standards are used to build and validate the model and the actual analysis step where the
model is applied to spectra of unknown samples.
Table6 Data on the Reference Spectra and the Unknown Spectrum Used
in the Filter-Fit Procedure (Fig. 16)
Spectrum Region of interest (keV) Used as
Pure Fe 4.258.12 FeKa Kb reference
Pure Ni 4.739.57 NiKaKb reference
Pure Cu 5.708.46 CuKa reference
8.469.81 CuKb reference
Pure Zn 6.599.09 ZnKa reference
9.0910.30 ZnKb reference
Pure Pb 8.3613.68 PbLa reference
SRM 1103 5.2211.26 Unknown
Table 7 Comparison of Spectrum Evaluation Results Using Filter-Fit Method and
Nonlinear Least-Squares Fitting; Ratios of the Intensity from the SRM 1130 Standard to
the Pure Element Given
Ratio of intensity in SRM 1104 to pure element
Element Filter t Nonlinear t % Di
Fe 0.00530.0002 0.00470.0001 11
Ni 0.00190.0002 0.002010.00008 6
Cu 0.5460.001 0.5510.001 0.9
Zn 0.3900.001 0.39120.0008 0.31
Pb 0.03100.0006 0.03080.0003 0.65
Copyright 2002 Marcel Dekker, Inc.
1. Theory
Multivariate spectroscopic calibration attempts to predicts properties of samples (con-
centration of analytes) based on measured spectral data via the following relation:
Y = XB F (70)
The Y
nm
matrix holds concentration of m analytes in the n samples. The X
np
matrix
represents the spectral data, with n measured spectral data each having p x-ray intensities
(channels) and F
nm
is the matrix the residuals, part of the Y matrix not explained by the
model. The regression coecients B can be calculated in several ways. The most
straightforward approach is the use of multiple linear regression (see Sec. IX). The least-
squares solution is given then given by
B = (X
/
X)
1
X
/
Y (71)
When the number of x variables exceeds the number of samples ( p > n) and=or when there is
a high degree of correlation between the x variables, the least-squares solution becomes
unstable or cannot be obtained due to the fact that the covariance matrix X
/
X cannot be
inverted. This is exactly the fact when we attempt to apply multivariate calibration in XRF.
The number of channels in the spectrum (1024 or more) largely exceeds the number of
samples andthe intensities of neighboring channels in each peak are very strongly correlated.
Toovercome this problem, the methodof singular value decompositioncanbe used. The
X matrix is decomposed in a number of linearly independent variables, know as principal
components. In principle, there are as many principal components as there are original vari-
ables (channels), but only the most signicant [i.e., having the highest variance (or eigenvalue)]
are retained. Using the matrix of this smaller number of principal components scores, rather
than the entire Xmatrix in Eq. (70), is known as principal components regression (PCR). The
disadvantage of PCR is that the selection of the principal components is based on how much
variance they explain of the X matrix. The rst few principal components may have little re-
lation with the concentrations that need to be predicted by the model.
Partial least-squares regression (PLSR) is a variant of PCR that largely overcomes
this problem. The methods uses two outer relations and one inner relation. The outer
relations describe the decomposition of the X and Y matrices:
X = TP
/
E =
A
a=1
t
a
p
/
a
E (72)
Y = UQ
/
F =
A
a=1
u
a
q
/
a
F (73)
T
nA
and U
nA
are the score matrices, the values of the A(A p) latent variables. P
/
Ap
and Q
/
Am
are the loading matrices describing the relation between the latent variables
(T and U) and the original variables (X and Y). The number of latent variables A in the
model is of crucial importance and its optimum value must be found by cross-validation.
The matrices E and F contain the residuals, part of the original spectral data and the
concentration data, respectively, not accounted for when using A latent variables. The
inner relation is written as:
u
a
= b
a
t
a
; a = 1; . . . ; A (74)
fromwhich the regression coecients Bcan be obtained. This operation can be seen as a least-
squares t between the Xblock and Yblock scores. The nal PLS model can thus be written as
Copyright 2002 Marcel Dekker, Inc.
Y = TBQ
/
F
+
(75)
A graphical representation of the PLS model is given in Figure 17. In the normal PLS
algorithm Y is a vector of concentrations for one element and a separate model is build
for each element. If all Y-variables are predicted simultaneously, as in the case of
Equation (70), the PLS2 algorithm is used. This method performs better when the con-
centrations in the samples are highly correlated.
The quality of the calibration model can be judged by the root mean square error of
prediction (RMSEP):
RMSEP =
1
n
n
i=1
(^ y
i
y
i
)
2
_ _
1=2
(76)
where ^ y
i
is the concentration in sample i predicted by the model y
i
is the true concentration
in the standard. To determine the optimum number of latent variables, the RMSEP is
calculated using PLS models with dierent numbers of latent variables A. The RMSEP
values are plotted against A and the value where a minimum or a plateau is reached is
taken. For small datasets, the calculation of the RMSEP is done using cross-validation.
When a large number of standards are available, they are split in a training set (ap-
proximately two-thirds of the samples) and a prediction set (approximately one-third of
the samples). The training set is used to build the PLS model and the RMSEP is calculated
based on the concentrations of the prediction set. Alternatively, for smaller calibration
sets, leave-one-out cross-validation can be used. Each sample is excluded once from the
dataset and predicted by the model built with the remaining samples. This is repeated until
all samples have been excluded once.
Geladi and Kowalski (1986) published a tutorial on PLS and its relation to other
regression techniques. A standard work on the PLS method is the book by Martens and
Figure17 Graphical representation of the PLS algorithm.
Copyright 2002 Marcel Dekker, Inc.
Naes (Martens and Naes, 1989). Theoretical and statistical aspects of the method can be
found in chemometrics literature (Manne, 1987; Lorber et al., 1987; Pratar and De Jong,
1997; Hoskuldsson, 1988).
2. Application
The PLS method is illustrated with the analysis of aqueous solution containing Ni, Cu,
and As with concentration in the ranges 1850, 0.545, and 520 g=L, respectively,
whereas Zn, Fe, and Sb were present in more or less constant amounts. Spectra were
acquired for 1000 s from 5-mL solution using an EDXRF spectrometer equipped with a
Rh x-ray tube operating at 30 kV and a Si(Li) detector having 160 eV resolution at MnKa.
From the 22 samples, 10 were used to build the PLS mode and the remaining were
used as the prediction set. Table 8 gives the composition of the samples in the calibration
set. Figure 18 shows part of the spectrum between 5 and 14 keV of sample number 9. The
CuKa line is considerably interfered by the NiKb line. Absorption eects can be expected
because the NiK absorption edge (8.33 keV) is just above the CuKa line (8.04 keV) and the
AsK lines can cause secondary uorescence of Cu and Ni.
A PLS 1 model is build for Cu. The X matrix consists of 451 x-ray intensities
(variables) between 5 and 14 keV (channels 250 to 700) of the 10 samples. The Y matrix
contains the Cu concentration of those samples.
In Figure 19, the RMSEP based on leave-one-out cross-validation is plotted as
function of the number of latent variables. A minimum error of 0.57 g=L is obtained for
three latent variables so that the PLS model with three latent variable is retained.
Figure 20 compares the true Cu concentration with the predicted concentration for the
calibration set. The PLS model predicts very accurately the Cu concentrations in the
range from 1 to 40 g=L. The so-called regression coecients of the PLS model show
which variables (channels) are used in the model. They are plotted in Figure 21. As could
be expected, the Cu concentrations are predicted from the channel content corresponding
to the CuK lines. The inuence of absorption and enhancement results in a small negative
contribution from the Ni peak and a small positive contribution from the As peaks, re-
spectively. The PLS model thus handles both the problem of spectral interference and
matrix eects. The Cu concentration predicted by the PLS model for the test set is given in
Table 9. Except for the two samples with the highest concentrations, the Cu concentration
Table 8 Composition of Samples Used to Build the PLS Model
Concentration (g=L)
Sample No. Ni Cu Zn
1 18.0 6.7 6.1
2 18.1 5.0 6.0
3 18.5 3.5 6.1
4 18.1 8.4 5.7
5 53.6 1.8 20.1
6 53.8 0.8 19.8
7 57.3 1.1 19.3
8 56.9 1.5 20.0
9 51.4 17.7 20.9
10 17.9 41.5 5.3
Copyright 2002 Marcel Dekker, Inc.
in the unknown samples is very well estimated. The predicted of the test set is generally
somewhat worse that the prediction of the calibration set. To build an accurate model, a
large number of standards spanning the concentration range of interest for each element is
required. This is certainly the major drawback of PLS for its application in XRF,
Figure18 X-ray spectrum of sample number 9 in the calibration set used to build the PLS model.
Figure 19 RMSEP values versus the number of latent variables for the prediction of the Cu
concentration.
Copyright 2002 Marcel Dekker, Inc.
Figure 20 Comparison of true and predicted Cu concentrations for the samples in the PLS
calibration set.
Figure 21 Regression coefficients of the PLS model for Cu, showing which variables (channels in
the x-ray spectrum) are used to predict Cu.
Copyright 2002 Marcel Dekker, Inc.
especially when solid samples are considered. This problem can to some extend be
overcome by building a calibration set via Monte Carlo simulation. Just as with the lter-
t method, standards and unknowns need to be measured under strictly identical spec-
trometer conditions. Changes in gain or resolution will cause systematic errors in the
calculated concentrations.
Swerts and Van Espen (1993) demonstrated the use of PLS for the determination of
S in graphite using a Rh x-ray tube excitation XRF equipped with a Si(Li)detector. Be-
cause of diraction eects, least-squares tting of the spectra was nearly impossible. Using
PLS, the sulfur content could be determined in a concentration range 260%, with an
accuracy of better than 5% relative standard deviation. Urba nski and Kowalska applied
the PLS method to the determination of Sr and Fe in powdered rock samples and to the
determination of the S and ash content in coal using a radioisotope XRF system equipped
with a low-resolution gas proportional counter. They also demonstrated the usefulness of
this method for the determination of the thickness and composition of Sn-Pb and Ni-Fe
layered structures (Urba nski and Kowalska, 1995). Molt and Schramm (1987) compared
principal components regression (PCR), PLS for the determination of S, exhibiting strong
interference from Mo, in aqueous and solid samples. The results were also compared with
quantitative analysis using the method developed by Lucas-Tooth and Price (1961).
Equally good results were obtained with all three methods. Similar results were obtained
by Lemberge and Van Espen for the determination of Ni, Cu, and As in liquid samples.
They demonstrated that taking the square root of the data improves the PLS model and
that the PLS method extracts information from the scattered excitation radiation to de-
scribe the matrix eects (Lemberge and Van Espen, 1999).
VII. LEAST-SQUARES FITTINGUSINGANALYTICAL FUNCTIONS
A widely used and certainly the most exible procedure for evaluating complex x-ray
spectra is based on least-squares tting of the spectral data with an analytical function.
The method is conceptually simple, but not trivial to implement and use.
Table 9 Comparison of True and Predicted Cu Concentration Using the PLS Model
Cu concentration (g=L)
Sample No. True Predicted Dierence
11 6.30 6.38 0.08
12 7.20 7.34 0.14
13 6.30 6.32 0.02
14 7.50 7.40 0.10
15 6.30 6.09 0.21
16 0.90 1.27 0.37
17 1.00 0.75 0.25
18 1.20 1.55 0.35
19 1.80 2.11 0.31
20 2.50 2.86 0.36
21 17.60 16.51 1.09
22 44.70 40.59 4.11
Copyright 2002 Marcel Dekker, Inc.
A. Concept
In this method, an algebraic function, including analytical importantly parameters such as
the net areas of the uorescence lines, is used as a model for the measured spectrum. The
object function (w
2
) is dened as the weighted sum of squares of the dierences between
this model y(i) and the measured spectrum y
i
over a region of the spectrum:
w
2
=
n
2
i=n
1
1
s
2
i
y
i
y(i; a
1
; . . . ; a
m
) [ [
2
(77)
where s
2
i
is the variance of data point i, normally taken as s
2
i
= y
i
, and a
j
are the para-
meters of the model. The optimum values of the parameters are those for which w
2
is
minimal. They can be found by setting the partial derivatives of w
2
with respect to the
parameters to zero:
@w
2
@a
j
= 0; j = 1; . . . ; m (78)
If the model is linear in all the parameters a
j
, these equations result in set of m linear
equations in the m unknowns a
j
, which can be solved algebraically. This is known as linear
least-squares tting. If the model is nonlinear in one or more of its parameters, a direct
solution is not possible and the optimum value of the parameters must be found itera-
tively. An initial value is given to the parameters and they are varied in some way until a
minimum for w
2
is obtained. The latter is equivalent to searching for a minimum in the
m1-dimensional w
2
hypersurface. This is known as nonlinear least-squares tting. The
selection of a suitable minimization algorithm is very important because it determines to a
large extent the performance of the method. A detailed discussion of linear and nonlinear
least-squares tting is given in Sec. IX.
The most dicult problem to solve when applying this least-squares procedure is the
construction an analytical function that accurately describes the observed spectrum. The
model must be capable of describing accurately the spectral data in the tting region. This
requires an appropriate model for the continuum, the characteristic lines of the elements
and for all other features present in the spectrum such as absorption edges, escape, and
sum peaks. Although the response function of the energy-dispersive detector is, to a very
good approximation, Gaussian, deviation from the Gaussian shape needs to be taken into
account. Failure to construct an accurate model will result in systematic errors, which
under certain conditions may lead to gross positive or negative errors in the estimated
peak areas. On the other hand, the tting function should remain simple, with as few
parameters as possible. Especially for the nonlinear tting, the location of the w
2
minimum
becomes problematic when a large number of parameters is involved.
In general, the tting model consists of two parts:
y(i) = y
B
(i)
P
y
P
(i) (79)
where y(i) is calculated content of channel i and the rst part describes the continuum and
the second part the contributions of all peaklike structures.
Because the tting functions for both linear and nonlinear least-squares tting have
many features in common, we treat the detailed description of the tting function for the
most general case of nonlinear least squares. Moreover, if the least-squares t is done
using the Marquardt algorithm, the linear least-squares t is computationally a particular
case of the nonlinear least-squares t. Programs based on this algorithm can perform
Copyright 2002 Marcel Dekker, Inc.
linear and nonlinear tting using the same computer code. A large part of the discussion
given here is based on the computer code AXIL, developed by the author for spectrum
tting of photon-, electron-, and particle-induced x-ray spectra (Van Espen et al., 1977a,
1977b, 1979b, 1986).
B. Description of the Continuum
To model the continuum, various analytical expressions are in use, depending on the
excitation conditions and on the width of the tting region. Except for electron micro-
scopy, it is virtually impossible to construct an acceptable physical model that describes
the continuum, mainly because of the large number of processes that contribute to it. For
this reason, very often some type of polynomial expression is used.
1. Linear Polynomial
A linear polynomial of the type
y
B
(i) = a
0
a
1
(E
i
E
0
) a
2
(E
i
E
0
)
2
a
k
(E
i
E
0
)
k
(80)
is useful to describe the continuum over a region 23 keV wide. Wider regions often
exhibit too much curvature to be described by this type of polynominal. In Eq. (80), E
i
is
the energy (in keV) of channel i [see Eq. (84)] and E
0
is a suitable reference energy, often
the middle of the tting region. Expressing the polynomial as a functin of (E
i
7E
0
) rather
than as a function of the channel number is done for computational reasons; (E
i
7E
0
)
3
is,
at most, of the order of 10
3
, whereas i
3
can be as high as 10
9
. Most computer programs
that implement a polynomial model for the continuum allow the user to specify the degree
of the polynomial; k =0, 1, and 2 produce respectively a constant, a straight line, and a
parabolic continuum. Values of k larger than 4 are rarely useful because such high-degree
polynomials tend to have physical nonrealistic oscillations. Equation (80) is linear in the
tting parameters a
0
, . . . , a
k
, so that this function can be used in linear as well as in
nonlinear least-squares tting.
2. Exponential Polynomial
A linear polynomial cannot be used to t the continuum over the entire spectrum or to t
regions of high positive or negative curvature. Higher curvature can be modeled by
functions of the type
y
B
(i) = a
0
exp[a
1
(E
i
E
0
) a
1
(E
i
E
0
)
2
a
k
(E
i
E
0
)
k
[ (81)
where k is the degree of the exponential polynomial. A value of k as high as 6 might be
required for an accurate description of a continuum from 2 to 16 keV. This function is
nonlinear in the tting parameters a
1
, . . . , a
k
and requires a nonlinear least-squares pro-
cedure and some initial guess of these parameters. Initial values for these nonlinear
parameters can be determined by rst estimating the shape of the continuum using one of
the procedures described in Sec. IV followed by a linear t of the logarithm of this con-
tinuum. These initial guesses are then further rened in the nonlinear tting procedure.
3. Bremsstrahlung Continuum
The exponential polynomial is not suitable for describing the shape of the continuum
observed in electron- and particle-induced x-ray spectra, mainly because of the high
Copyright 2002 Marcel Dekker, Inc.
curvature at the low-energy region of the spectrum. The continuum results from photons
emitted from the sample by the retardation of fast electrons. The slope of the emitted
continuum is essentially an exponentially decreasing function according to Kramers for-
mula. At low energies, the emitted photons are strongly absorbed by the detector windows
and by the sample. A suitable function to describe such a radiative continuum is an
exponential polynominal multiplied by the absorption characteristics of the spectrometer:
y
B
(i) = a
0
exp[a
1
(E
i
E
0
) a
k
(E
i
E
0
)
k
[T
a
(E
i
) (82)
A detailed discussion of the function T
a
(E) is given on page 288. To be physically correct,
the absorption term must be convoluted with the detector response function, because
the sharp edges due to absorption by detector windows (Si or Au) or elements present in
the sample are smeared out by the nite resolution of the detector.
4. ContinuumRemoval
An alternative to an algebraic description of the continuum is to estimate the continuum
rst using one of the procedures outlined in Sec. IV and to substract this continuum from
the measured spectrum before the actual least-squares tting. To correctly implement the
least-squares t after subtraction of the continuum, the weights 1=s
2
i
[Eq. (77)] must be
adjusted. If y
/
i
represents the spectral data after subtraction of the continuum,
y
/
i
= y
i
y
B
(i), the variance of y
/
i
is given by s
/ 2
i
= s
2
i
s
2
y
B
(i)
. A reasonable approximation
for s
2
y
B
(i)
is y
B
(i) itself, so that the total variance becomes s
/ 2
i
= y
i
y
B
(i). If this adjust-
ment of the weights is not made, the uncertainty in the net peak areas are understimated,
especially for small peaks on a high continuum.
It is rather dicult for an inexperienced user to select the appropriate continuum
model for a given spectrum. The following might serve as a general guideline. For tting
regions 23 keV wide, a linear polynomial continuum is often adequate. To t large re-
gions of XRF spectra, the exponential polynomial provides the most accurate results, with
k typically, between 4 and 6. The same holds for the bremsstrahlung continuum of SEM-
EDX and PIXE spectra. The simplest method from the users point of view is the
continuum stripping, but this method does not provide optimum results. A slight under-
estimation or overestimation might occur, resulting in large relative errors in the area
determination of small peaks (Vekemans et al., 1995).
C. Description of Fluorescence Lines
Because the response function of most solid-state detectors is predominantly Gaussian, all
mathematical expressions used to describe the uorescence lines involve this function.
When dealing with the K lines of high-atomic-number elements, such as Pb or U, the in-
uence of the natural line shape becomes appreciable and the use of the more complicated
Voigt prole is required (Wilkinsin, 1971; Gunnink, 1977; Pessara and Debertin, 1981).
1. Single Gaussian
A Gaussian peak is characterized by three parameters: the position, width, and height or
area. It is desirable to describe the peak in terms of its area rather than its height because
the area is directly related to the number of x-ray photons detected, whereas the height
depends on the spectrometer resolution. The rst approximation to the prole of a single
peak is then given by
Copyright 2002 Marcel Dekker, Inc.
A
s
2p
_ exp
(x
i
m)
2
2s
2
_ _
(83)
where A is the peak area (counts), s is the width of the Gaussian expressed in channels,
and m is the location of the peak maximum. The often used FWHM is related to s by the
factor 2
2 ln 2
_
or FWHM=2.35s. In Eq. (83), the peak area is a linear parameter; the
width and position are nonlinear parameters. This implies that a nonlinear least-squares
procedure is required to nd optimum values for the latter two parameters. Using a linear
least-squares method assumes that the position and width of the peak are know with high
accuracy from calibration.
To describe part of a measured spectrum, the tting function must contain a number
of such functions, one for each peak. For 10 elements and 2 peaks (Ka and Kb) per element,
we would need to optimize 60 parameters. It is highly unlikely that such a nonlinear least-
squares t will terminate successfully at the global minimum. To overcome this problem,
the tting function can be written in a dierent way as shown in the next subsection.
2. Energyand Resolution Calibration Function
The rst obvious step is to drop the idea of optimizing the position and width of each peak
independently. In x-ray spectrometry, the energies of the uorescence lines are know with
an accuracy of 1 eV or better. The pattern of peaks observed in the spectrum is directly
related to elements present in the sample. Based on those elements, we can predict all x-ray
lines that constitute the spectrum and their energies. The peak function [Eq. (83)] is
therefore written in terms of energy rather than channel number. Dening ZERO as the
energy of channel 0 and expressing the spectrum GAIN in electronvolts=channel, the
energy of channel i is given by
E(i) = ZERO (GAIN)i (84)
and the Gaussian peak can be written as
G(i; E
j
) =
GAIN
s
2p
_ exp
[E
j
E(i)[
2
2s
2
_ _
(85)
with E
j
the energy (in eV) of the x-ray line and s the peak width given by
s
2
=
NOISE
2:3548
_ _
2
3:58(FANO)E
j
(86)
In this equation, NOISE is the electronic contribution to the peak width (typically
80100 eV FWHM) with the factor 2.3548 to convert to s units, FANO is the Fano factor
( ~0.114), and 3.85 is the energy required to produce an electronhole pair in silicon. The
term GAIN=s
2p
_
in Eq. (85) is required to normalize the Gaussian so that the sum over all
channels is unity.
For linear least-squares tting, ZERO, GAIN, NOISE, and FANO are physically
meaningful constants. In the case of nonlinear least squares, they are parameters to be
rened during the tting. The advantage of optimizing the energy and resolution calibra-
tion parameters rather than the position and width of each peak is a vast reduction of the
dimensionality of the problem. The nonlinear t of 10 peaks would now involve 14 para-
meters compared to 30. Even more importantly, all information available in the spectrumis
used to estimate ZERO, GAIN, NOISE, and FANO and thus the positions and the widths
Copyright 2002 Marcel Dekker, Inc.
of all peaks. Imagine a small, severely overlapping doublet with a well-dened peak on both
sides of this doublet. These two peaks will contribute most to the determination of the four
calibration parameters, virtually xing the position and the width of the two peaks in the
doublet. As a consequence, their areas can be determined much more accurately.
Referring to our discussion on information content in Sec. II, we did not obtain this
extra performance for nothing. We have supplied extra information: the energy of the
peaks and the two calibration relations [Eqs. (84) and (86)]. Fitting with Eq. (85) requires
that the extra information we supply is indeed correct.
With modern electronics, the linearity of the energy calibration [Eq. (84)] holds very
well in regions above 2 keV. Fitting the entire spectrum, including the low-energy region,
might require more complex energy calibration functions. To t PIXE spectra from 1 to
30 keV, Maenhaut and Vandenhaut (Maenhaut and Vandenhaut, 1986) suggested the
following function: i = C
1
C
2
E C
3
exp(C
4
E).
The relation between the square of the peak width and the energy [Eq. (86)] is based
on theoretical considerations. The relation holds very well if the doublet splitting of the
x-ray lines is taken into account. The Ka
1
Ka
2
separation increases from a negligible value
for Ca (3.5 eV) to nearly 100 eV for Mo. The observed peak shape of a K line is actually an
envelope of two peaks. This envelope can be represented rather well by a single Gaussian,
but failing to take this doublet splitting into account (i.e., tting with a single Gaussian
where doublets are required) will result in peak widths that do not obey Eq. (86). To
illustrate this, the observed width of a number of Ka lines as function of the x-ray energy is
presented in Figure 22. The dotted line represents the width of the Ka doublet tted as one
peak. The solid (straight) line shows the width of the individual lines in the doublet.
3. Response Function for an Element
A second modication to the tting function that will reduce the number of tting
parameters is modeling an entire element, rather than single peaks. A number of lines can
be considered as logically belonging together, such the Ka
1
and Ka
2
of the above-men-
tioned doublets or all K lines of an element. This group can be tted with one area
parameter A representing the total number of counts of all the lines in the group.
The spectrum of an element can then be represented by
y
P
(i) = A
N
P
j=1
R
j
G(i; E
j
) (87)
where G are the Gaussians for the various lines with energy E
j
and R
j
the relative in-
tensities of the lines. The summation runs over all lines in the group (N
P
) with
R
j
= 1.
The transition probabilities of all lines originating from a vacancy in the same (sub)
shell (K, L
I
, L
II
, . . .) are constants, independent of the excitation. However, the relative in-
tensities depend on the absorption in the sample and in the detector windows. To take this
into account, the x-ray attenuation must be included in Eq. (87). The relative intensity ratios
are obtained by multiplying the transition probabilities with an absorption correction term:
R
/
j
=
R
j
T
a
(En
j
)
j
R
j
T
a
(En
j
)
(88)
Contributions from various subgroups (i.e., between K and L, between L
I
and L
II
, etc.)
depend on the type of excitation (photons, electrons, protons) and on the excitation en-
ergy. General values cannot be given and must be determined for the particular excitation
Copyright 2002 Marcel Dekker, Inc.
if one wants to combine lines of dierent subgroups. Transition probabilities of lines in
various groups, determined experimentally and calculated from rst principles, can be
found in the literature (McCrary et al., 1971; Salem et al., 1970; Salem and Wimmer, 1970;
Scoeld 1970, 1974a, 1974b).
In Figure 23, the t of a tungsten L line spectrum using 24 transitions from the 3 L
subshells is shown. The relative intensities of the L lines within each subshell were taken
from the literature (Scoeld, 1974b). The t was thus done with one peak-area parameter
for the L
1
, the L
2
and the L
3
sublevels.
Fitting elements rather than individual peaks enhances the capability of the method
to resolve overlapping peaks. The area of the CrKa peak, interfered by a VKb peak, can be
obtained with higher precision (lower standard deviation) because the area of the VKb
peak is related to the area of the (interference free) VKa peak. Again, we have introduced
more a priori knowledge into our model. If this information (the relative intensity ratio) is
not correct, we will introduce systematic errors (e.g., the CrKa peak area, although having
a low standard deviation, will not be correct). In practice, there will be a trade-o between
the gain in accuracy and the gain in precision. Errors in the values of the transition
probabilities and in the value of the absorption correction term are suciently small so
that small to moderate high peaks (up to 10
5
counts peak area) can be tted as one group.
4. Modified Gaussians
When tting very large peaks with a Gaussian, the deviation from the pure Gaussian shape
becomes signicant. In Figure 24, a MnK spectrum with 10
7
counts in the MnKa peak is
shown. One observes a tailing on the low-energy side of the peaks and a continuum that is
Figure 22 FWHM of various Ka lines, tted as single peak and as a Ka
1
Ka
2
doublet.
Copyright 2002 Marcel Dekker, Inc.
Figure23 Fit of a complex L line spectrum of tungsten. In total, 24 transitions, divided over the 3
L subshells, are required for the description on the spectrum.
Figure 24 MnK line spectrum with very good counting statistics. The dierence from the
Gaussian response is obtained by subtracting all Gaussian peaks. From this, the peak-shape
correction is calculated.
Copyright 2002 Marcel Dekker, Inc.
higher at the low energy side of the Ka peak than at the high energy side of the Kb peak.
The observed peak shape is partially caused by nonideal behavior of the detector. For low-
energy lines ( <10 keV), incomplete charge collection and other detector artifacts play an
important role. For higher energetic lines, Compton scattering in the detector also con-
tributes. Another part of the deviation from the Gaussian shape is attributed to phe-
nomena taking place in the sample itself. X-ray satellite transitions such as KLM radiative
Auger transitions on the low-energy side of the Ka and peak and KMM transitions on the
low-energy side of the Kb contribute to the overall shape of the spectrum. In the literature,
considerable attention has been given to the explanation and to the accurate mathematical
description of the observed peak shape (McNelles and Compbell, 1975; Van Espen et al.,
1977b; Wielopolski and Gardner, 1979; Campbell et al., 1985, 1987, 1997, 1998, Gardner
et al., 1986; Yacout et al., 1986; Campbell, 1996).
Failure to account for the deviation from the Gaussian peak shape causes a number
of problems when tting x-ray spectra. Small peaks sitting on the tail of large ones (e.g.,
NiKa in front of CuKa) cannot be tted accurately, resulting in large systematic errors for
the small peak. Because the least-squares method seeks to minimize the dierence between
the observed spectrum and the tted function, the tail might become lled up with peaks of
elements that are not really present. Also, the continuum over the entire range of the
spectrum becomes dicult to describe. This is illustrated in Figure 25a, where a spectrum
Figure 25 Fit of the spectrum of a brass sample (NIST SRM 1106) (a) fitted with a simple
Gaussians and (b) fitted with Gaussians including tail and step functions.
Copyright 2002 Marcel Dekker, Inc.
of a brass sample (NIST SRM 1106) is tted with simple Gaussians and a linear con-
tinuum. An unrealistically high area for the NiKa peak is obtained in this way.
A number of analytical functions have been proposed to account for the true line
shape. Nearly all of them include a at shelf and an exponential tail, both convoluted with
the Gaussian response function. Similar functions are also used to describe the peak shape
of g-ray spectra (Phillips and Marlow, 1976).
To account for the deviation from the Gaussian peak shape, the Gauss function G(i,
E
j
) in Eq. (85) can be replaced by
P(i; E
jk
) = G(i; E
jk
) f
S
S(i; E
jk
) f
T
T(i; E
jk
) (89)
where G(i, E
j
) is the Gaussian part given by Eq. (85) and S and T are the step and tail
functions, respectively,
S(i; E
jk
) =
GAIN
2E
jk
erfc
E(i) E
jk
s
2
_
_ _
(90)
T(i; E
jk
) =
GAIN
2gs exp[1=(2g
2
)[
exp
E(i) E
jk
gs
_ _
erfc
E(i) E
jk
s
2
_
1
g
2
_
_ _
(91)
In these equations, s represents spectrometer resolution [Eq. (86)] and g is the broadening
of the exponential tail. The parameters f
S
and f
T
in Eq. (89) describe the fraction of the
photons that arrive in the step and the tail, respectively. The complement of the error
function used to convolute the step and the tail is dened as
erfc(u) = 1 erf(u) = 1
2
p
_
_
u
0
e
t
2
dt (92)
and can be calculated via series expansion (Press et al., 1988). Figure 25b shows a t of a
brass spectrum when these step and tail functions are included. In this case, a much more
realistic area for the NiKa is obtained and the continuum is described more correctly.
Campbell et al. (1987) used a Gaussian, a short-tail, and a long-tail exponential to t
the peak in energy-dispersive spectra obtained from Ka
1
and La
1
lines selected by Bragg
reection from a curved crystal. In this way, the inuence of the doublet structure and the
satellite lines is eliminated. Excellent ts with reduces chi-square values between 1.02 and
1.16 were obtained for peaks having 10
6
counts.
Fitting real x-ray spectra with modied Gaussians as given by Eq. (89) increases
dramatically the number of parameters that need to be optimized ( f
S
, f
T
, g) for each peak.
The phenomena contributing to the deviation from the Gaussian shape depend on the
energy of the detected x-ray line, so that these parameters can be expressed as smooth
functions of energy (Wielopolski and Gardner, 1979; Yacout et al., 1986). These relations
can be established using pure element spectra with very good counting statistics. An ad-
ditional problem is that the tailing of the Kb peaks tend to be larger due to the more
intense KMM radiative transition, thus requiring a separate relation for Ka and Kb tails.
A further modication of the tting function involves the use of a Voigt prole to
account for the natural line width of the x-ray lines. This becomes important when tting K
lines of higher-Z elements (Ba and above) where the natural linewidth becomes consider-
able ( >20 eV) compared to the detector resolution. The peak prole, G(i; E
j
) in Eq. (89)
must then be replaced by a Voigt prole (the convolution of a Lorenzian and a Gaussian):
GAIN
s
2p
_ K
E(i) E
jk
s
2
_ ;
a
L
2s
2
_
_ _
(93)
with Kthe real part of the complex error function and a
L
the natural width of the x-ray line:
Copyright 2002 Marcel Dekker, Inc.
K(x; y) = Ree
z
2
erfc(iz)|; z = x iy (94)
Voigt proles can be calculated using numerical approximations (Schreier, 1992). The
imaginary part of the complex error function can be used to calculate the derivatives
necessary in the least-squares-tting algorithm.
An alternative procedure to describe the peak shape is used in the AXIL program
(Van Espen et al., 1977b). The deviation from the Gaussian shape is stored as a table on
numerical values, representing the dierence of the observed shape and the pure Gaussian.
The table extends from zero energy up to the high-energy side of the Kb peak and is
normalized to the area of the Ka peak. The deviation is obtained from pure element
spectra having very good counting statistics. (Ka ~10
7
counts). Preferably, thin lms are
used to keep the continuum as low as possible and to avoid absorption eects. The area,
position, and width of all peaks in the spectrum are determined by tting Gaussians on a
constant continuum over the full width at tenth maximum (FWTM) of the peaks. The
Gaussian contributions are then stripped from the spectrum. The resulting non-Gaussian
part as shown in Figure 24 is further smoothed to reduce the channel to channel uc-
tuations and subsequently used as a numerical peak-shape correction. The tting function
for an element is then given by
y
P
(i) = A R
/
Ka
G(i; E
Ka
) C(i) [ [
N
P
j=2
R
/
j
G(i; E
j
)
_ _
(95)
where C(i) is the numerical peak-shape correction at channel i. Values in the table are
interpolated to account for the dierence between the energy scale of the correction and
the actual energy calibration of the spectrum. Similar to the parameters of the non-
Gaussian analytical functions, the shape of the numeric correction seems to vary slowly
from one element to another. This allows us to interpolate the peak-shape correction for
all elements from a limited set of experimentally determined corrections.
A major disadvantage of this method is that it is quite dicult and laborious to
obtain good experimental peak-shape corrections. Although they are, in principle, de-
tector dependent, experience has proven that the same set of corrections can be used for
dierent detectors with reasonable success, proving the fundamental nature of the ob-
served non-Gaussian shape. Another disadvantage is that the peak-shape correction for
the K becomes underestimated if strong dierential absorption takes place because the
peak-shape correction is only related to the area of the Ka peak. Also, it is nearly im-
possible to apply this method to the description of L-line spectra. A mayor advantage
however is the computational simplicity of the method and the fact that no extra para-
meters are required in the model.
5. Absorption Correction
The absorption correction term T
a
(E), used in the Eqs. (82) and (88), includes the x-ray
attenuation in all layers and windows between the sample surface and the active area of the
detector. For high energetic photons also, the transparency of the detector crystal needs to be
taken into account. In x-ray uorescence, the attenuation in the sample, causing additional
changes in the relative intensities, can also be considered, providing the sample composition
is known. The total correction term is thus composed of a number of contributions:
T
a
(E) = T
Det
T
Path
T
Sample
(96)
The detector contribution for a Si(Li) detector is given by
Copyright 2002 Marcel Dekker, Inc.
T
Det
(E) = e
m
Be
(rd)
Be
e
m
Au
(rd)
Au
e
m
Si
(rd)
Si
(1 e
m
Si
(rD)
Si
) (97)
where m, r, and d are the mass-attenuation coecient, the density, and the thickness of the
Be window, the gold contact layer, and the silicon dead layer. In the last term, D is the
thickness of the detector crystal.
Any absorption in the path between the sample and the detector can be modeled in a
similar way. For an air path, the absorption is given by
T
Path
(E) = e
m
air
(rd)
air
(98)
The mass-attenuation coecient of air can be calculated assuming a composition of 79.8%
N
2
, 19.9% O
2
, and 0.3% Ar. The density of air in standard conditions is 1.2610
73
g=cm
3
.
In PIXE setups, it is common practice to use additional absorbers between the sample and
the detector, with one absorber sometimes having a hole (funny lter). The absorption
behavior of such a structure can be modeled by
T
Path
(E) = e
m
f
(rd)
f
[h (1 h)e
m
ff
(rd)
ff
[ (99)
where the rst term accounts for the absorption in (all) solid lters and the second term
accounts for the absorption of the lter having a hole. The fraction of the detector solid
angle subtended by the hole is denoted by h.
The sample absorption correction in the case of x-ray uorescence, assuming ex-
citation with an (equivalent) energy E
0
, is given by
T
sample
(E) =
1 e
w
s
(rd)
s
w
s
(rd)
s
(100)
and the sample attenuation coecient is given by
w
s
=
m
s
(E)
sin y
1
m
s
(E
0
)
sin y
2
(101)
This sample attenuation coecient can only be calculated if the weight fractions of the
constituting elements are known, which is often not the case because the aim of the
spectrum evaluation is to obtain the net peak areas from which the concentrations are to
be calculated. Although x-ray attenuation in solid samples might become very large, it is
important to realize that not the absolute values of the absorption corrections but their
ratio is of importance in Eq. (88). This ratio changes less dramatically especially because
the energy dierence of the lines involved is small. For an innitely thin Fe sample, the
Kb=Ka intensity ratio is 0.134. In an innitely thick Fe matrix, the absorption correction
term T
a
(E
Kb
) and T
a
(E
Ka
) are respectively 161 and 134, assuming MoKa excitation,
causing the Kb=Ka ratio to change to 0.161. Therefore, a rough estimate of the sample
composition is often sucient. Van Dyck and Van Grieken (1983) demonstrated the
integration of spectrum analysis with quantitative analysis. The sample absorption term
is recalculated based on the estimated sample composition, and the spectrum tting is
repeated.
6. Sumand Escape Peaks
As indicated in Sec. II, Si escape peaks can be modeled by a Gaussian with an energy
1.750 keV below the parent peak. The area relative to the area of the parent peak can be
calculated from the escape fraction [Eq. (4)]:
Copyright 2002 Marcel Dekker, Inc.
Z =
N
e
N
p
=
f
1 f
(102)
Including the escape peaks, the description of the uorescence of an element becomes
y
P
(i) = A
N
P
J=1
R
/
j
[G(i; E
j
) ZG((i; E
j
1:750)[ (103)
Also, various polynomial type functions expressing the escape ratio as a function of the
energy of the parent peak are in use. The coecients of the function can be determined by
least-squares tting from experimental escape ratios. For spectra obtained with a Ge
detector, one needs to account in a similar way for both the GeKa and GeKb escape peak
for elements above arsenic.
The incorporation of the sum peaks in the tting model is more complex. The
method discussed below was rst implemented by Johansson in the HEX program
(Johansson, 1982). Initially, the spectrum is tted without considering pileup peaks. The
peaks are then sorted according to their height and the n largest peaks are retained. Peaks
that diers less than 50 eV are combined to one peak. Using Eqs. (5) and (6), the relative
intensities of all possible n(n 1)=2 pileup peaks and their energies are calculated and the
m most intense are retained. Knowing the relative intensities and the energies of these m
pileup peaks, they can be included in the tting model as one pileup element. In the next
iteration, the total peak area A of this pileup element is obtained. The construction of the
pileup element can be repeated during the next iterations as more reliable peak areas
become available. In Figure 26, part of an PIXE spectrum is shown tted with and without
sum peaks included.
D. Special Aspects of Least-Squares Fitting
1. Constraints
When using nonlinear least-squares tting, it can be advantages to impose limits on the
tting parameters to eliminate physically meaningless results. Some illustration of what
can happen if too much freedom is given to the tting parameters is provided by Statham
(1978).
The incorporation of the energy and resolution calibration functions [Eqs. (84) and
(86)] in the tting model already places a severe constraint on the t, avoiding the situation
in which two peaks that are very close together swapping their positions or becoming one
broader peak.
The tting parameter can be eectively constrained by dening the real physically
meaningful parameter P
j
[e.g., the GAIN in Eq. (84)] as an arctangent function of the
tting parameter a
j
(McCarthy and Schamber, 1981; Nullens et al., 1979):
P
j
= P
0
j
2L
j
p
arctan a
j
(104)
where P
0
j
is the expected value (initial guess) of the parameter P
j
, and L
j
denes the range.
As a result of this transformation, the parameter P
j
will always be in the range P
0
j
L
j
.
Apart from signicantly adding to the mathematical complexity, such transformation has
the disadvantage that the w
2
minimum, although lying within the selected interval, cannot
be reached because the path toward the minimum passes a forbidden region. Also, this
constraint makes no distinction between the more probable values of the parameter lying
near the center of the interval and the unlikely values at the limits.
Copyright 2002 Marcel Dekker, Inc.
An alternative approach, proposed by Nullens et al. (1979), relies on the modica-
tion of the curvature of the w
2
surface. The tting parameters, such as ZERO, GAIN,
NOISE, and FANO, are considered as random variables with an expected value equal to
the initial guess a
0
j
and having an uncertainty Da
j
. They can be included in the expression
of w
2
, just as the observed data points y
i
s
i
:
w
2
=
i
1
s
2
i
y
i
y(i) [ [
2
j
1
(Da
j
)
2
[a
0
j
a
j
[
2
(105)
Using this expression in the Marquardt nonlinear least-squares-tting algorithm results in
modied equations for the diagonal terms of the a matrix and for the b vector (see Sec. IX):
a
jj
=
i
1
s
2
i
@y
0
(i)
@a
j
_ _
2
1
(Da
j
)
2
(106)
and
b
j
=
i
1
s
2
i
[ y
i
y
0
(i)[
@y
0
(i)
@a
j
a
0
j
a
j
(Da
j
)
2
(107)
If no signicant peaks are present in the tting interval, the derivatives of the tting
function with respect to the position and width of the peaks are zero. Therefore, the
Figure 26 Fit of part of a PIXE spectrum having very high count rates for Fe and Zn. (Top:
without considering sum peaks; bottom: sum peaks included in the fitting model.)
Copyright 2002 Marcel Dekker, Inc.
second term in b
j
will dominate, causing the parameter estimate in the next iteration to be
such that a
j
tends toward a
0
j
(the initial value). If well-dened peaks are present, however,
the second term in b will be negligibly small compared to the rst term and the method
acts as if no constraints were present.
Another way to look at this is by considering the curvature of the w
2
surface near the
minimum. Figure 27 shows a cross section through the w
2
surface along the peak position
parameter. The variation of the w
2
values, with and without constraints, are shown for a
large (10,000 counts) and a small (100 counts) peak. The true peak position is at channel
100 and the FWHM of the peak is eight channels. The constraint, Da, on the peak position
is one channel. From Figure 27, it is evident that the w
2
minimum is much better dened
for a small peak when the constraint is applied, whereas this has no inuence for a large
peak. This method has been implemented in the AXIL program to constraint the energy
and resolution calibration parameters.
2. Weightingthe Fit
The weights w
i
used in the least-square methods [Eq. (77)] are dened in terms of the
true (population) standard deviation of the data y
i
, which can be obtained directly from
the data itself:
w
i
=
1
s
2
i
-
1
y
i
(108)
This approximation is valid for moderate to good statistics. When regions of the spectrum
with very bad statistics (low channel contents) are tted, estimating the weight from the
measured channel content causes a systematic bias leading to an underestimate of the peak
areas. To overcome this problem, Phillips suggested estimating the weights from the
Figure27 Effect of the use of a constraint on the shape of the w
2
response surface. (Left: marginal
effect for a large peak; right: important contribution for a small peak.)
Copyright 2002 Marcel Dekker, Inc.
average of three channels (Phillips, 1978). In the AXIL program, the weights are initially
based on the measured channel content, but when the overall w
2
value of the t falls
below 3, the weights are calculated from the tted channel content. This is based on the
idea that when the calculated spectrum approached the measured spectrum, the calculated
channel contents are a better estimate of the true channel content than the measured
values. The eect of the weighting on the t is considerable, as can be seen in Figure 28.
E. Examples
To illustrate the working of the nonlinear least-squares-tting method, an articial spec-
trum with CuK and ZnK lines is tted with four Gaussian peaks on a constant continuum.
Using the Marquardt algorithm, the position, width, and area of each peak are de-
termined. The tting function thus is
y(i) = a
0
i
4
j=1
a
j
exp
(i a
j1
)
2
2a
2
j2
_ _
(109)
with i the channel number (independent variable) and a
j
the parameters to be determined,
13 in total. In Table 10, the values of the parameters to generate the spectrum (true va-
lues), the initial guesses for the nonlinear parameters, and the nal tted values are given.
The initial guesses were deliberately taken rather far away from the true values. Figure 29
(top) shows the tted spectrum after the rst and second iterations. During the second
iteration, the Marquardt algorithm evolved into a gradient search, drastically changing the
position and width of the peaks. Even after ve iterations, the calculated spectrum deviates
considerably from the measured spectrum, as can be seen from Figure 29 (bottom).
Figure 28 The effect of weighing the least-squares fit is shown on part of an PIXE spectrum with
a very small number of counts per channel.
Copyright 2002 Marcel Dekker, Inc.
Finally, after 10 iterations, a perfect match between them is obtained, with a reduced w
2
value of 0.96.
From Table 10, it is evident that the t was quite successful, with all peak areas,
positions, and widths estimated correctly within the calculated standard deviation. One
observes that the uncertainties in the peak areas are approximately equal to the square
root of the peak area and that the position and the width of the peaks are estimated very
precisely (within 0.01 channel or 0.2 eV), especially for the larger peaks.
By observing how the iteration procedure changes the peak width and position
parameters, one can imagine that something might go wrong. Especially if the spectrum is
more complex, chances are high that the iteration stops in a false minimum or even drifts
away completely. In both cases, physically incorrect parameter estimates will be obtained.
In practice, it is course possible to give much better initial estimates for the peak position
and width parameters than used in this example.
In a next example, a complex spectrum (geological reference material JG1, excited
with MoK x-rays from a secondary target system) is evaluated using nonlinear least-
squares tting. In Figure 30, the spectrum and the t are shown together with the residuals
of the t (see p. 298). Due to the large number of overlapping lines, the method used in the
rst example (tting the position and width of each peak independently) is not applicable
in this case. For the description of the spectrum from 1 to 18 keV, the uorescence lines of
21 elements were used. Au, Hg, Pb, and Th were treated each as one L group
(L
1
L
2
L
3
) and Al, Si, Ti, V, Cr, Mn, Ni, Cu, Zn, Ga, Al, Br, Sr, Rb, Y as one K group
(Ka Kb); K, Ca, and Fe were tted with individual Ka and Kb peaks. The coherently
scattered MoKa radiation was tted with a Gaussian and the incoherent scattered MoKa
radiation is tted with a Voigt prole. Including escape and sum peaks, this amounts to
well over 100 peak proles. Step and tail functions [Eq. (89)] were included for all peaks,
using expressions to relate the step and tail heights and the tail width to the energy of the
peak. The continuum is described by an exponential function [Eq. (81)] with six para-
meters. The least-squares t thus performed required the renement of 37 parameters
Table 10 True Value of Spectrum Parameters Used to Generate the Articial Spectrum in
Figure 29, Initial Guesses and Fitted Values of the Peak Area, Position, and Width Obtained by
Nonlinear Least-Squares Fitting
Parameter True value Initial guess Fitted value
Area (counts)
CuKa 100,000 0 100,134321
CuKb 13,400 0 13,163169
ZnKa 30,000 0 30,092213
ZnKb 4,106 0 413883
Position (channel number)
CuKa 402.05 395 402.030.01
CuKb 445.25 450 445.350.06
ZnKa 431.55 435 431.590.04
ZnKb 478.60 485 478.680.09
Width (channels)
CuKa 3.913 3 3.910.01
CuKb 4.033 3 3.990.05
ZnKa 3.995 3 4.020.03
ZnKb 4.123 3 4.060.08
Copyright 2002 Marcel Dekker, Inc.
(4 calibration parameters, 27 peak areas, 6 continuum parameters, 3 step and tail para-
meters, and 1 Voigt parameter). The minimum w
2
value of 1.47 is obtained after 16
iterations. The residuals indicate an overall good t with most of the residuals in the 73
to 3 interval, without any systematic patterns. Its interesting to note that the tted
continuum is well below the base of the peaks, especially in the region from channel 200 to
600. The continuum describes correctly the small scattered bremsstrahlung contribution
from the x-ray tube above 12 keV (channel 600) and the Compton scattering in the de-
tector at the low-energy side. Most of the apparent continuum in this secondary-target
EDXRF spectrum is due to incomplete charge collection and tailing phenomena of the
scattered Mo excitation radiation and the uorescence lines.
F. Evaluation of Fitting Results
In order to understand and appreciate the capabilities and limitations of least-squares
tting, either using library spectra, linear, or nonlinear analytical functions, it is useful to
study the eect of random and systematic errors in some detail. Random errors are as-
sociated with the uncertainty s
i
of the channel content y
i
. As will be seen, these un-
certainties inuence the precision of the net peak areas and determine the ultimate
resolving power of the least-squares method. Systematic errors, on the other hand,
Figure 29 Artificial CuK and ZnK line spectrum tted with a nonlinear least-squares procedures
to optimize peak area, position, and width. (Top: Fitted spectrum after rst and second iteration;
bottom: after fth and nal iterations.)
Copyright 2002 Marcel Dekker, Inc.
are caused by discrepancies between the tting model and the observed data and cause
inaccuracies in the net peak areas.
1. Error Estimate
Section IX explains how the least-squares-tting method (linear as well as nonlinear)
allows the estimation of the uncertainties in the tted parameters. These uncertainties
result from the propagation of the statistical uctuations in the spectral data into the
parameters. Intuitively, one could come to the conclusion that the standard deviation of
the peak area should be close to the square root of the peak area. This is indeed the case
for a large, interference-free peak on a low continuum, but if the peak is sitting on a high
continuum and=or is seriously overlapped by another peak, the uncertainty in the esti-
mated peak area will be much larger.
A properly implemented least-squares method not only correctly estimates the net
peak areas but also their uncertainty, taking into account the continuum and the degree of
peak overlap, providing, of course, that the tting model is capable to describe the
measured spectrum. The closer together the peaks are, the higher the uncertainties in the
two peak areas will become. (Theoretically, in the limit of complete overlap, the un-
certainty will be innite and the area of the two peaks will have complete erratic values,
but their sum will still represent correctly the total net peak area of the two peaks; in
practice, the curvature matrix a will be singular so that the matrix inversion fails.)
Figure 30 Example of the fit of a complex x-ray spectrum (thin film deposit of geological
reference material JG1 excited with a Mo secondary-target system). The residuals are plotted as an
indicator of the quality of the fit.
Copyright 2002 Marcel Dekker, Inc.
The uncertainties in the net peak areas can be used to decide if a peak is indeed
present in the spectrum and to calculate the detection limit. A peak area is statistically not
signicant dierent from zero if its value is in the range 3s. Any value above 3s gives
clear evidence that the peak is present; any value below 73s would indicate that there is
something wrong with the model because truly negative peak areas are physically mean-
ingless. Because the uncertainty in the net peak area includes the inuence of continuum
and peak overlap, it can be used to calculate the a posteriori detection limits of the ele-
ments (peaks) present in the spectrum. Three situations can occur:
1. Estimated peak area >36 standard deviation
? report: areastandard deviation
2. 736 standard deviation _estimated area _36 standard deviation
? report: detection limit equal to 36 standard deviation
3. Area <736 standard deviation
? revise the tting model
2. Criteria for Qualityof Fit
From the denition of the least-squares method, it follows that the w
2
value [Eq. (75)]
estimates how well the model describes the data. The reduced w
2
value, obtained by di-
viding w
2
by the number of degrees of freedom;
w
2
n
=
1
n
w
2
=
1
n m
w
2
(110)
has an expected value of 1 for a perfect t. The number of degrees of freedom equals the
number of data points (n) minus the number of parameters (m) estimated during the t.
Because w
2
is also a random variable, the observed value will often be slightly larger or
smaller than 1. Actually, w
2
follows (approximately) a chi-square distribution and the 90%
condence interval is given by
w
2
n;0:95
_ w
2
n
_ w
2
n;0:05
(111)
where w
2
n;0:95
and w
2
n;0:05
are the tabulated w
2
values at the 95% and 5% condence intervals,
respectively. For n =20, the condence interval is 0.5431.571, and for n =200, it is
0.8411.170. An observed value of w
2
n
in this interval indicates that the model describes the
experimental data within the statistical uncertainty of the data. In other words, all re-
maining dierences between the data and the t can be attributed to the noise uctuations
in the channel contents and are statistically not signicant. When tting complex spectra
with good statistics, much higher chi-square values will be obtained due to small im-
perfections in the continuum or peak model. This does not mean that the estimated peak
areas are not useful any more, but high reduced w
2
values ( >3) might indicate that the
tting model needs improvement.
The reduced chi-square value as dened in Eq. (110) gives an estimate of the overall
t quality over the entire tting region. Locally, in some part of the spectrum, the t might
actually be worse than indicated by this value. It is therefore useful to dene the chi-square
value for each peak separately:
w
2
P
=
1
n
1
n
2
m
i=n
2
1
s
2
i
[y
i
y(i)[
2
(112)
Copyright 2002 Marcel Dekker, Inc.
where n
1
and n
2
are the boundaries of the peak at FWTM and n
2
n
1
approximates the
number of degrees of freedom. High values of w
2
P
indicate that the peak is tted poorly and
the resulting peak area should be used with caution. In this case (w
2
P
> 1), it is advisable to
give a conservative estimate for the uncertainty in the net peak area, by multiplying the
calculated uncertainty with the square root of the w
2
value:
s
A
/ = s
A
w
2
P
_
(113)
Although the w
2
values gives an indication of the goodness of t, visual inspection of
the t is highly recommended. Because of the large dynamic range of the data, a plot of the
spectrum and the t on a linear scale nearly always give the impression of a perfect t. A
plot of the logarithm or the square root of the data is more appropriate. The best method
is to plot the residuals of the t as in done in Figure 30. The residuals are dened as
r
i
=
y
i
y(i)
s
i
(114)
It is the sum of the squares of these residuals that were minimized by the least-squares
method. Values in excess of 3 or 73 and especially the presence of a pattern in the
residuals then indicate poorly tted regions.
G. Available Computer Codes
In the literature, a number of computer programs for spectrum evaluation based on the
least-squares method are reported. Without attempting to be complete, the main char-
acteristics of a number of programs are summarized in the following paragraphs. Wa tjen
made a compilation of the characteristics of eight computer packages for PIXE analysis
(Wa tjen, 1987).
An intercomparison of ve computer program for the analysis of PIXE spectra re-
vealed a very good internal consistency among the ve programs (Campbell et al., 1987).
PIXE spectra of biological, environmental, and geological samples were used and their
complexity put high demands of the spectrum analysis procedures. The following pro-
grams were tested: AXIL, University of Gent, Belgium; HEX, University of Lund, Swe-
den; SESAM-X (Marburg, FRG), the Guelph program; PIXAN, Lucas Heights. It was
concluded that the most serious disagreement occurred for small peaks on the low-energy
tails of very large peaks, pointing to a need to a more accurate description of the tail
functions. Also, very good agreement between the linear (SESAM-X) and the nonlinear
least-squares approach was observed.
The Los Alamos PIXE Data Reduction software (Duy et al., 1987) contains three
components. The K and L relative x-ray intensities of the elements making up the sample
are computed taking into account the detector and sample absorption. Using Gaussian
peak shape, the energy and resolution calibrations of the spectrum are calculated. With the
relative peak areas and the calibration functions obtained is this way, the spectrum is tted
using a Gaussian peak shape and a polynomial continuum. Escape and pileup peaks can
be included. A linear least-squares t is done with the relative elemental concentrations
and the polynomial continuum coecient as unknowns. The continuum and the relative
concentrations are constrained to non-negative values and all elements having x-ray lines
in the spectrum interval considered are included.
The PIXASE computer package (Zolnai and Szabo , 1988) performs spectrum ana-
lysis using nonlinear least-squares tting. Elements are represented by groups of lines with
Copyright 2002 Marcel Dekker, Inc.
xed, absorption-corrected, relative intensities and including escape peaks. Each peak is
modeled by a Gaussian with the addition of a exponential tail and an error function as
step. The square of the peak width is a rst-order polynomial of the peak energy and the
position is a second-order polynomial of the peak energy. The continuum is described as
the sum of an exponential polynomial and two simple exponentials. Pileup eects are
treated as one pileup element. The nonlinear least-squares tting is done using a simple
grid search technique. The search space of each parameter is limited by user-supplied
minimum and maximum values. Fitting a large series of similar spectra linear least-squares
tting using a library of calculated spectrum components can be done.
Bombelka et al. (1987) described an PIXE analysis program based on linear least
squares. The peak shape includes a Gaussian low-energy tail function to account for the
incomplete charge collection and the escape peak. The position and the square of the
width of the peaks are given by rst-order linear function of the energy. The uorescence
lines of an element are modeled as a sum of those peak shapes with relative intensity ratios
corrected for absorption in the detector windows and absorbers. The continuum is com-
posed of a fourth-order exponential polynomial multiplied by the x-ray attenuation term
(bremsstrahlung continuum) and a second-order linear polynomial. Pileup is taken into
account as a pileup element. The energy and resolution calibration parameters are ob-
tained from selected peaks in the spectrum. The parameters of the exponential continuum
are calculated from continuum spectra. The parameters obtained by the linear least-
squares t are the amplitude parameters of each element and of one pileup element, the
amplitude parameter of the bremsstrahlung continuum, and the linear polynomial con-
tinuum parameters. The computer implementation for PIXE is called SESAMX and is
highly interactive with graphical representation of the spectral data. Other computer
program for tube-excited (Breschinsky et al., 1979) and synchrotron radiation- (Petersen
et al., 1986) excited XRF were developed based on this code.
SAMPO-X (Aarnio and Lauranto, 1989) is intended for the analysis of electron
induced x-ray spectra and is based on the well-known SAMPO code original developed for
g-ray spectroscopy (Routti and Prussin, 1969). A Gaussian with two exponential tails, as
in the original SAMPO program, is used to represent the peaks. The height and the po-
sition parameters are obtained by the nonlinear least-squares t. The peak width and the
tail parameters are obtained from shape-calibration tables by interpolation. The con-
tinuum is modeled by the semi-empirical electron bremsstrahlung intensity function
proposed by Pella et al. (1985). The thickness of the detector beryllium window and the
atomic number of the sample, which occur in this formula, are adjusted by the least-
squares t. The program also includes an element identication based on the energy and
intensity of the tted peaks and a standard ZAF matrix correction.
Jensen and Pind (1985) described a program for the analysis of energy-dispersive x-
ray spectra. The program uses a sum of Gaussians, one for each uorescence line. The
continuum is subtracted rst using a linear, parabolic, or exponential function tted from
peak-free regions in the spectrum. The peak width is obtained from a calibration function
which expressed the log of the peak width as a linear function of the peak position. The
width calibration is done using nonoverlapping peaks in a calibration spectrum or in the
spectrum to analyze. Peak positions are determined using a peak search method or entered
by the operator with the aid of a graphical display of the spectrum. The peak heights are
then determined using linear least-squares tting.
The computer code developed at the technical University of Graz (Marageter et al.,
1984a) is primary intended for the evaluation of x-ray uorescence spectra. A Gaussian
response function with an low-energy tail is used to describe the peaks. The square of the
Copyright 2002 Marcel Dekker, Inc.
peak width is a linear function of the peak energy. A straight-line equation relates the peak
position to the energy of the peaks. A parabola is used to describe the continuum, whereas
absorption edges are modeled by a complementary error function. The tting parameters
are the peak heights, the three continuum parameters, the height of the absorption edges,
and the two energy calibration parameters. Nonlinear least-squares tting is done with the
Marquardt algorithm using a tangent transformation to constrain the tting parameters to
physical meaningful values (Nullens et al., 1979). Provision for escape peaks and Auger
peaks is also made (Marageter et al., 1984b).
The AXIL program (Van Espen et al., 1977a) was originally developed for the ana-
lysis of x-ray uorescence spectra and later modied to allow the evaluation of electron-
and particle-induced x-ray spectra (Van Espen et al., 1979b, 1986). It uses the Marquardt
algorithm for nonlinear least-squares tting with a modied (constrained) chi-square
function. Linear, exponential, and bremsstrahlung polynomials can be used to model the
continuum as well as continuum stripping. X-ray lines are described by Gaussian functions
with an optional numerical peak-shape correction. Escape and sum peaks can be included
in the model. The peak position and the square of the peak width is related to the x-ray
energy by linear functions. Provision is made to correct for the absorption in detector
windows, lters, and the sample. In a later version, an orthogonal polynomial continuum
model was added, as well as step and tail functions to describe the deviation from the
Gaussian peak shape (Vekemans et al., 1994). The code was implemented as a Windows
application. An example of the screen output of this program is given in Figure 31.
VIII. METHODS BASED ONTHE MONTE CARLOTECHNIQUE
Monte Carlo techniques for the simulation of x-ray spectra are becoming more and more
popular, particularly because of the fast computers available today. These simulated
spectra are useful for studying the behavior and performance of various spectrum pro-
cessing methods. The Monte Carlo technique also can be used in quantitative analysis
procedures, as will be discussed in Sec. VIII. B.
A. Simulation of X-ray Spectra
During the development and test phase of a spectrum processing method, it is often ad-
vantageous to use computer-simulated spectra. For these spectra, such features as posi-
tion, width, and area of peaks are exactly known in advance. They can be generated to any
desired complexity. In order to make any real use of them, the simulated spectra must
possess the same channel-to-channel variation according to a Poisson distribution as ex-
perimental spectra.
A simple and adequate procedure consist of rst calculating, over the channel range
of interest, the ideal spectrum y
0
using, for example, a polynomial for the continuum and a
series of Gaussians as given by Eq. (109). More complex functions, including a more
physically realistic model for the continuum and tailed peaks, can be used if desired. The
next step is to add or subtract some number of counts from each channel content so as to
obtain data with the desired counting statistical noise. In other words, the true content
y
0
i
needs to be converted into a random variable, y
i
, so that is obeys a Poisson distribution
[Eq. (1)] with N
0
= y
0
.
Poisson-distributed random variables can be generated by various computer algo-
rithms. An example (Press, 1988) is given in Sec. X. For y
0
> 30 the Poisson-distributed
Copyright 2002 Marcel Dekker, Inc.
Figure 31 Screen capture of the spectrum analysis program WinAxil, showing the fitted spectrum and the results obtained.
Copyright 2002 Marcel Dekker, Inc.
random variable can be approximated by the much easier to calculate normal-distributed
random variable. The probability to observe y counts in a channel, assuming a normal
distribution, is given by
P(y) =
1
s
2p
_ e
(ym)
2
2s
2
(115)
with m = s
2
= y
0
. For large y
0
the normal distribution is a very good approximation of the
Poisson distribution. Even for small channel contents this approximation is quite sa-
tisfactory. The probability of observing 6 counts, assuming the true value is 4, is 0.10
according to a Poisson distribution, whereas it is 0.12 according to a normal distribution.
The problem of adding counting statistics to the calculated spectrum is thus reduced
to calculating a normal-distributed random variable y with mean value m = y
0
and var-
iance s
2
= y
0
. Starting from a uniformly distributed random variable U in the interval
(0, 1), which can be generated by a pseudo-random number generator, normal distri-
buted random variables with zero mean and unit variance can be obtained with the Box-
Muller method (Press, 1988).
n
1
= 2U
i
1
n
2
= 2U
i1
1
r = n
2
1
n
2
2
(116)
if r <1
z
1
= n
1
2 ln r=r
_
z
2
= n
2
2 ln r=r
_
n
1
and n
2
and r are calculated from two uniformly distributed random numbers U
i
and
U
i1
. If r is less than 1, two normally distributed random numbers z
1
and z
2
can be cal-
culated. If r _ 1; n
1
; n
2
; and r are recalculated using a new set of uniform random numbers.
The normally distributed random number y, with mean m = y
0
and variance s
2
= y
0
, is
then obtained by simple scaling:
y = m zs = y
0
z
y
0
_
(117)
Applying this to all channels will produce the desired counting statistics. Since z is nor-
mally distributed with mean 0 and unit variance, z can be negative as well as positive. The
count rate in each channel will thus be increased or decreased randomly and proportional
to
_
y
0
. The nal step in the computer simulation is to convert the real numbers that were
used during the calculation to integers.
Another interesting procedure is to generate articial spectra from parent spectra.
The parent spectrum is a spectrum acquired for a very long time so that it exhibits ex-
tremely good counting statistics (high channel content). A large number of child spectra
with lower and varying counting statistics can be generated by the procedure explained
below. This method is useful to study the eect of counting statistics on spectrum-
processing algorithms (Ryan, 1988).
From the parent spectrum y
i
, which might be rst smoothed to reduce the noise even
further providing not too much distortion is introduced, the normalized cumulative dis-
tribution function Y
j
is calculated
Y
j
=
j
i=0
y
i
n
i=0
y
i
(118)
Copyright 2002 Marcel Dekker, Inc.
n being the number of channels in the spectrum and Y
j
in the interval (01). To generate
the child spectrum we select N times a channel i according to the equation
i = Y
1
(U) (119)
where Y
1
is the inverse cumulative distribution and U is a uniformly distributed random
number. Each time channel i is selected, one count is added to that channel. Since N is the
total number of counts in the child spectrum, the counting statistics can be controlled by
varying N. However, Y
1
cannot be expressed as an analytical function, but for each
random number U we select the channel i for which Y
i
_ U < Y
i1
. Figure 32 shows an
EDXRF spectrum from a 0.187 mg=cm
2
pellet of IAEA Animal Blood reference material
acquired for 1000 s with a Tracor Spectrace 5000 instrument, which is used as a parent
spectrum. The cumulative distribution function and a child spectrum simulated with
N=3610
4
are also shown. The total number of counts in the original spectrum is 3610
6
.
The child spectrum is equivalent to a spectrum that would have been acquired for 10 s.
Some computer routines useful for simulation experiments are given in Sec. X.
Figure 32 Simulation of a child spectrum from a parent. (Top: original spectrum; middle:
the cumulative distribution; bottom: generated child.)
Copyright 2002 Marcel Dekker, Inc.
B. SpectrumEvaluation Using Monte CarloTechniques
Another interesting application of the Monte Carlo technique is the simulation of the
complete response of an x-ray uorescence setup. Given an excitation spectrum and the
excitation-detection geometry the interactions of the primary photons with the sample are
simulated and the events giving rise to detectable phenomena are registered. With such a
Monte Carlo simulation, the intensity of the characteristic lines and the scattered excitation
spectrum are estimated, taking primary and secondary eects (absorption, enhancement,
and single and multiple scattering) into account. The obtained spectrumis one as seen by an
ideal detector with innite resolution. This spectrum can then be convolution with the
response function of a real detector to mimic a typical observed pulse-height spectrum.
Apart from a detailed Monte Carlo simulation code, this technique requires detailed
knowledge of the excitation spectrum and a very accurate detector response function.
Doster and Gardner were among the rst to apply this technique to simulate the
complete spectral response of an EDXRF system (Doster and Gardner, 1982a; Doster and
Gardner, 1982b; Yacout and Dunn, 1987). More recently, Janssens and coworkers de-
veloped highly ecient computer codes for the simulation of conventional and synchro-
tron energy-dispersive x-ray uorescence systems, allowing such conditions as polarized
radiation and heterogeneous samples to be taken into consideration (Janssens et al., 1993;
Vincze et al., 1993; Vincze et al., 1995a, 1995b; Vincze et al., 1999). Figure 33 shows the
results of the application of this Monte Carlo simulation for a NIST SRM 1155 steel
sample excited with a ltered Rh-anode x-ray tube. At the top the simulated spectrum as
seen by a perfect detector is shown, at the bottom the spectrum after convolution with a
suitable detector response function is compared with a real measured spectrum of this
sample. Except for some sum peaks that occur in the measured spectrum, an excellent
agreement between the simulated and measured uorescence lines, scattered peaks, and
continuum is obtained.
These simulated x-ray spectra can be used in various ways in quantitative analysis.
Using the complete spectral response of spectrometer, one can try to nd a sample
composition that minimizes (in the least-squares sense) the dierence between the simu-
lated and the measured spectrum. The analysis involves the following steps: (1) simulation
of the X-ray intensities over the expected composition range of the unknowns; (2) con-
volution with the detector response function to obtain spectra; (3) construction of a w
2
map (weighted sum of squares of dierences between simulated spectra and experimental
spectrum) as a function of the sample composition; (4) interpolation of w
2
for the com-
position corresponding to the minimum. An interesting aspect of this method is that all the
information present in the spectrum (characteristic lines, scattered radiation, continuum)
is considered. Doster and Garnder demonstrated an analytical accuracy of the order of
2% absolute for the analysis of Cr-Fe-Ni alloys with a
109
Cd radioisotope system (Doster
and Gardner, 1982a). Based on this work, Yacout and Dunn (Yacout and Dunn, 1987)
demonstrated the use of the inverse Monte Carlo method, which requires in principle only
one simulation to analyze a set of similar samples.
Another method is called Monte Carlolibrary least-squares analysis (Verghese
et al., 1988). Starting from an initial guess of the composition of an unknown sample, a
spectrum is simulated taking into account all the interactions in the sample. During the
simulation one keeps track of the response of each element to construct library spectra.
After the simulation these library spectra are used to obtain the elemental concentrations
by linear least-squares tting (see Sec. VI). In the concentration in the unknown samples
dier too much from the initial assumed concentration, the simulation is repeated.
Copyright 2002 Marcel Dekker, Inc.
In contrast to the normal library least-squares method, this method has the advantage that
the library spectra are simulated for a composition close to the composition of the spec-
trum to analyzed, rather than measured from standards. This eliminates the necessity of
applying the top-hat lter and problems related to changes in Kb=Ka ratios, and again the
method combines spectrum evaluation with quantitative analysis.
Finally the ability to simulate x-ray spectra that agree very well with real measured
spectra opens the possibility to used them as standards for the quantitative analysis
based on partial least-squares regression (see Sec. VI.B). Indeed, as this method only
functions correctly if the PLS model is built using a large number of standards covering
the entire concentration domain, it seems advantageous to use simulated spectra for this.
All the inter-element interactions can be accounted for by the simulation and only a few
real standards are required to scale the simulated spectra.
Figure 33 Monte Carlo simulation of the spectral data from an NIST SRM 1155 sample excited
with a Rh-anode x-ray tube. (Top: simulated spectrum as seen by an ideal detector; bottom:
simulated spectrum after convolution with the detector response function and comparison with
measured spectrum.)
Copyright 2002 Marcel Dekker, Inc.
IX. THE LEAST-SQUARES-FITTINGMETHOD
The aim of the least-squares method is to obtain optimal values for the parameters of a
function that models the dependence of experimental data. The method has its roots in
statistics but is also considered part of numerical analysis. The least-squares parameter
estimate, also known as curve tting, plays an important role in experimental science. In
x-ray uorescence, it is used in many calibration procedures and it forms the basis of a
series of spectrum analysis techniques. In this section, an overview of the least-squares
method with emphasis on spectrum analysis is given.
Based on the type of tting function, one makes a distinction between linear and
nonlinear least-squares tting because one requires numerical techniques of dierent
complexity to solve the problem in the two cases. The linear least-squares method deals
with the tting of functions that are linear in the parameters to be estimated. For this
problem, a direct algebra solution exists. If the tting function is not linear in one or more
of the parameters, one uses nonlinear least-squares tting and the solution can only be
found iteratively. A group of linear functions of general interest are the polynomials, the
straight line being the simplest case. The special case of orthogonal polynomials will also
be considered. If more than one independent variable (x
1i
; x
2i
; . . . ; x
mi
) is associated with
each measurement of the dependent variable y
i
, one speaks of multivariate regression.
Spectrum analysis using the library function (e.g., lter-t method) belongs to this cate-
gory. If analytical function (e.g., Gaussians) are tted to a spectrum, the method of linear
or nonlinear least square is used, depending on whether nonlinear parameters, such as the
peak position and width, are determined.
A. Linear-Least Squares
Considered the problem of tting experimental data with the following linear function:
y = a
1
X
1
a
2
X
2
a
m
X
m
(120)
This function covers all linear least-squares problems. If m = 2; X
1
= 1, and X
2
= x, the
straight-line equation y = a
1
a
2
x is obtained. For m > 2 and X
k
= x
k1
, Eq. (120) is a
polynomial y = a
1
a
2
x a
3
x
2
a
m
x
m1
to be tted to the experimental data
points x
i
; y
i
; s
i
i = 1; . . . ; n. If X
1
; . . . ; X
m
represents dierent independent variables, the
case of multiple linear regression is dealt with. Because of this generality, we will discuss
the linear least-squares method based on Eq. (120) in detail.
Assume a set of n experimental data points:
x
1i
; x
2i
; . . . ; x
mi
; y
i
; s
i
; i = 1; . . . ; n (121)
with x
ki
the value of the kth independent variables X
k
in measurement i, assumed to be
known without error, and y
i
the value of the dependent variable measured with un-
certainly s
i
. The optimum set of parameters a
1
; . . . ; a
m
that gives a least-squares t of
Eq. (120) to these experimental data are those values of a
1
; . . . ; a
m
that minimize the chi-
square function:
w
2
=
n
i=1
1
s
2
i
(y
i
a
1
X
1
a
2
X
2
a
m
X
m
)
2
(122)
The minimum is found by setting the partial derivatives of w
2
with respect to the para-
meters to zero:
Copyright 2002 Marcel Dekker, Inc.
@w
2
@a
k
= 2
n
i=1
1
s
2
i
(y
i
a
1
X
1
a
2
X
2
a
m
X
m
)X
k
= 0; k = 1; . . . ; m (123)
Dropping the weights 1=s
2
i
temporarily for clarity, we obtain a set of m simultaneous
equations in the m unknown a
k
:
y
i
X
1
= a
1
X
1
X
1
a
2
X
2
X
1
a
m
X
m
X
1
y
i
X
2
= a
1
X
1
X
2
a
2
X
2
X
2
a
m
X
m
X
2
.
.
.
y
i
X
m
= a
1
X
1
X
m
a
2
X
2
X
m
a
m
X
m
X
m
(124)
where the summations run over all experimental data points i. These equations are known
as the normal equations. The solutionthe values of a
k
can easily be found using matrix
algebra. Because two (column) matrices are equal if their corresponding elements are
equal, the set of equations can be in matrix form as
y
i
X
1
y
i
X
2
.
.
.
y
i
X
m
_
_
_
_
=
a
1
X
1
X
1
a
2
X
2
X
1
a
m
X
m
X
1
a
1
X
1
X
2
a
2
X
2
X
2
a
m
X
m
X
2
.
.
.
a
1
X
1
X
m
a
2
X
2
X
m
a
m
X
m
X
m
_
_
_
_
(125)
The right-hand column matrix can be written as the product of a square matrix a and a
column matrix a:
y
i
X
1
y
i
X
2
.
.
.
y
i
X
m
_
_
_
_
=
X
1
X
1
X
2
X
1
X
m
X
1
X
1
X
2
X
2
X
2
X
m
X
2
.
.
.
X
1
X
m
X
2
X
m
X
m
X
m
_
_
_
_
a
1
a
2
.
.
.
a
2
_
_
_
_
(126)
or
b = aa (127)
This equation can be solved for a by premultiplying both sides of the equation with
the inverse matrix a
1
,
a
1
b = a
1
aa = Ia (128)
or, I being the identity matrix,
a = a
1
b (129)
Introducing the weights again, the elements of the matrices are given by
b
j
=
n
i=1
1
s
2
i
y
i
X
j
; j = 1; . . . ; m (130)
a
jk
=
n
i=1
1
s
2
i
X
k
X
j
; j = 1; . . . ; m; k = 1; . . . ; m (131)
and
Copyright 2002 Marcel Dekker, Inc.
a
j
=
m
k=1
a
1
jk
b
k
; j = 1; . . . ; m (132)
where a
1
jk
are the elements of the inverse of the matrix a
1
.
The uncertainty in the estimate of a
j
is due to the uncertainty of each measurement
multiplied by the eect that measurement has on a
j
:
s
2
a
j
=
n
i=1
s
2
i
@a
j
@y
i
_ _
2
(133)
Because a
1
jk
is independent of y
i
, the partial derive is simply
@a
j
@y
i
=
1
s
2
i
m
k=1
a
1
jk
X
k
(i) (134)
After some rearrangements, it can be shown that
s
2
aj
=
m
k=1
m
l=1
a
1
jk
a
1
jl
n
i=1
1
s
2
i
X
k
X
l
_ _
(135)
the term in the brackets being a
kl
and
s
2
aj
=
l
a
1
jk
a
1
jl
a
kl
= a
1
jj
(136)
This results in the simple statement that the variance (square of uncertainty) of a tted
parameter a
j
is given by the diagonal element j of the inverse matrix a
1
. The o-diagonal
elements are the covariances. For this reason, a
1
is often called the error matrix. Simi-
larly, a is called the curvature matrix because the elements are a measure of the curvature
of the w
2
hypersurface in the m-dimensional parameter space. It can easily be shown that
1
2
@
2
w
2
@a
j
@a
k
=
i
1
s
2
i
X
k
X
j
= a
jk
(137)
If the uncertainties in the data points s
i
are unknown and are the same for all data
points (s
i
= s), these equations can still be used by setting the weights w
i
= 1=s
2
i
to 1.
Assuming the tting model is correct, s can be estimated from the data:
s
2
i
= s
2
- s
2
=
1
n m
i
(y
i
a
1
X
1
a
2
X
2
a
m
X
m
)
2
(138)
The uncertainties in the parameters are then given by
s
2
a
j
= s
2
a
1
jj
(139)
If the uncertainties in the data points are known, the reduced w
2
value can be calculated as
a measure of the goodness of t:
w
2
n
=
1
n m
i
1
s
2
i
(y
i
a
1
X
1
a
2
X
2
a
m
X
m
)
2
=
1
n m
w
2
(140)
The expected value of w
2
n
is 1.0, but due to the random nature of the experimental data values
slightly smaller or greater than 1 will be observed even for a perfect t. w
2
n
follows a chi-
square distribution with n mdegrees of freedom, and a 90%condence interval is dened by
Copyright 2002 Marcel Dekker, Inc.
w
2
(n; P = 0:95) _ w
2
n
_ w
2
(n; P = 0:05) (141)
where w
2
(n; P) is the (tabulated) critical value of the w
2
distribution for n degrees of freedom
at a condence level P. Observed w
2
n
values outside this interval indicate a deviation
between the t and the data that cannot be attribute to random statistical uctuations.
B. Least-Squares Fitting Using Orthogonal Polynomials
A special group of linear functions are orthogonal polynomials. Orthogonality means that
the polynomial terms are uncorrelated, and this has some distinct advantages. Let P
j
(x
i
)
be an orthogonal polynomial of degrees j; a function can then be constructed as a sum of
these orthogonal polynomials of successive higher degree:
y(i) =
m
j=0
C
j
P
j
(x
i
) (142)
The least-squares estimates of the coecient C
j
are determined by minimizing the weighted
sum of squares:
w
2
=
n
i=1
w
i
(y
i
y(i))
2
(143)
which results in a set of m1 normal equations in the m1 unknown. Because the P
j
(x
i
)
are a set of orthogonal polynomials, they have the property that
n
i=1
w
i
P
j
(x
i
)P
k
(x
i
) = g
k
d
jk
(144)
with g
k
a normalization constant and d
jk
= 0 for j ,= k.
Because of this property, the matrix of the normal equations is diagonal and the
polynomial coecients are directly obtained from
C
j
=
n
i=1
w
i
y
i
P
j
(x
i
)
g
j
(145)
and the variance of the coecient is given by
s
2
cj
=
1
g
j
(146)
Another advantage of the use of orthogonal polynomials is that the addition of one
extra term C
m1
P
m1
does not change the values of the already determined coecients
C
0
; . . . ; C
m
. Further, if the y
i
are independent, then also the C
j
are independent; that is, the
variancecovariance matrix is also diagonal. As a result, much higher-degree orthogonal
polynomials can be tted compared to ordinary polynomials without running into pro-
blems with ill-conditioned normal equations and oscillating terms.
Orthogonal polynomials can constructed by following recurrence relation:
P
j1
(x
i
) = (x
i
a
j
)P
j
(x
i
) b
j
P
j1
(x
i
) j = 0; . . . ; m1 (147)
a
j
and b
j
are constants independent of y
i
given by
a
j
=
n
i=1
w
i
x
i
[P
j
(x
i
)[
2
g
j
; j = 0; . . . ; m (148)
Copyright 2002 Marcel Dekker, Inc.
b
j
=
n
i=1
w
i
x
i
P
j
(x
i
)P
j1
(x
i
)
g
j1
; j = 0; . . . ; m (149)
Further, the normalization factor is given by
g
j
=
n
i=1
w
i
[P
j
(x
i
)[
2
(150)
and
b
0
= 0 and P
0
(x
i
) = 1 (151)
Thus, an example of a rst-order orthogonal polynomial is
C
0
P
0
C
1
P
1
= C
0
C
1
(x
i
a
0
) (152)
with
a
0
=
n
i=1
w
i
x
i
_
w
i
(153)
C. Nonlinear Least Squares
In this part, we consider the tting of a function that is nonlinear in one or more tting
parameters. Examples of such functions are a decay curve,
y(x) = a
1
exp(a
2
x) (154)
or a Gaussian on a linear background,
y(x) = a
1
a
2
x a
3
exp
(x a
4
)
2
2a
2
5
_ _
(155)
Equation (154) is nonlinear in a
2
; Eq. (155) is nonlinear in the parameters a
4
and a
5
.
Equation (154) is representative for a group of functions for which linear least-squares tting
can be applied after suitable transformation. Fitting with Eq. (155) implies the application of
truly nonlinear least-squares tting with iterative optimization of the tting parameters.
1. Transformationto Linear Functions
Taking the logarithm of Eq. (154),
ln y = ln a
1
a
2
x (156)
and dening y
/
= ln y; and a
/
1
= ln a
1
, a linear (straight line) tting function is obtained,
y
/
= a
/
1
a
2
x (157)
and the method discussed earlier can be applied, but not without making the following
important remark. We have transformed our original data y
i
to y
/
i
= ln y
i
. Consequently,
also the variance s
/2
i
has been changed according to the general error propagation formula:
s
/ 2
i
= s
2
i
@y
/
@y
_ _
2
(158)
Copyright 2002 Marcel Dekker, Inc.
or in this particular case,
s
/ 2
i
=
s
2
i
y
i
(159)
Thus, even if all original data points had the same uncertainty (s
i
= s) and unweighed
linear least-squares tting could have been used, after the transformation a weighted linear
least-squares t is required. The results of the t are the parameters and the associated
uncertainties of the transformed equation, and to obtain the original parameter, we must
perform a back transformation with the appropriate error propagation,
a
1
= e
a
/
1
(160)
s
2
a
1
= s
2
a
/
1
@a
1
@a
/
1
_ _
2
= a
1
s
2
a
/
1
(161)
2. General Nonlinear Least Squares
For the general case of least-squares tting with a function that is nonlinear in one or more
of its tting parameters, no direct solution exists. Still, we can dene the object function w
2
:
w
2
=
i
1
s
2
i
[y
i
y(x
i
; a)[
2
(162)
whose minimumwill be reached when the partial derivative with respect to the parameters are
zero; however, this will result in a set of m equations for which no general solution exists. The
other approachtothe problemis thentoconsider w
2
as acontinuous functionof the parameters
a
j
(i.e., w
2
will take a certain value for each set of values of the parameters a
j
for a given dataset
x
i
; y
i
; s
2
i
). w
2
thus forms a hypersurface in the m-dimensional space, formed by the tting
parameter a
j
. This surface must be searched to locate the minimum of w
2
. Once found, the
corresponding coordinate values of the axes are the optimumvalues of the tting parameters.
The problem of nonlinear least-squares tting is thus reduced to the problem of
nding the minimum of a function in an m-dimensional space. Any algorithm that per-
forms this task should operate according to the following:
1. Given some initial set of values for the parameters a
ini
evaluate w
2
:
w
2
old
= w
2
(a
ini
)
2. Find a new set of values a
new
such that w
2
new
< w
2
old
.
3. Test the minimum of w
2
value:
if w
2
new
is the (true) minimum accept a
new
as the optimum values of the t
else w
2
old
= w
2
new
and repeat Step 2.
From the scheme, the iterative nature of the nonlinear least-squares tting methods be-
comes evident. Moreover, it shows some other important aspects of the method: Initial
values are required to start the search, we need a procedure to obtain a new set of
parameters which preferably are such that w
2
is decreasing, and we need to be sure that the
true minimum, not some local minimum, is nally reached.
A variety of algorithms has been proposed, ranging from brute-force mapping
procedures dividing the m-dimensional parameter space in small cells and evaluating w
2
in
each point, to more subtle simplex search procedures (Fiori et al., 1981). The most
Copyright 2002 Marcel Dekker, Inc.
important group of algorithms is nevertheless based on the evaluation of the curvature
matrix. The gradient method and the rst-order expansion will be discussed briey, as they
form the basis of the most widely used LeverbergMarquardt algorithm (Marquardt,
1963; Bevington and Robinson, 1992; Press et al., 1988).
a. The Gradient Method
Having a tting function y = y(x; a) and w
2
dened as a function of the m parameters a
j
,
w
2
= w
2
(a) =
n
i=1
1
s
2
i
[y
i
y(x
i
; a)[
2
(163)
the gradient of w
2
in the m-dimensional parameter space is given by
Hw
2
=
j
@w
2
@a
j
j (164)
where j is the unit vector along the axis j and the components of the gradient are given by
@w
2
@a
j
= 2
i
1
s
2
i
[y
i
y(x
i
; a)[
@y
@a
j
(165)
It is convenient to dene
b
j
=
1
2
@w
2
@a
j
(166)
The gradient gives the direction in which w
2
increases most rapidly. A method to locate the
minimum can thus be developed on this basis. Given the current set of parameters a
j
, a
new set of parameters a
/
j
is calculated (for all j simultaneously):
a
/
j
= a
j
Da
j
b
j
(167)
which follows the direction of steepest descent and guarantees a decrease of w
2
(at least if
the appropriate step sizes Da
j
are taken).
The gradient method works quite well away from the minimum, but near the
minimum, the gradient becomes very small (at the minimum, even zero). Fortunately, the
method discussed next behaves in the opposite way.
b. First-Order Expansion
If we write the tting function y(x
i
; a) as a rst-order Taylor expansion of the parameters
a
j
around y
0
,
y(x; a) = y
0
(x; a)
j
@y
0
(x; a)
@a
j
da
j
(168)
we obtain an (approximation) to the tting function which is linear in the parameter in-
crements da
j
. y
0
(x; a) is the value of the tting function for some initial set of parameter a.
Using this function, we can now express w
2
as
w
2
=
i
1
s
2
i
y
i
y
0
(x
i
; a)
j
@y
0
(x
i
; a)
@a
j
da
j
_ _
2
(169)
and we can use the method of linear least squares to nd the parameters da
j
so that w
2
will
be minimal. We are thus tting the dierence y
/
i
= y
i
y
0
(x
i
; a) with the derivatives as
Copyright 2002 Marcel Dekker, Inc.
variables and the increments da
j
as unknowns. With reference to the section on linear
least-squares tting [Eq. (122)],
X
j
=
@y
0
(x
i
)
@a
j
(170)
and [Eq. (130) and (131)]
b
j
=
n
i=1
1
s
2
i
[y
i
y
0
(x
i
)[
@y
0
(x
i
)
@a
j
(171)
a
jk
=
n
i=1
1
s
2
i
@y
0
(x
i
)
@a
j
@y
0
(x
i
)
@a
k
(172)
dening a set of m normal equations in the unknowns da
j
,
b = ada (173)
with solution
da
j
=
m
k=1
a
1
jk
b
k
(174)
It is not very dicult to prove that
b
j
=
1
2
@w
2
0
@a
k
(175)
(i.e., the component of the gradient of w
2
at the point of expansion) and
a
jk
-
1
2
@
2
w
2
0
@a
j
da
k
(176)
Thus, a
jk
in Eq. (172) is the rst-order approximation to the curvature matrix whose
inverse is the error matrix.
The rst-order expansion of the tting function is closely related to the rst-order
Taylor expansion of the w
2
hypersurface itself:
w
2
= w
2
0
j
@w
2
0
@a
j
da
j
(177)
where w
2
0
is the w
2
function at the point of expansion:
w
2
0
=
n
i=1
1
s
2
i
[y
i
y
0
(x
i
; a)[
2
(178)
At the minimum, the partial derivation of w
2
with respect to the parameter a
k
will be zero:
@w
2
@a
k
=
@w
2
0
@a
k
j
@
2
w
2
0
@a
j
@a
k
da
k
= 0 (179)
This results in a set of equations in the parameters da
k
:
@w
2
0
@a
k
=
m
j=1
@w
2
0
@a
j
@a
k
da
k
(180)
Copyright 2002 Marcel Dekker, Inc.
b
k
=
a
jk
da
k
(181)
which is the same set of equations, except that in the expansion of the tting function, only
a rst-order approximation of the curvature matrix is used.
Because near the minimum the rst-order expansion of the w
2
surface is a good ap-
proximation, we canconclude that alsothe rst-order expansionof the tting function(whichis
computationally more elegant because only derivatives of functions and not of w
2
are required)
will yield parameter increments da
j
which will direct us toward the minimum. For each linear
parameter in the tting function, the rst-order expansion of the function in this parameter is
exact and the calculated increment da
j
will be such that the newvalue a
j
da
j
is optimum(for
the given set of nonlinear parameters which might not be at their optimum value yet).
c. The Marquardt Algorithm
Based on the observation that away from the minimum the gradient method is eective
and near the minimum the rst-order expansion is useful, Marquardt developed an al-
gorithm that combines both methods using a scaling factor l that moves the algorithm
either in the direction of the gradient search or into the direction of rst-order expansion
(Marquardt, 1963).
The diagonal terms of the curvature matrix are modied as follows:
a
/
jk
=
a
jk
(1 l); j = k
a
jk
; j ,= k
_
(182)
where a
jk
is given by Eq. (172) and the matrix equation to be solved for the increments da
j
is
b
j
=
k
a
/
jk
da
k
(183)
When l is very large (l 1), the diagonal elements of a dominate and Eq. (183) reduces to
b
j
- a
/
jj
da
k
(184)
or
da
k
-
1
a
/
jj
b
j
-
1
a
/
jj
@w
2
@a
k
(185)
which is the gradient, scaled by a factor a
/
jj
. On the other hand, for small values of
l (l 1), the solution is very close to rst-order expansion.
The algorithm proceeds as follows:
1. Given some initial values of the parameters a
j
, evaluate w
2
= w
2
(a) and initialize
l =0.0001.
2. Compute b and a matrices using Eqs. (171) and (172).
3. Modify the diagonal elements a
/
jj
= a
jj
l and compute da.
4. If w
2
(a da) _ w
2
(a)
increase l by a factor of 10 and repeat Step 3;
If w
2
(a da) < w
2
(a)
decrease l by a factor of 10
accept new parameters estimates a a da and repeat Step 2.
The algorithm thus performs two loops: the inner loop incrementing l until w
2
starts to
decrease and the outer loop calculating successively better approximations to the optimum
values of the parameters. The outer loop can be stopped when w
2
decreases by a negligible
absolute or relative amount.
Copyright 2002 Marcel Dekker, Inc.
Once the minimum is reached, the diagonal elements are an estimate of the un-
certainty in the tting parameters just as in the case of linear least squares:
s
2
aj
= a
1
jj
(186)
which is equal to (a
/
jj
)
1
providing the scaling factor l is much smaller than 1.
In Sec. X, a number of computer programs for linear and nonlinear least-squares
tting are given. Further information can be found in many textbook (Press et al., 1988).
The book by Bevington and Robinson (1992) contains a very clear and practical discus-
sion of the least-squares method.
X. COMPUTER IMPLEMENTATIONOF VARIOUS ALGORITHMS
In this section, a number of computer routines related to spectrum evaluation are listed.
The calculation routines are written in FORTRAN. Some example programs, calling this
FORTRAN routines, are written in C. The programs were tested using Microsoft FORTRAN
version 4.0A and C version 5.1. Most of the routines are written for clarity rather than
optimized for speed or minimum space requirement.
A. Smoothing
1. Savitskyand Golay Polynomial Smoothing
The subroutine SGSMTH calculates a smoothed spectrum using a second-degree poly-
nomial lter (see Sec. III.B.2).
SUBROUTINE SGSMTH (Y, S, NCHAN, ICH1, ICH2, IWID)
INTEGER*2 NCHAN, ICH1, ICH2, IWID
REAL*4 Y(0:NCHAN1), S(0:NCHAN1)
REAL C(20:20)
C -- Calculate filter coefficients
IW =MIN(IWID, 41)
M =IW=2
SUM =FLOAT( (2*M 71)*(2*M 1)*(2*M 3) )
DO 10 J =M, M
C(J) =FLOAT( 3*(3*M*M 3*M 71 75*J*J) )
10 CONTINUE
C -- Convolute spectrum with filter
JCH1 =MAX( ICH1, M )
JCH2 =MIN( ICH2, NCHAN 71 7M )
DO 30 I =JCH1, JCH2
S(I) =0.
Input: Y Original spectrum
NCHAN Number of channels in the spectrum
ICH1, ICH2 First and last channel number to be smoothed
IWID Width of the filter (2m1), IWID<42
Output S Smoothed spectrum, only defined between ICH1 and ICH2
Copyright 2002 Marcel Dekker, Inc.
DO 20 J = 7M, M
S(I) =S(I) C(J)*Y(I J)
20 CONTINUE
S(I) =S(I)=SUM
30 CONTINUE
RETURN
END
2. LowStatistics Digital Filter
The subroutine LOWSFIL smooths a spectrum using the low statistics digital lter algo-
rithm (see Sec. III.B.3).
SUBROUTINE LOWSFIL (Y, S, NCHAN, ICH1, ICH2, IFWHM)
INTEGER*2 NCHAN, ICH1, ICH2, IFWHM
REAL Y(0:NCHAN 71), S(0:NCHAN 71)
LOGICAL NOKSLOPE, STOOHIGH
REAL AFACT, FFACT, MFACT, RFACT
PARAMETER (AFACT =75., FFACT =1.5, MFACT =10., RFACT =1.3)
C -- Adjust smoothing region
IW =NINT( FFACT*IFWHM )
JCH1 =MAX(ICH1 7IW, IW)
JCH2 =MIN(ICH2 IW, NCHAN 71 7IW)
DO 100 I =JCH1, JCH2
IW =NINT( FFACT*IFWHM )
SUML =0.
SUMR =0.
DO 20 J =I 7IW, I 71
SUML =SUML Y(J)
20 CONTINUE
DO 30 J =I 1, I IW
SUMR =SUMR Y(J)
30 CONTINUE
C -- Adjust window
50 CONTINUE
SUMT =SUML Y(I) SUMR
IF( SUMT .GT. MFACT ) THEN
SLOPE =(SUMR 1.)=(SUML 1.)
NOKSLOPE =SLOPE.GT.RFACT .OR. SLOPE .LT. 1.=RFACT
STOOHIGH =SUMT .GT. AFACT*SQRT(Y(I))
IF( NOKSLOPE .OR. STOOHIGH .AND. IW .GT. 1 ) THEN
SUML =SUML 7Y(I 7IW)
SUMR =SUMR 7Y(I 7IW)
IW =IW 71
GOTO 50
Input: Y Original spectrum
NCHAN Number of channels
ICH1, ICH2 First and last channel to be smoothed
IFWHM FWHM in channels of a peak in the middle of the smoothing region
Output: S Smoothed spectrum, only defined between ICH1 and ICH2
Copyright 2002 Marcel Dekker, Inc.
ENDIF
ENDIF
C -- Smoothed value
S(I) =SUMT=FLOAT (2*IW 1)
100 CONTINUE
C -- Copy data points that could not be smoothed
DO 110 I =ICH1, JCH1 71
S(I) =Y(I)
110 CONTINUE
DO 120 I =JCH2 1, ICH2
S(I) =Y(I)
120 CONTINUE
RETURN
END
For each channel i in the spectrum, two windows, one on each side of the channel of width
f6FWHM(E
i
) channels are considered. In both windows, the channel contents are
summed, yielding a left sum L and a right sum R. Both windows are subsequently reduced
in width until either the total sum S = L y
i
R falls below some constant minimum M
or until two conditions are met:
1. S is less than a cuto value N = A
y
i
_
, with A a constant.
2. The slope (R 1)=(L 1) lies between 1=r and r, with r a constant.
The minimum constant M sets the base degree of smoothing in a region of vanishing
counts. The rst condition ensures that smoothing is conned to the low statistics region
of the spectrum; the second condition avoids the incorporation of the edges of the peaks in
the averaging.
When the above conditions are satised, the average S=(2f6FWHM1) is adopted
as a smoothed channel count. The following parameters were found to yield good results
when treating PIXE spectra: f =1.5, A=75, M=10, and r =1.3.
B. Peak Search
The subroutine LOCPEAKS locates peaks in a spectrum using positive part of tophyhat
lter (see Sec III.C).
SUBROUTINE LOCPEAKS (Y, NCHAN, IWID, R, IPOS, NPEAKS, MAXP)
INTEGER*2 NCHAN, IWID, NPEAKS, MAXP
INTEGER*2 IPOS (MAXP)
REAL Y (NCHAN), R
C - - Width of filter (number of channels in the top)
Input: Y Spectrum
NCHAN Number of channels in the spectrum
R Peak search sensitivity factor, typical 2 to 4
IWID Width of the filter, approx. equal to the FWHM of the peaks
MAXP Maximum number of peaks to locate (size of array IPOS)
Output: NPEAK Number of peaks found
IPOS Peak positions (channel number)
Copyright 2002 Marcel Dekker, Inc.
C must be odd, and at least 3
NP =MAX (IWID=2)*2 1, 3)
NPEAKS =0
C -- Calculate half width and start and stop channel
N =NP=2
I1 =NP
I2 =NCHAN-NP
C -- Initialize running sums
I =I1
TOTAL =0.
TOP =0.
DO 20 K = 7N*2, NP
TOTAL =TOTAL Y(I1 K)
20 CONTINUE
DO 22 K = 7N, N
TOP =TOP Y(I1 K)
22 CONTINUE
C - - Loop over all channels
LASTPOS =0
SENS =R*R
FI =0.
FNEXT =0.
SNEXT =0.
DO 100 I =I1 1, I2
TOP =TOP 7 Y(I-N-1) Y(I N)
TOTAL =TOTAL 7Y(I-NP) Y(I NP)
FPREV =FI
FI =FNEXT
SI =SNEXT
FNEXT =TOP TOP 7TOTAL
SNEXT =TOTAL
C Significant?
IF( FI.GT.0. .AND. (FI*FI.GT.SENS*SI) ) THEN
C Find maximum
IF( FI .GT. FPREV .AND. FI .GT. FNEXT ) THEN
IF( FPREV.GT.0. .AND. FNEXT.GT.0. ) THEN
C and store (ch# is array index 1 and FI refers to I1)
NEWPOS =I2
IF( NEWPOS.GT.LASTPOS 2 ) THEN
NPEAKS =NPEAKS 1
IPOS(NPEAKS) =NEWPOS
LASTPOS =NEWPOS
IF( NPEAKS .EQ. MAXP ) RETURN
ENDIF
ENDIF
ENDIF
ENDIF
100 CONTINUE
RETURN
END
The routine is optimized for speed and requires no other arrays than the spectrum and a
table to store the peak maxima found. This is achieved by using a variant of the top-hat
Copyright 2002 Marcel Dekker, Inc.
lter; that is, for a lter width of 5, the coecients are 71, 71, 1, 1, 1, 1, 1,
71, 71, 71. The next point in the ltered spectrum can be calculated from the current
by subtracting and adding and only the value of the previous, the current, and the next
points in the ltered spectrum are retained. This makes the routine quite cryptic, but it
works very fast and is reliable.
C. ContinuumEstimation
1. Peak Stripping
The subroutine SNIPBG, a variant of the SNIP algorithm, calculates the continuum via peak
stripping (see Sec. IV.A).
SUBROUTINE SNIPBG (Y, YBACK, NCHAN, ICH1, ICH2, FWHM, NITER)
INTEGER*2 NCHAN, ICH1, ICH2, NITER
REAL*4 Y(0:NCHAN-1), YBACK(0:NCHAN-1), FWHM
PARAMETER (SQRT2 =1.4142, NREDUC =8)
C -- Smooth spectrum
IW =NINT (FWHM)
I1 =MAX (ICH1 7IW, 0)
I2 =MIN (ICH2 IW, NCHAN-1)
CALL SGSMTH (Y, YBACK, NCHAN, I1, I2, IW)
C -- Square root transformation over required spectrum region
DO 10 I =I1, I2
YBACK(I) =SQRT (MAX(YBACK(I), 0.))
10 CONTINUE
C -- Peak stripping
REDFAC =1.
DO 30 N =1, NITER
C.. Set width, reduce width for last NREDUC iterations
IF( N .GT. NITER-NREDUC ) REDFAC =REDFAC=SQRT2
IW =NINT (REDFAC*FWHM)
DO 20 I =ICH1, ICH2
I1 =MAX (I 7IW, 0)
I2 =MIN (I IW, NCHAN-1)
YBACK(I) =MIN(YBACK(I), 0.5* (YBACK(I1) YBACK(I2)))
20 CONTINUE
30 CONTINUE
C -- Back transformation
DO 40 I =ICH1, ICH2
YBACK(I) =YBACK(I)*YBACK(I)
INPUT: Y Spectrum
NCHAN Number of channels in the spectrum
ICH1,ICH2 First and last channels of region to calculate the continuum
FWHM Width parameter for smoothing and stripping algorithm, set it
to average FWHM of peaks in the spectrum, typical value 8.0
NITER Number of iterations of SNIP algorithm, typical 24
Output: YBACK Calculated continuum in the region ICH1ICH2
Comment: Uses subroutine SGSMTH
Copyright 2002 Marcel Dekker, Inc.
40 CONTINUE
RETURN
END
2. Orthogonal Polynomial Continuum
The subroutine OPOLBAC ts the continuum of a pulse-height spectrum using an ortho-
gonal polynomial. Continuum channels are selected by adjusting the weight of the t (see
Sec. IV.C).
The routine calls ADJWEIG to adjust the weights. Further, the subroutine ORTPOL is
used to t the polynomial. The iteration (adjustment of weights) stops when all coefcients
c
j
change less than one standard deviation or when the maximum number of iterations is
reached.
SUBROUTINE OPOLBAC (NPTS, X, Y, W, YBACK, WORK1, WORK2,
> NDEGR, A, B, C, COLD, SC, FAILED, RCHISQ, R)
INTEGER NPTS, NDEGR
REAL*4 X(NPTS),Y(NPTS),W(NPTS),YBACK(NPTS)
REAL*4 WORK1 (NPTS), WORK2 (NPTS)
REAL*4 A (NDEGR), B(NDEGR), C(NDEGR), COLD(NDEGR), SC(NDEGR)
REAL*4 RCHISQ, R
LOGICAL*2 FAILED
PARAMETER (MAXADJ =20)
LOGICAL*2 NEXT
C -- Initialize
DO 10 J =1, NDEGR
COLD(J) =0.
10 CONTINUE
DO 20 I =1, NPTS
YBACK(I) =0.
20 CONTINUE
C -- Main iteration loop
DO 100 K =1, MAXADJ
C .. Calculate weights
CALL ADJWEIG (NPTS, Y, W, YBACK, R, NBPNTS)
Input: NPTS Number of data points (channels)
X Array of channel numbers
Y Array of spectrum
R Adjustable parameter [Eqs. (51) and (52)], typical value 2
Output: YBACK Array of fitted continuum
W Array of weights
A,B Coefficients of the orthogonal polynomial
C Fitted parameters of the orthogonal polynomial
SC Uncertainties of C
FAILED Logical variable TRUE if no convergence after MAXADJ weight
adjustments
RCHISQ Reduced chi-square value of the fitted continuum
Workspace: WORK1, WORK2 of size NPTS
Copyright 2002 Marcel Dekker, Inc.
C .. Fit orthogonal polynomial
CALL ORTPOL (NPTS, X, Y, W, YBACK, WORK1, WORK2,
> NDEGR, A, B, C, SC, SUMSQ)
RCHISQ =SUMSQ=FLOAT (NBPNTS 7NDEGR)
S =SQRT (RCHISQ)
C .. Test if further adjustments of weights is required
NEXT =.FALSE.
DO 30 J =1, NDEGR
SC(J) =S * SC(J)
IF( ABS(COLD(J)-C(J)) .GT. SC(J) ) NEXT =.TRUE.
COLD(J) =C(J)
30 CONTINUE
C .. convergence
IF( .NOT.NEXT ) THEN
FAILED =.FALSE.
RETURN
ENDIF
100 CONTINUE
C -- No convergence after MAXADJ iterations
FAILED =.TRUE.
RETURN
END
SUBROUTINE ADJWEIG (NPTS, Y, W, YFIT, R, NBPNTS)
C * * Adjust weights to emphasize the continuum
INTEGER NPTS, NBPNTS
REAL*4 Y(NPTS), W(NPTS), YFIT(NPTS), R
NBPNTS =0
C -- Loop over all data points
DO 10 I =1, NPTS
IF (YFIT(I) .GT. 0.) THEN
IF (Y(I).LE. YFIT(I) R*SQRT (YFIT(I))) THEN
C .. Point is considered as continuum
W(I) =1./YFIT(I)
NBPNTS =NBPNTS 1
ELSE
C .. Point is NOT considered as continuum.
W(I) =1./(Y(I) 7YFIT(I))**2
ENDIF
ELSE
C .. Continuum <=0, weight based original data
C (initial condition)
W(I) =1./MAX(Y(I),1.)
NBPNTS =NBPNTS 1
ENDIF
10 CONTINUE
RETURN
END
D. Filter-Fit method
The C program FILFIT is a test implementation of the lter-t method (see Sec. VI). This
program simply coordinates all input and output, allocates the required memory, and calls
two FORTRAN routines that do the actual work. The subroutine TOPHAT returns the
Copyright 2002 Marcel Dekker, Inc.
convolute of the spectrum with the top-hat lter or the weights (the inverse of the variance
of the ltered spectrum). The general-purpose subroutine LINREG is called to perform the
multiple linear least-squares t. The output includes the reduced w
2
value, the parameters
of the t a
j
(which is an estimate of the ratio of the intensity in the analyzed spectrum to
the intensity in the standard for the considered X-ray lines), and their standard deviation.
The routine GETSPEC reads the spectral data and must be supplied by the user.
#include <stdio.h>
#include <malloc.h>
#include <float.h>
#include <math.h>
void fortran TOPHAT( );
void fortran LINREG( );
float spec[2048];
main( )
{
int nchan, first_ch_fit, last_ch_fit, width, ierr;
int i, first_ch_ref, last_ch_ref, ref, num_ref, num_points;
int filter_mode =0, weight_mode =1, ioff;
float meas_time, ref_meas_time, *scale_fac;
float *x, *xp, *y, *w, *yfit, *a, *sa, chi;
double *beta, *alpha;
char filename [64];
// input width of tophat filter
scanf(%hd, &width);
// input spectrum to fit and fitting region
scanf(%s, filename);
scanf(%hd %hd, &first_ch_fit, &last_ch_fit);
nchan =Getspec(spec, filename, &meas_time);
num_points =last_ch_fit 7first_ch_fit 1;
// filter spectrum and store in y[ ]
y =(float *)calloc(num_points, sizeof(float));
TOPHAT(spec, y, &nchan, &first_ch_fit, &last_ch_fit, &width, &filter_mode);
//calculate weights of fit and save in w[ ]
w =(float *)calloc(num_points, sizeof(float));
TOPHAT(spec, w, &nchan, &first_ch_fit, &last_ch_fit, &width, &weight_mode);
//read reference spectra, filter and store in x[ ]
scanf(%hd, &num_ref);
scale_fac =(float *)calloc(num_ref, sizeof(float));
x =(float *)calloc(num_points*num_ref, sizeof(float));
for(ref =0; ref <num_ref; ref) {
scanf(%s, filename);
nchan =Getspec(spec, filename, &ref_meas_time);
scale_fac[ref] =ref_meas_time/meas_time;
scanf(%hd %hd, &first_ch_ref, &last_ch_ref);
if (first_ch_ref <first_ch_fit) first_ch_ref =first_ch_fit;
if(last_ch_ref >last_ch_fit) last_ch_ref =last_ch-fit;
ioff =ref*num_points first_ch_ref 7first_ch_fit;
xp =x ioff;
TOPHAT(spec, xp, &nchan, &first_ch_ref, &last_ch_ref, &width, &filter_mode);
}
Copyright 2002 Marcel Dekker, Inc.
// perform least squares fit
yfit =(float *)calloc(num_points, sizeof(float));
a =(float *)calloc(num_ref, sizeof(float));
sa =(float *)calloc(num_ref, sizeof(float));
beta =(double *)calloc(num_ref, sizeof(double));
alpha =(double *)calloc(num_ref* (num_ref 1)/2, sizeof(double));
LINREG (y,w,x &num_points, &num_ref, &num_points, &num_ref, yfit, a, sa, &chi,
&ierr, beta, alpha);
if( ierr = =0 ) {
printf(Filter fit: Chi-square =%f\n, chi);
printf(Standard Int. in analyse spectrum/Int. in standard\n);
for(i =0; i <num_ref; i )
printf( %hd %f n%f\n, i 1, a[i]*scale_fac[i], sa[i]*sacle_fac[i]);
for(i =0; i <num_points; i ){
printf(%4hd %7.0f %9.2f, first_ch_fit i, y[i],yfit[i]);
for (ref =0;ref <num_ref; ref )
printf(%7.0f, x[ref * num_points i]);
printf(\n);
}
}
}
SUBROUTINE TOPHAT (IN, OUT, NCHAN, IFRST, ILAST, IWIDTH, MODE)
INTEGER*2 NCHAN, IFRST, ILAST, IWIDTH, MODE
REAL*4 IN (NCHAN), OUT(1)
C ** Tophat filter of width IWIDTH, MODE =0 calculate filtered specturm,
C MODE ! =0 calculate weights (1/variance of filtered spectrum)
C - - Calculate filter constants.
IW =IWIDTH
IF(MOD(IW,2).EQ.0)IW =IW 1
FPOS =1./FLOAT(IW)
KPOS =IW/2
IV =IW/2
FNEG = 71./FLOAT(2*IV)
KNEG1 =IW/2 1
KNEG2 =IW/2 IV
N =0
C - - Loop over all requested channels.
DO 30 I =IFRST 1, ILAST 1
C . . Central positive part,
YPOS =0.
DO 10 K = 7KPOS, KPOS
IK =MIN(MAX(I K,1),NCHAN)
YPOS =YPOS IN(IK)
10 CONTINUE
C . . Left and right negative part,
YNEG =0.
DO 20 K =KNEG1, KNEG2
IK =MIN(MAX(I 7K,1),NCHAN)
YNEG =YNEG IN(IK)
IK =MIN(MAX(I K,1),NCHAN)
YNEG =YNEG IN(IK)
Copyright 2002 Marcel Dekker, Inc.
20 CONTINUE
N =N 1
IF(MODE.EQ.0) THEN
OUT(N) =FPOS * YPOS FNEG * YNEG
ELSE
VAR =FPOS*FPOS*YPOS FNEG*FNEG*YNEG
OUT(N) =1./MAX(VAR,1.)
ENDIF
30 CONTINUE
RETURN
END
E. Fitting Using Analytical Functions
The C program NLRFIT is an example implementation of the nonlinear spectrum tting
using an analytical function (see Sec. VII). The program only coordinates input and
output. The actual tting is done using the Marquardt algorithm with the FORTRAN
subroutine MARQFIT. The tting function consists of a polynomial continuum width NB
terms and NP Gaussians. The continuum parameters and the area, position, and with of
each Gaussian are optimized during the t. The tting function is calculated using the
routine FITFUNC. The derivatives of the tting function with respect to the parameters are
calculated by the routine DERFUNC.
// Program NLRFIT
#include <stdio.h>
#include <malloc.h>
#include <float.h>
#include <math.h>
#define MAX_PERKS 10
#define MAX_CHAN 1024
void fortran MARQFIT();
float spec [MAX_CHAN];
// Fortran common block structure COMMON/FITFUN/NB, NP
struct common_block {short NB, NP;};
extern struct common_block fortran FITFUN;
main( )
{
char specfile[64];
int nchan, first_ch_fit, last_ch_fit, nb, np;
int i, j, n, num_points, num_param, ierr, max_iter;
float ini_pos[MAX_PEAKS], ini_wid[MAX_PEAKS];
float *x, *xp, *y, *w, *yfit, *a, *sa, chi, lamda, crit_dif;
float *b, *beta, *deriv, *alpha;
double *work;
// Input of parameters and spectral data
scanf(%s, specfile);
scanf(%hd %hd %hd %f, &first_ch_fit, &last_ch_fit, &max_iter, &crit_dif);
scanf(%hd %hd, &np, &nb);
for(i =0; i <np; i )
scanf(%f %f, &ini_pos[i], &ini_wid[i]);
Copyright 2002 Marcel Dekker, Inc.
nchan =GetSpec(spec, specfile);
num_points =last_ch_fit 7first_ch_fit 1;
num_param =nb 3*np;
//Allocate memory for y[ ], w[ ], x[ ]
y =(float *)calloc(num_points, sizeof(float));
w =(float *)calloc(num_points, sizeof(float));
x =(float *)calloc(num_points, sizeof(float));
// Store independent variable (spectrum), weights and dep. var (channel #)
for(i =first_ch_fit, n =0; i < =last_ch_fit; i , n ) {
y[n] =spec[i];
w[n] =(spec[i] >0.) ? 1./spec[i] : 1.;
x[n] =(float)i;
}
// allocate memory for other arrays required
yfit =(float *)calloc(num_points, sizeof(float));
a =(float *)calloc(num_param, sizeof(float));
sa =(float *)calloc(num_param, sizeof(float));
b =(float *)calloc(num_param, sizeof(float));
beta =(float *)calloc(num_param, sizeof(float));
deriv =(float *)calloc(num_param, sizeof(float));
alpha =(float *)calloc(num_param*(num_param 1)/2, sizeof(float));
work =(double *)calloc(num_param*(num_param 1)/2, sizeof(double));
// initialize, all linear parameters to zero, peak position and width
// to their initial guesses
lamda =0.001;
for(i =0; i <np; i ){
a[nb np i] =ini_pos[i];
a[nb 2*np i] =ini_wid[i];
}
// perform least squares fit
FITFUN.NP =np;
FITFUN.NB =nb;
MARQFIT(&ierr, &chi, &lamda, &crit_dit, &max_iter, x, y, w, yfit, &num_points, a,
sa, &num_param, b, beta, deriv, alpha, work);
if( ierr = =0 ) {
printf(\nNon-linear fit: Chi-square =%f\n, chi);
printf(Polynomial continuum parameters\n);
for(i =0; i <nb; i )
printf(%hd %f n %f\n, i 1, a[i], sa[i]);
printf(Peak parameters Area Position Width\n);
for(i =0; i <np; i ){
printf(%hd %10.0f n % 710.0f, i 1, a[nb i], sa[nb i]);
printf(%10.3f n % 710.3f, a[nb np i], sa[nb np i]);
printf(%10.3f n % 710.3f\n, a[nb 2*np i], sa[nb 2*np i]);
}
for (i =0; i <num_points; i )
printf(%4hd %7.0f %9.2f\n, first_ch_fit i, y[i], yfit[i]);
}
}
Copyright 2002 Marcel Dekker, Inc.
SUBROUTINE FITFUNC(X, YFIT, NPTS, A, NTERMS )
REAL*4 X(NPTS), YFIT(NPTS), A(NTERMS)
COMMON /FITFUN/ NB, NP
C ** Fitting function, polynomial continuum and NP gaussians
C with position, width and area as parameters
PARAMETER(SQR2PI =2.50663)
C - - Loop over all channels
DO 100 I =1, NPTS
C . . continuum
YFIT(I) =A(1)
DO 20 J =2, NB
YFIT(I) YFIT(I) A(J) * X(I)**(J 71)
20 CONTINUE
C . . Peaks
DO 30 K =1, NP
AREA =A(NB K)
POS =A(NB NP K)
SWID =A(NB 2*NP K)
Z =((POS 7X(I))/SWID)**2
IF( Z.LT.50.) THEN
G =EXP( 7Z/2.)/SWID/SQR2PI
YFIT(I) =YFIT(I) AREA*G
ENDIF
30 CONTINUE
100 CONTINUE
RETURN
END
SUBROUTINE DERFUNC(X, NPTS, A, NTERMS, DERIV, I)
REAL*4 X(NPTS), A(NTERMS), DERIV(NTERMS)
COMMON/FITFUN/NB, NP
C ** Derivatives of fitting function: polynomial continuum and NP Gaussians
C with postion, width and area as parameters
PARAMETER (SQR2PI =2.50663)
C -- Derivatives of function with respect to the continuum parameters
DERIV(1) =1.
DO 10 J =2, NB
DERIV(J) =X(I)**(J 71)
10 CONTINUE
C -- Derivatives of function with respect to the peak parameters
DO 30 K =1, NP
AREA =A(NB K)
POS =A(NB NP K)
SWID =A(NB 2*NP K)
Z =((POS 7X(I))/SWID)**2
IF(Z.LT.50.) THEN
G =EXP( 7Z/2.)/SWID/SQR2PI
C . . Peak area
DERIV(NB K) =G
C . . Peak position
DERIV(NB NP K) = 7AREA*G*(POS 7X(I))/SWID/SWID
C . . Peak width
DERIV(NB 2*NP K) =AREA*G*(Z 71.)/SWID
ELSE
Copyright 2002 Marcel Dekker, Inc.
DERIV(NB K) =0.
DERIV(NB NP K) =0.
DERIV(NB 2*NP K) =0.
ENDIF
30 CONTINUE
RETURN
END
F. Monte Carlo Methods
1. UniformRandom-Number Generator
The function URAND is a FORTRAN function returning uniform distributed random
numbers in the interval 0 _U<0. The random-number generator is based on Knuths
subtractive method (Press et al., 1988) (see Sec. VIII. A)
REAL*4 FUNCTION URAND (ISEED)
INTEGER*2 ISEED
REAL*4 UTABLE(56)
REAL*4 UBIT, USEED
PARAMETER (UBIG =4000000., USEED =1618033.)
SAVE I1, I2, UTABLE, INIT
C -- Initialize table
IF (ISEED.LT.0 .OR. INIT.EQ.0) THEN
U =USEED FLOAT(ISEED)
U =MOD(U,UBIG)
UTABLE(55) =U
UTMP =1.
DO 10 I =1, 54
II =MOD(I*21, 55)
UTABLE(II) =UTMP
UTMP =U 7UTMP
IF(UTMP.LT.0.) UTMP =UTMP UBIG
U =UTABLE(II)
10 CONTINUE
DO 30 K =1, 4
DO 20 I =1, 55
UTABLE(I) =UTABLE(I) 7UTABLE(1 MOD(I 30,55))
IF(UTABLE(I).LT.0) UTABLE(I) =UTABLE(I) UBIG
20 CONTINUE
30 CONTINUE
I1 =0
I2 =31
ISEED =1
INIT =1
ENDIF
C -- Get next random number
I1 =I1 1
IF(I1.EQ.56) I1 =1
Input: ISEED Set to any negative number to initialize the random generator
Output: URAND Uniform random number in the interval 0 _URAND<1
Copyright 2002 Marcel Dekker, Inc.
I2 =I2 1
IF(I2.EQ.56) I2 =1
U =UTABLE(I1) 7UTABLE(I2)
IF(U.LT.0.)U =U UBIG
UTABLE(I1) =U
URAND =U/UBIG
RETURN
END
2. Normal Distributed RandomDeviate
The function NRAND returns normal distributed random number with zero mean and unit
variance, using the BoxMuller method (see Sec. VIII. B).
REAL*4 FUNCTION NRAND (ISEED)
INTEGER*2 ISEED
SAVE NEXT, FAC, V1, V2
IF (NEXT.EQ.0 .OR. ISEED.LT.0) THEN
10 CONTINUE
V1 =2. * URAND(ISEED) 71.
V2 =2. * URAND(ISEED) 71.
R =V1*V1 V2*V2
IF(R.G.E.1. .OR. R. EQ. 0.) GOTO 10
FAC =SQRT( 72. * LOG(R)/R)
NRAND =V1*FAC
NEXT =1
ELSE
NRAND =V2*FAC
NEXT =0
ENDIF
RETURN
END
3. Poisson Distributed RandomDeviate
The function PRAND can be used to produce approximately Poisson distributed random
deviates. For small numbers ( <20), the direct method is used; for larger numbers, the
Poisson distribution is approximated by the Normal distribution.
Input: ISEED Set to any negative number to initialize the random sequence
Output: NRAND Normal distributed random deviate with zero mean and unit
variance.
Input: Y (population) mean of deviate
ISEED Set to any negative number to initialize the random sequence
Output: PRAND Poisson distributed random deviate with mean Y
Copyright 2002 Marcel Dekker, Inc.
REAL *4 FUNCTION PRAND(Y, ISEED)
INTEGER*2 ISEED
REAL*4 Y
REAL*4 NRAND, URAND
IF(Y.LT.20.) THEN
C -- Use direct method
G =EXP( 7Y)
PRAND = 71.
T =1.
10 CONTINUE
PRAND =PRAND 1.
T =T * URAND(ISEED)
IF(T.GT.G) GOTO 10
ELSE
C -- Approximate by normal distribution
PRAND =Y SQRT(Y) * NRAND(ISEED)
ENDIF
RETURN
END
G. Least-Squares Procedures
1. Linear Regression
Subroutine LINREGis a general-purpose (multiple) linear regression routine (see Sec. IX. A).
SUBROUTINE LINREG (Y, W, X, N, M, NMAX, MMAX,
> YFIT, A, SA, CHI, IERR, BETA, ALPHA)
INTEGER*2 N, M, NMAX, MMAX, IERR
REAL*4 Y(N), W(N), YFIT(N), A(M), SA(M), CHI
REAL*4 X(NMAX, MMAX)
REAL*8 BETA(M), ALPHA(1)
c Accumulate BETA and ALPHA matrices
JK =0
DO 10 J =1, M
BETA(J) =0.0D0
DO 2 I =1, N
Input: Y Array of dependent variable
W Array of weights 1/s
2
i
)
X Matrix of independent variables
N Number of data points
M Number of independent variables (columns of X)
NMAX,MMAX Size of X matrix
Output: YFIT Array of fitted Y values
A Estimated least-squares parameters
SA Standard deviation of A
CHI Chi-square value
IERR Error condition, 1 if fit failed (singular matrix)
Workspace: BETA of size M
Copyright 2002 Marcel Dekker, Inc.
BETA(J) =BETA(J) W(I)*Y(I)*X(I, J)
2 CONTINUE
DO 6 K =1, J
JK =JK 1
ALPHA(JK) =0.0D0
DO 4 I =1, N
ALPHA(JK) =ALPHA(JK) W(I)*X(I,K)*X(I,J)
4 CONTINUE
6 CONTINUE
10 CONTINUE
c Invert ALPHA matrix
CALL LMINV (ALPHA, M, IERR)
IF(IERR .EQ. 71) THEN
RETURN
ENDIF
c Calculate fitting parameters A
DO 20 J =1, M
A(J) =0.
JJ =J*(J 71)/2
DO 12 K =1, J
JK =K JJ
A(J) =A(J) ALPHA(JK)*BETA(K)
12 CONTINUE
DO 14 K =J+1, m
JK =J K* (K 71)/2
A(J) =A(J) ALPHA(JK)*BETA(K)
14 CONTINUE
20 CONTINUE
c Calculate uncertainties in the parameters
DO 30 J =1, M
JJ =J*(J 1)/2
SA(J) =DSQRT(ALPHA(JJ))
30 CONTINUE
c Calculate fitted values and Chi-square
CHI =0.
DO 40 I =1, N
YFIT(I) =0.
DO 32 J =1, M
YFIT(I) =YFIT(I) A(J)*(I, J)
32 CONTINUE
CHI =CHI W(I)*(YFIT(I) 7Y(I))**2
40 CONTINUE
CHI =CHI/FLOAT(N 7M)
RETURN
END
2. Orthogonal Polynomial Regression
Subroutine ORTPOL ts an orthogonal polynomial to a set of data points [x
i
, y
i
, w
i
] (see
Sec. IX.B).
Copyright 2002 Marcel Dekker, Inc.
SUBROUTINE ORTPOL (NPTS, X, Y, W, YFIT, PJ, PJMIN, NDEGR, A, B, C, SC,
> SUMSQ)
INTEGER NPTS, NDEGR
REAL*4 X(NPTS), Y(NPTS), W(NPTS), YFIT(NPTS)
REAL*4 PJ(NPTS), PJMIN(NPTS)
REAL*4 A(NDEGR), B(NDEGR), C(NDEGR), SC(NDEGR), SUMSQ
C -- Initialize
DO 10 I =1, NPTS
PJ(I) =1.
PJMIN(I) =0.
YFIT(I) =0.
10 CONTINUE
GAMJMIN =1.
C -- Loop over all polynomial terms
DO 100 J =1, NDEGR
C .. Accumulate normalization factor, A and B constants for term j
GAMJ =0.
A(J) =0.
B(J) =0.
DO 20 I =1, NPTS
GAMJ =GAMJ W(I)*PJ(I)*PJ(I)
A(J) =A(J) W(I)*X(I)*PJ(I)*PJ(I)
B(J) =A(J) W(I)*X(I)*PJ(I)*PJMIN(I)
20 CONTINUE
A(J) =A(J) / GAMJ
B(J) =B(J) / GAMJMIN
C .. Least squares estimate of coefficient C
C(J) =0.
DO 30 I =1, NPTS
C(J) =C(J) W(I)*Y(I)*PJ(I)
30 CONTINUE
C(J) =C(J)=GAMJ
SC(J) =SQRT(1.=GAMJ)
C .. Contribution of this term to the fit
DO 40 I =1, NPTS
YFIT(I) =YFIT(I) C(J)*PJ(I)
40 CONTINUE
C .. Next polynomial term
IF (J .LT. NDEGR) THEN
Input: NPTS Number of data points
X Array of independent variable
Y Array of dependent variable
W Array of weights (w
i
=1s
2
i
)
NDEGR Degree of orthogonal polynomial to be fitted
Output: A, B Parameters of the orthogonal polynomials
C Fitted orthogonal polynomial coefficients
SC Standard deviation of C
SUMSQ Chi-square value
Workspace: PJ, PJMIN of size NPTS
Copyright 2002 Marcel Dekker, Inc.
DO 50 I =1, NPTS
PJPLUS =(X(I) 7A(J))*PJ(I) 7B(J)*PJMIN(I)
PJMIN(I) =PJ(I)
PJ(I) =PJPLUS
50 CONTINUE
GAMJMIN =GAMJ
ENDIF
100 CONTINUE
C -- Weighted sum of squares value
SUMSQ =0.
DO 110 I =1, NPTS
SUMSQ =SUMSQ W(I)*(Y(I) 7YFIT(I))**2
110 CONTINUE
RETURN
END
3. Nonlinear Regression
The subroutine MARQFIT performs nonlinear least-squares-tting according to the Mar-
quardt algorithm (see Sec. IX.C).
The routine requires two user-supplied subroutines: FITFUNC to evaluate the tting
function y(i) with the current set of parameters a and the DERFUNC to calculate the de-
rivatives of the tting function with respect to the parameters.
SUBROUTINE MARQFIT (IERR, CHISQR, FLAMDA, CRIDIF, MAXITER,
> X, Y, W, YFIT, NPTS, A, SA, NTERMS,
> B, BETA, DERIV, ALFA, ARR)
INTEGER*2 IERR, NPTS, NTERMS
REAL*4 CHISQR, FLAMDA, CRIDIF
REAL*4 X(NPTS), Y(NPTS), W(NPTS), YFIT(NPTS)
REAL*4 A(NTERMS), SA(NTERMS)
REAL*4 B(1), BETA(1), DERIV(1), ALFA(1)
REAL*8 ARR(1)
PARAMETER (FLAMMAX =1E4, FLAMMIN =1E 76)
Input: MAXITER Maximum number of iterations
X Array of independent variable
Y Array of dependent variable
W Array of weights (w
i
= 1=s
2
i
)
NPTS Number of data points
NTERMS Number of parameters
A Array of initial values of the parameters
Output: IERR Error status, 71 indicates failure of fit
CHISQR Reduced chi-square value
FLAMDA Marquardt control parameter
YFIT Array of fitted data points
A Least-squares estimate of the fitting parameters
SA Standard deviation of A
CRIDIF Minimum percent difference in two chi-square values
to stop the iteration
Workspace: B, BETA, DERIV of size NTERMS
ALFA, ARR of size NTERMS + (NTERMS 1)=1
Copyright 2002 Marcel Dekker, Inc.
C -- Evaluate the fitting function YFIT for the current parameters
C and save the Chi-square value
NITER =0
CALL FITFUNC(X, YFIT, NPTS, A, NTERMS)
CHISQR =CHIFIT(Y, YFIT, W, NPTS, NTERMS)
FLAMDA =0.
C -- Set ALFA and BETA to zero, save the current value of the parameters A
100 CONTINUE
NITER =NITER 1
CHISAV =CHISQR
DO 110 J =1, NTERMS
B(J) =A(J)
BETA(J) =0.
110 CONTINUE
DO 112 J =1, NTERMS*(NTERMS 1)=2
ALFA(J) =0.
112 CONTINUE
C -- Accumulate Alpha and Beta matrices
DO 120 I =1, NPTS
D =Y(I) 7YFIT(I)
C .. Calculate derivatives at point i
CALL DERFUNC (X, NPTS, A, NTERMS, DERIV, I)
DO 120 J =1, NTERMS
BETA(J) =BETA(J) W(I)*D*DERIV(J)
JJ =J*(J 71)=2
DO 120 K =1, J
JK =JJ K
ALFA(JK) =ALFA(JK) W(I)*DERIV(J)*DERIV(K)
120 CONTINUE
C -- Test and scale ALFA matrix
DO 140 J =1, NTERMS
JJ =J*(J 71)=2
JJJ =JJ J
IF(ALFA(JJJ) .LT. 1.E 720) THEN
DO 130 K =1, J
JK =JJ K
ALFA(JK) =0.
130 CONTINUE
ALFA (JJJ) =1.
BETA(J) =0.
ENDIF
SA(J) =SQRT (ALFA(JJJ))
140 CONTINUE
DO 160 J =1, NTERMS
JJ =J*(J 71)=2
DO 150 K =1, J
JK =JJ K
ALFA(JK) =ALFA(JK)=SA(J)=SA(K)
150 CONTINUE
160 CONTINUE
C -- Store ALFA in ARR, modify the diagonal elements with FLAMDA
200 CONTINUE
DO 210 J =1, NTERMS
Copyright 2002 Marcel Dekker, Inc.
JJ =J*(J 71)=2
DO 205 K =1, J
JK =JJ K
ARR (JK) =DBLE(ALFA(JK))
205 CONTINUE
JJJ =JJ J
ARR(JJJ) =DBLE(1. FLAMDA)
210 CONTINUE
C -- Invert matrix ARR
CALL LMINV (ARR, NTERMS, IERR)
IF (IERR .NE. 0) RETURN
C -- Calculate new values of parameters A
DO 220 J =1, NTERMS
DO 220 K =1, NTERMS
IF(K .GT. J) THEN
JK =J K*(K 71)=2
ELSE
JK =K J*(J 71)=2
ENDIF
A(J) =A(J) ARR(JK) = SA(J)*BETA(K) = SA(K)
220 CONTINUE
C -- Evaluate the fitting function YFIT for the new parameters and Chi-square
CALL FITFUNC(X, YFIT, NPTS, A, NTERMS)
CHISQR =CHIFIT(Y, YFIT, W, NPTS, NTERMS)
IF (NITER .EQ. 1) FLAMDA =0.001
C -- Test new parameter set
IF (CHISQR .GT. CHISAV) THEN
C .. Iteration NOT succesful, increase flamda and try again
FLAMDA =MIN(FLAMDA * 10., FLAMMAX)
DO 300 J =1, NTERMS
A(J) =B(J)
300 CONTINUE
GOTO 200
ENDIF
C .. Iteration succesful, decrease LAMDA
FLAMDA =MAX(FLAMDA=10., FLAMMIN)
C .. Get next better estimate if required
PERDIF =100.*(CHISAV 7CHISQR)=CHISQR
IF (NITER .LT. MAXITER .AND. PERDIF .GT. CRIDIF) GOTO 100
C -- Calculate standard deviations and return
DO 320 J =1, NTERMS
JJ =J*(J 2)=2
SDEV =DSQRT(ARR(JJ)) = SA(J)
SA(J) =SDEV
320 CONTINUE
RETURN
END
FUNCTION CHIFIT(Y, YFIT, W, NPTS, NTERMS)
REAL*4 Y(NPTS), YFIT(NPTS), W(NPTS)
C ** Evaluate chi-square
CHI =0.
DO 300 I =1, NPTS
Copyright 2002 Marcel Dekker, Inc.
CHI =CHI W(I) * (Y(I) 7YFIT(I))**2
300 CONTINUE
CHI =CHI = FLOAT(NPTS 7NTERMS)
CHIFIT =CHI
RETURN
END
4. Matrix Inversion
Subroutine LMINV is a general-purpose routine to invert a symmetric matrix.
SUBROUTINE LMINV ( ARR, N, IERR )
INTEGER*2 N, IERR
REAL*8 ARR(1)
REAL*8 DIN, WORK, DSUM, DPIV
INTEGER*2 I, IND, IPIV, J, K, KEND, KPIV, L, LANF,
> LEND, LHOR, LVER, MIN
KPIV =0
DO 10 K =1, N
KPIV =KPIV K
IND =KPIV
LEND =K 71
DO 4 I =K, N
DSUM =0.DO
IF (LEND .GT. 0) THEN
DO 2 L =1, LEND
DSUM =DSUM ARR(KPIV 7L) * ARR(IND 7L)
2 CONTINUE
ENDIF
DSUM =ARR(IND) 7DSUM
IF (I .EQ. K) THEN
IF (DSUM .LE. 0.D0) THEN
IERR = 71
RETURN
ENDIF
DPIV =DSQRT(DSUM)
ARR(KPIV) =DPIV
DPIV =1.D0 = DPIV
ELSE
ARR(IND) =DSUM * DPIV
ENDIF
IND =IND I
4 CONTINUE
10 CONTINUE
IERR =0
IPIV =N*(N 1)=2
Input: ARR Upper triangle and diagonal of real symmetric matrix stored in
linear array, size =N+(N1)=2.
N Order of matrix (number of columns)
Output: IERR Error status, IERR=0 inverse obtained, IERR=1 singular
matrix
Copyright 2002 Marcel Dekker, Inc.
IND =IPIV
DO 20 I =1, N
DIN =1.DO = ARR(IPIV)
ARR(IPIV) =DIN
MIN =N
KEND =I 71
LANF =N 7KEND
IF (KEND .GT. 0) THEN
J =IND
DO 14 K =1, KEND
WORK =0.D0
MIN =MIN 71
LHOR =IPIV
LVER =J
DO 12 L =LANF, MIN
LVER =LVER 1
LHOR =LHOR L
WORK =WORK ARR(LVER) * ARR(LHOR)
12 CONTINUE
ARR(J) = 7WORK * DIN
J =J 7MIN
14 CONTINUE
ENDIF
IPIV =IPIV 7MIN
IND =IND 71
20 CONTINUE
DO 30 I =1, N
IPIV =IPIV I
J =IPIV
DO 24 K =I, N
WORK =0.D0
LHOR =J
DO 22 L =K, N
LVER =LHOR K 7I
WORK =WORK ARR(LVER) * ARR(LHOR)
LHOR =LHOR L
22 CONTINUE
ARR(J) =WORK
J =J K
24 CONTINUE
30 CONTINUE
IERR =0
RETURN
END
REFERENCES
Aarnio PA, Lauranto H. Nucl Instrum Methods A276:608 (1989).
Bertin EP. Principles and Practice of X-ray Spectrometric Analysis. New York: Plenum Press, 1970.
Bevington PR, Robinson DK. Data Reduction and Error Analysis for the Physical Sciences.
New York: McGraw-Hill, 1992.
Black WW. Nucl Instrum Methods 71:317, 1969.
Copyright 2002 Marcel Dekker, Inc.
Bloomeld DJ, Love G. X-Ray Spectrom 14:8, 1985.
Bombelka E, Koenig W, Richter FW. Nucl Instrum Methods B22:21, 1987.
Breschinsky R, Krush E, Wehrse R. Diplomarbeit, Fachbereich Physik. Bremen: Universita t
Bremen, 1979.
Brook D, Wynne RJ. Signal Processing Principles and Applications. London: Edward Arnold, 1988.
Campbell JL. X-Ray Spectrom 24:307, 1995.
Campbell JL. Nucl Instrum Methods B109=110:71, 1996.
Campbell JL, Perujo A, Millman BM. X-Ray Spectrom 16:195, 1987.
Campbell JL, Maxwell JA, Papp T, White G. X-Ray Spectrom 26:223, 1997.
Campbell JL, Millman BM, Maxwell JA, Perujo A, Teesdale WJ. Nucl Instrum Methods B9:71,
1985.
Campbell JL, Maenhaut W, Bombelka E, Claytan E, Malmqvist K, Maxwell JA, Pallon J,
Vandenhaute J. Nucl Instrum Methods B14:204, 1986.
Campbell JL, Cauchon G, Le py M-C, McDonald L, Plagnard J, Stemmler P, Teesdale WJ, White G.
Nucl Instrum Methods A418:394, 1998.
Cirone R, Gigante GE, Gualtieri G. X-Ray Spectrom 13:110, 1984.
Clayton E, Duerden P, Cohen DD. Nucl Instrum Methods B22:64, 1987.
Doster JM, Gardner RP. X-Ray Spectrom 11:173, 1982a.
Doster JM, Gardner RP. X-Ray Spectrom 11:181, 1982b.
Duy CJ, Rogers PSZ, Benjamin TM. Nucl Instrum Methods B22:91, 1987.
Enke CG, Nieman TA. Anal Chem 48:705A, 1976.
Fiori CE, Myklebust RL, Gorlen K. NBS Spec Pub 604, 233, 1981.
Gardner RP, Yacout AM, Zhang J, Verghese K. Nucl Instrum Methods A242:299, 1986.
Geladi P, Kowalski BR. Anal Chim Acta 185:1, 1986.
Gertner I, Heber O, Zajfman J, Zajfman D, Rosner B. Nucl Instrum Methods B36:74, 1989.
Gunnink R. Nucl Instrum Methods 143:145, 1977.
Heckel J, Scholz W. X-Ray Spectrom 16:181, 1987.
Hertogen J, De Donder J, Gijbels R. Nucl Instrum Methods 115:197, 1974.
Hoskuldsson A. J Chemometr 2:211, 1988.
Jansson PA. Deconvolution with Applications in Spectroscopy. New York: Academic Press, 1984.
Janssens K, Van Espen P. Anal Chim Acta 184:117, 1986.
Janssens K, Dorrine W, Van Espen P. Chemometr Intell Lab Syst 4:147, 1988.
Janssens K, Vincze L, Van Espen P, Adams F. X-Ray Spectrom 22:234, 1993.
Jenkins R, Gould RW, Gedcke D. Quantitative X-Ray Spectrometry. New York: Marcel Dekker,
1981.
Jensen BB, Pind N. Anal Chim Acta 117:101, 1985.
Johansson GI. X-Ray Spectrom 11:194, 1982.
Joy DC. Rev Sci Instrum 56:1772, 1985.
Kajfosz J, Kwiatek WM. Nucl Instrum Methods B22:78, 1987.
Lemberge P, Van Espen PJ. X-Ray Spectrom 28:77, 1999.
Lorber A. Wangen LE, Kowalski BK. J Chemometr 1:19, 1987.
Lucas-Tooth HJ, Price BJ, Metallurgia 64: 149, 1961.
Maenhaut W, Vandenhaute J. Bull Soc Chim Belg 95:407, 1986.
Manne R. J Chemometr Intell Lab Syst 2:187, 1987.
Marageter E, Wegscheider W, Mu ller K. Nucl Instrum Methods B1:137, 1984a.
Marageter E, Wegscheider W, Mu ller K. X-Ray Spectrom 13:78, 1984b.
Marquardt DW. J Soc Ind Appl Math 11:431, 1963.
Martens H, Naes T. Multivariate Calibration. Chichester: John Wiley, 1989.
Massart DL, Vandeginste BGM, Deming SN, Michotte Y, Kaufman K. Chemometrics: A
Textbook. Amsterdam: Elsevier, 1988.
McCarthy JJ, Schamber FH. NBS Spec Publ 604:273, 1981.
McCrary JH, Singman LV, Ziegler LH, Looney LD, Edmonds CM, Harris CE. Phys Rev
A4:1745, 1971.
Copyright 2002 Marcel Dekker, Inc.
McCullagh H. Report EGG-PHYS-5890, Idaho National Engineering Laboratory, Idaho Falls, ID,
1982.
McNelles LA, Campbell JL. Nucl Instrum Methods 127:73, 1975.
Molt K, Schramm R. Fresenius J Anal Chem 359:61, 1997.
Nielson KK. X-Ray Spectrom 7:15, 1978.
Nullens H, Van Espen P, Adams F. X-Ray Spectrom 8:104, 1979.
Nunez J, Rebollo Neira LE, Plastino A, Bonetto RD, Gue rin DMA, Alvarez AG. X-Ray Spectrom
17:47, 1988.
Op De Beeck JP, Hoste J. Atomic Energy Rev 13:151, 1975.
Pella PA Feng L, Small JA. X-Ray Spectrom 14:125, 1985.
Pessara W, Debertin K. Nucl Instrum Methods 184:497, 1981.
Petersen W, Ketelsen P, Kno chel A. Nucl Instrum Methods A245:535, 1986.
Phillips GW. Nucl Instrum Methods 153:449, 1978.
Philips GW, Marlow KW. Nucl Instrum Methods 137:525, 1976.
Pratar A, De Jong S. J Chemometr 11:311, 1997.
Press WH, Flannery BP, Teukolsky SA, Vetterling WT. Numerical Recipes in C, The Art of
Scientic Computing. Cambridge: Cambridge University Press, 1988.
Reed SJB, Ware NG. J Phys E5:582, 1972.
Robertson A, Prestwich WV, Kennett TJ. Nucl Instrum Methods 100:317, 1972.
Routti JT, Prussin SG. Nucl Instrum Methods 72:125, 1969.
Ryan CG, Clayton E, Grin WL, Sie SH, Cousens DR. Nucl Instrum Methods B34:396, 1988.
Salem SI, Wimmer RJ. Phys Rev A2:1121, 1970.
Salem SI, Saunders BG, Melson C. Phys Rev A1:1563, 1970.
Savitzky A, Golay MJE. Anal Chem 36:1627, 1964.
Schamber FH. In: Dzubay T, ed. X-Ray Fluorescence Analysis of Environmental Analysis. Ann
Arbor, MI: Ann Arbor Science, 1977, p. 241.
Schreier F. J Quant Spectros Radiat Transfer 48:743, 1992.
Schwalbe LA, Trussell HJ. X-Ray Spectrom 10:187, 1981.
Scoeld JH. Phys Rev 179:9, 1970.
Scoeld JH. Phys Rev A9:1041, 1974a.
Scoeld JH. Phys Rev A10:1507, 1974b.
Sherry WM, Vander Sande JB. X-Ray Spectrom 6:154, 1977.
Smith DGW, Gold CM, Tomlinson DA. X-Ray Spectrom 4:149, 1975.
Statham PJ. X-Ray Spectrom 5:16, 1976a.
Statham PJ. X-Ray Spectrom 5:154, 1976b.
Statham PJ. X-Ray Spectrom 7:132, 1978.
Statham PJ, Nashashibi T. In: Newbury DE, ed. Microbeam Analysis. San Francisco: San Francisco
Press, 1988, p. 50.
Steenstrup S. J Appl Crystallogr 14:226, 1981.
Swerts J, Van Espen P. Anal Chem 65:1181, 1993.
Urba nski P, Kowalska E. X-Ray Spectrom 24:70, 1995.
Van Dyck P, Van Grieken R. X-Ray Spectrom 12:111, 1983.
Van Espen P, Adams F. X-Ray Spectrom 5:123, 1976.
Van Espen P, Nullens H, Adams F. Anal Chem 51:1325, 1979a.
Van Espen P, Nullens H, Adams F. Nucl Instrum Methods 142:243, 1977a.
Van Espen P, Nullens H, Adams F. Nucl Instrum Methods 145:579, 1977b.
Van Espen P, Nullens H, Adams F. X-Ray Spectrom 9:126, 1980.
Van Espen P, Janssens K, Nobels J. Chemometr Intell Lab Syst 1:109, 1986.
Van Espen P, Nullens H, Maenhaut W. In: Newbury DE, ed. Microbeam Analysis. San Francisco,
San Francisco Press, 1979b, p. 265.
Van Espen P, Nullens H, Adams F. Anal Chem 51:1580, 1979c.
Vekemans B, Janssens K, Vincze L, Adams F, Van Espen P. X-Ray Spectrom 23:275, 1994.
Vekemans B, Janssens K, Vincze L, Adams F, Van Espen P. Spectrochim Acta 50B:149, 1995.
Copyright 2002 Marcel Dekker, Inc.
Verghese K, Mickael M, He T, Gardner RP. Adv X-Ray Anal 31:461, 1988.
Vincze L, Janssens K, Adams F. Spectrochim Acta 48B:553, 1993.
Vincze L, Janssens K, Adams F, Jones KW. Spectrochim Acta 50B:1481, 1995a.
Vincze L, Janssens K, Vekemans B, Adams F. Spectrochim Acta 54B:1711, 1999.
Vincze L, Janssens K, Adams F, Rivers ML, Jones KW. Spectrochim Acta 50B:127, 1995b.
Wa tjen U. Nucl Instrum Methods B22:29, 1987.
Wilkinson DH. Nucl Instrum Methods 95:259, 1971.
Wielopolski L, Gardner RP. Nucl Instrum Methods 165:297, 1979.
Yacout AM, Dunn WL. Adv X-Ray Anal 30:113, 1987.
Yacout AM, Gardner RP, Verghese K. Nucl Instrum Methods A243:121, 1986.
Yacout AM, Gardner RP, Verghese K. Adv X-Ray Anal 30:121, 1987.
Yule PH. Nucl Instrum Methods 54:61, 1967.
Zolnai L, Szabo , Gy. Nucl Instrum Methods B34:118, 1988.
Copyright 2002 Marcel Dekker, Inc.
5
Quantication of Innitely Thick Specimens
by XRF Analysis
Johan L. de Vries*
Eindhoven, The Netherlands
Bruno A. R. Vrebos
Philips Analytical, Almelo, The Netherlands
I. INTRODUCTION
Quantitative x-ray uorescence (XRF) analysis involves the conversion of measured
uorescent intensities to the concentration of the analytes. In most of the literature on this
subject, the measured intensities at this stage of the analytical procedure are assumed to be
corrected for background and line overlap. The same assumptions will be made here,
although some comments will be made on these topics throughout this chapter. Fur-
thermore, energy-dispersive and wavelength-dispersive x-ray spectrometers dier sig-
nicantly with respect to the counting channel and the resolution or resolving power that
can be achieved. This has a direct inuence on, for example, the line overlap correction
procedures that need to be applied (if any) to obtain net intensities. Wavelength-dispersive
x-ray uorescence spectrometry has been discussed in Chapter 2; energy-dispersive systems
are the subject of Chapters 3 (for x-ray tube excitation) and 7 (for radioisotope excitation).
The software provided by the vendors of x-ray spectrometry systems for their customers
compensates to some extent for the peculiarities of each design and considerable dier-
ences in the software packages oered with the instruments are to be expected. It is
therefore dicult to give guidelines that are generally applicable.
In some cases, most notably those where the intensity of the background is essen-
tially constant (between specimens), the correction for background can be neglected. This
is usually the case when the dierence in terms of composition of the standard specimens
and the unknowns is rather small [i.e., the concentration ranges (not necessarily the
concentration levels) of all components are rather small]. This condition of similarity
between specimens is sucient to ensure constant background, but it is not a requirement,
as, in certain cases, the intensity of the background is (nearly) constant even when the
concentration range of one or more analytes is quite large. Note that the only requirement
*Retired.
Copyright 2002 Marcel Dekker, Inc.
here is for the background to be constant; at this point, no restrictions on the magnitude of
the background have been imposed. Under these circumstances, correction for back-
ground is not required and the gross count rate (or intensity) can be used with equal
success as the net intensity for the determination of the specimens composition.
The situation with line overlap is somewhat more complicated, as it is almost never
constant. This is due to the intensity of the overlapping line(s): It is very rarely a constant
for dierent specimens. There are several methods to correct for line overlap, and the
preference of the individual analyst as well as the type of spectrometer and its associated
software (if any) are some of the factors inuencing the nal choice.
The basic principles of quantitative analysis have not changed much since the early
years of x-ray uorescence analysis, especially for the wavelength-dispersive systems.
Energy-dispersive systems have seen more changes due to, for example, development of
dierent solid-state detectors and primary optics, such as capillaries (see Chapter 11), and
the increasingly more easily available computing power which allowed for more com-
prehensive spectral treatment programs at the ngertips of the analysts (see also
Chapter 4). In general, the method has always involved some calibration step, during
which the intensities of selected elements in a suite of standard specimens are measured
according to a scheme or a recipe that has been optimized earlier. These intensities are
combined with the composition of the standard specimens, and the calibration curves for
the elements of interest are then constructed. These calibration curves might include a
correction for matrix eects of some sort. The resulting calibrations could then be used for
the analysis of unknown specimens with compositions similar to the standards. This
method is rather time-consuming, but it yields, in general, the highest possible analytical
accuracy. Also, approximate knowledge of the composition of the specimens prior to the
analysis is assumed. Such assumptions can be made very implicitly (e.g., when selecting the
calibration curve to be used for analysis). If the method is routinely used and the number
of unknown specimens to be analyzed is rather large, the benets of this instrumental
technique become very clear, when compared to classical wet chemistry methods, which
tend to be much more time-consuming.
Since the second half of the 1980s, interest has been growing for a dierent kind of
quantitative analysis on wavelength-dispersive spectrometers, in which the elements pre-
sent in the specimens and their approximate concentration levels are not known prior to
the analysis. Semiquantitative analysis consists of an intimate combination of qualitative
and quantitative analyses. During the qualitative analysis, the presence of the elements
present in the specimen is established. This can be done in a variety of ways. Qualitative
analysis is based on the collection of a spectrum of the specimen, followed by peak search
and peak identication. Alternatively, the intensities of a rather large and xed set of
characteristic lines are collected. The lines are selected by the manufacturer and allow for
the determination of most of the commonly occurring elements. In both cases, the net
intensities are then determined and a quantication is made, involving matrix correction
and general calibration lines for each of the elements. These calibration lines have been
determined earlier from a set of standard specimens, although some manufacturers insist
on calling their method standardless. The packages that are available commercially
dier on the basis of the qualitative analysis (accumulating a spectrum or measurement on
xed wavelength positions), the number of standard specimens used, the exibility of the
spectrometer conguration that they can cope with, and the method used to correct for
matrix eect.
The focus of energy-dispersive systems has always been somewhat more qualitative
and more oriented toward research, exibility, and the analysis of small batches of
Copyright 2002 Marcel Dekker, Inc.
unknown specimens. The software allowed for quantitative analysis without the need to
have standards similar to the unknowns. These systems are thus more exible in their use
in that respect. This does not mean that such systems are unsuitable for routine analysis in
industry; many of them actually are used for that purpose.
In this chapter, several methods for the quantication of innitely thick specimens
are discussed. For all methods described here, it is assumed that the specimens are
innitely thick: The intensity of the radiation from the specimen is constant and is not
aected by increasing the thickness of the specimen. The analysis of specimens with less-
than-innite thickness is discussed in Chapter 6. Also, the specimens are assumed to be
homogeneous: Their composition is the same throughout the specimen. This assumption is
valid for a large range of material types found in many applications. The most important
exception are specimens with one or more thin coatings.
II. CORRELATIONBETWEENCOUNT RATE ANDSPECIMENCOMPOSITION
A. Introduction
In quantitative analysis, the measured x-ray uorescent intensity of a given element is
converted into its weight concentration in the specimen. As a rst approximation, one
would expect a linear relationship. Each atom of the analyte element i has the same
probability of being excited by the primary photons and emitting its characteristic photons
with wavelength l
i
. Indeed, if separate atoms or ions (e.g., in a gas or in a very diluted
solution) are considered, the following relation holds:
I
i
K
i
W
i
1
where I
i
is the measured intensity of the uorescent radiation of the analyte i and W
i
is the
weight fraction of the analyte i in the specimen.
The proportionality constant K
i
consists of many physical and instrumental factors
such as the following:
The intensity and the distribution of wavelengths of the photons in the primary beam
The probability that an atom i emits its characteristic radiation l
i
The probability that these photons l
i
pass through the measuring channel:
collimators, diracting crystal, pulse-height window
The probability that these photons are being detected and registered.
For a given instrument, the voltage and power on the x-ray tube remain constant for the
analyte i and they can be determined by measuring the uorescent intensity of the pure
element i. However, we are generally dealing with compact specimens where the atoms are
arranged into chemical compounds. Both the primary x-rays and the uorescent x-rays
will be absorbed by the dierent atoms in the specimen.
B. General Relationship Between Intensity and Concentration
1. Primary Fluorescenceby Monochromatic Radiation
First, excitation by monochromatic radiation will be considered. Let the intensity of the
incident beam with wavelength l
0
at the surface of the specimen be given by I
0
(l
0
). This
beam strikes the surface of the specimen at an angle c
0
(see also Fig. 1). A parallel beam is
assumed here and the specimen is considered to be extending to innity in all three
dimensions. The incident radiation is gradually absorbed by the specimen, and at a layer at
Copyright 2002 Marcel Dekker, Inc.
depth t below the surface, the remaining fraction of the intensity I
t
(l
0
) is given by the
LambertBeer law:
I
t
l
0
I
0
l
0
expm
s
l
0
r
s
t csc c
0
2
m
s
(l
0
) is the mass-attenuation coecient of the specimen for photons with wavelength l
0
(the subscript s refers to the specimen) and r
s
is the density of the specimen. Note that due
to the angle of incidence, the path length traveled is given by t csc c
0
. The mass-attenuation
coecient m
s
(l
0
) in this equation is thus calculated for the specimen; this is done by adding
the mass-attenuation coecients for all elements j present in the specimen, each multiplied
with its mass fraction W
j
:
m
s
l
0
X
n
j1
m
j
l
0
W
j
3
where n is the total number of elements present in the specimen. The fraction of the in-
cident beam absorbed by the analyte i in the layer between t and (t dt) is given by
W
i
m
i
l
0
r
s
csc c
0
dt 4
It is assumed here that the composition of the specimen is uniform throughout. In other
words, W
i
and W
j
are independent of the position of the layer (t, t dt) within the
specimen.
Only a fraction of the photons absorbed creates vacancies in the K shell; this fraction
is given by (r
iK
1),r
iK
, where r
iK
is the absorption jump ratio of the K shell of element i.
The fraction of K vacancies emitting x-rays is given by the uorescence yield o
iK
. The
fraction of Ka photons in the total of x-rays emitted for the analyte is given by the
transition probability f
iKa
. These factors can be combined in a factor designated Q
i
(l
0
, l
i
):
Q
i
l
0
. l
i
o
i
r
iK
1
r
iK
f
iKa
5
In the above derivation, it is assumed that the characteristic line of interest is a Ka. If
another line is considered, the relevant changes need to be made.
The characteristic photons thus generated are isotropically emitted in all directions,
without a preferential direction. Only a fraction is emitted toward the detector. If O is the
solid angle as viewed by the collimator detector system, expressed in steradians, then the
fraction is given by O,4p. The angle O should be small enough so that the beam can be
considered to be a parallel beam leaving the specimen at a single well-dened angle c
00
with the surface. The fraction of characteristic photons with wavelength l
i
not absorbed
between the layer (t, t dt) and the surface is given by
Figure 1 Schematic representation of the geometry involved in the calculation of primary
fluorescence emission.
Copyright 2002 Marcel Dekker, Inc.
expm
s
l
i
r
s
t csc c
00
6
All photons reaching the surface and propagating in the direction indicated by the
angle c
00
are assumed to be detected. If the detector has absorbing elements (such as
windows) or detection eciencies dierent from unity, then these can also be taken into
account. Also, the attenuation by the medium (e.g., air) between specimen and detector
can be calculated using similar expressions.
The intensity of element i, as excited by the incident beam with wavelength l
0
, is
labeled P
i
(l
0
) (explicitly denoting the primary uorescence eect) and is given by
P
i
l
0
I
0
l
0
expm
s
l
0
r
s
t csc c
0
m
i
l
0
W
i
Q
i
l
0
. l
i
r
s
cscc
0
dt
O
4p
expm
s
l
i
r
s
t csc c
00
7
Combining factors leads to
P
i
l
0
I
0
l
0
m
i
l
0
W
i
Q
i
l
0
. l
i
csc c
0
O
4p
expm
s
l
0
csc c
0
m
s
l
i
csc c
00
r
s
tr
s
dt 8
The contribution of all layers between the surface and the bottom of the specimen have to
be summed. This canbe done by integratingthe above expressionover dt, from0 (the surface)
to the bottom. In practice, for x-rays, the thickness of bulk specimens can be considered to
be innite, so the upper limit is 1. Note that t is always used in combination with r
s
, so the
integration will be done over r
s
t. Taking all constant factors outside the integral, one obtains
P
i
l
0
I
0
l
0
m
i
l
0
W
i
csc c
0
Q
i
l
0
. l
i
O
4p
Z
1
0
expm
s
l
0
csc c
0
m
s
l
i
csc c
00
r
s
tr
s
dt 9
From a textbook on calculus,
Z
1
0
expax dx
1
a
10
Taking a (m
s
(l
0
) csc c
0
m
s
(l
i
) csc c
00
) and noting that r
s
dt d(r
s
t), the following
expression is obtained for the primary uorescence of the analyte i in the specimen s:
P
i
l
0
I
0
l
0
m
i
l
0
W
i
csc c
0
Q
i
l
0
. l
i
O,4p
m
s
l
0
csc c
0
m
s
l
i
csc c
00
11
Often, the element specic factors, given by Eq. (5), are combined with the instrument
specic factor O,4p. This leads to a simpler expression for the primary uorescence:
P
i
l
0
K
i
I
0
l
0
m
i
l
0
m
s
l
0
Gm
s
l
i
W
i
12
where K
i
O,4pQ
i
l
0
. l
i
and G csc c
00
,csc c
0
sin c
0
, sin c
00
.
In the above derivation, the following assumptions have been made:
(a) The specimen is completely homogeneous.
(b) The specimen extends to innity in three dimensions.
Copyright 2002 Marcel Dekker, Inc.
(c) The primary rays are not scattered on their way to the layer dt.
(d) No enhancement eects occur.
(e) The characteristic radiation is not scattered on its way to the specimen surface.
The simplest case is excitation of Ka (or Kb) photons. For the characteristic photons
associated with the L lines, other eects, such as CosterKronig transitions and so forth
have to be taken into account when describing the fraction of absorbed primary photons
that give rise to characteristic photons. For a more detailed discussion on these eects,
refer to Chapter 1.
2. Secondary Fluorescence Excitedby Monochromatic Radiation
Under certain conditions, the characteristic photons of an element j can excite atoms of
the analyte i. This will lead to additional characteristic photons of element i. The term
additional is used here in the sense that these photons are not considered in the above
derivation for the primary uorescence and, consequently, are not predicted by Eq. (12). It
is beyond the scope of this chapter to derive the mathematical expressions following a
rigorous approach, but certain aspects of secondary uorescence emission are readily seen
from the above derivation and from Figure 2. In Figure 2, two layers are now considered.
The layer on the left, indicated by j, at a depth of t
1
with a thickness of dt
1
, is where the
primary uorescence of element j is excited. This is described by Eqs. (2)(5). Obviously,
the element considered at this stage is the enhancing element j, so the fundamental
parameters for j need to be used instead of these for i in Eqs. (2)(5). Once the uorescent
radiation is created, it will be absorbed on its way through the specimen. The precise
direction is of no concern, as we are now no longer considering the fraction that travels
toward the detector. The attenuation along its path is described by equations such as
Eq. (2) or Eq. (6). Keep in mind that the angles and the distances involved are now dif-
ferent and that the wavelength l for which the mass attenuation needs to be calculated
[Eq. (3)] is now l
j
instead of l
0
. At the second layer, indicated by i, at a depth of t
2
with a
thickness of dt
2
, part of this (primary uorescence) radiation will be absorbed. Again, this
eect is described by an equation such as Eq. (4), with the angle c
0
replaced by the angle of
this path. The factors to be applied in order to lead to characteristic radiation of the
analyte i are given by Q
i
(l
j
, l
i
); Eq. (6) describes the attenuation (wavelength is now l
i
) on
the way toward the detector. These factors can easily be recalculated. The complication is
in the description of the geometry, especially in the description of the distance betwee
n
the
two layers: The position of the two layers relative to one another is not restricted, the path
of the photon l
j
does not have to be in the plane of the gure, nor is the second layer
Figure 2 Schematic representation of the geometry involved in the calculation of secondary
fluorescence emission.
Copyright 2002 Marcel Dekker, Inc.
always at a larger depth than the rst layer. The angle X
0
in Figure 2 can take any value
between 0 and 2p, and the three paths shown are not necessarily in one plane. This aspect
has been discussed by Li-Xing (1984).
Anyway, after integration, the nal result S
ij
l
0
. l
j
is given by
S
ij
l
0
. l
j
I
0
l
0
csc c
0
m
j
l
0
W
j
Q
j
l
0
. l
j
m
i
l
j
W
i
Q
i
l
j
. l
i
O,4p
2m
s
l
0
csc c
0
m
s
l
i
csc c
00
sin c
0
m
s
l
0
ln 1
m
s
l
0
m
s
l
j
sin c
0
!
sin c
00
m
s
l
j
ln 1
m
s
l
j
m
s
l
j
sin c
00
!
13
where S
ij
l
0
. l
j
describes the secondary uorescence (enhancement) of the analyte i by
characteristic photons with wavelength l
j
; these photons have been excited by primary
photons with wavelength l
0
.
The photons to be considered for the enhancement of the analyte are not limited to
Ka only, as the Kb lines and other characteristic lines that have sucient energy to excite
the shell of interest of the analyte i must also be taken into account. For this reason, all
individual contributions need to be added in order to calculate the total enhancement of
the analyte i by element j:
S
ij
l
0
X
i
S
ij
l
0
. l
j
14
There are two criteria to be satised for a characteristic photon l
j
to be able to cause
secondary uorescence:
1. It must be excited by the incident photon l
0
(this means that the energy of the
incident photon must be higher than the energy of the absorption edge
associated with l
j
).
2. The energy of the photon l
j
must be higher than the energy of the absorption
edge for the analyte.
The derivation of the intensities of characteristic x-rays as a function of specimen
composition has rst been done by Sherman (1955). The resulting equations, however,
were rather unwieldy. Equations (9) and (13) have been published later by Shiraiwa and
Fujino (1966) and Sparks (1976).
In practice, many computer programs only consider a few lines for enhancement,
and lines that have a low transition probability f are usually neglected. Data regarding
these transition probabilities for a line within a series can be found, for example, in
Appendix II of Chapter 1. The enhancement phenomenon will be more pronounced if the
x-rays of the enhancing elements are only slightly more energetic than the energy of the
absorption edge of the element i. The enhancement may contribute up to 4050% of
the total uorescent radiation I
i
, especially where the concentration of the enhancing
elements is much greater than the concentration of the analyte. This applies even more for
the light elements, where the primary spectrum may not be very eective as the most
intense wavelengths are far away from the absorption edges of these light elements.
The eect of scattered radiation is generally ignored, although its contribution can
also be calculated (Pollai et al., 1971).
If the characteristic line of the analyte considered is one of its L lines, then it is
possible that other lines of the analyte itself enhance the characteristic line considered. The
energy of the K edge of La (atomic number Z57) is 38.9 keV (see Chapter 1), so the K
lines of all elements with Z<57 are excited if, for example, an x-ray tube is used at
Copyright 2002 Marcel Dekker, Inc.
voltages above 40 keV, as is commonly the case with wavelength-dispersive spectrometers.
Under these conditions, the L lines of these elements are enhanced by the K lines. The
situation for elements with high atomic numbers is more complex. Let us consider Pb
(Z82). Lead is very often determined through its La
1
or Lb
1
line because the K lines of
Pb are too energetic to be measured with reasonable eciency by scintillation or Si(Li)
detectors. Also, excitation voltages higher than 88 keV are required in order to excite the K
lines. Thus, one would not readily expect enhancement by Pb on its own L lines. From the
data in Table 1, however, it follows that the PbLa
1
line (L
3
M
5
) is enhanced by PbLg
1
(L
2
N
4
) because its energy (14.76 keV) is higher than the energy of the L
3
edge (13.04 keV).
The total excited characteristic intensity is then given by adding the primary and the
secondary uorescence contributions:
I
i
l
0
P
i
l
0
X
j
S
ij
l
0
. l
j
15
Also, tertiary uorescence is possible, where the incident photon excites element k (pri-
mary uorescence), whose radiation excites element j (secondary uorescence), whose
radiation, in turn, excites element i, causing tertiary uorescence. This contribution is
generally lower than 3% of the total uorescence and is commonly ignored, as shown by
Shiraiwa and Fujino (1967, 1974).
3. Excitationby Continuous Spectra
If the incident beam is polychromatic rather than monochromatic, Eq. (15) needs to be
calculated for each wavelength. Wavelengths longer than the wavelength of the absorption
edge l
edge,i
of the analyte cannot excite uorescence, so these need not to be considered. If
Jl is the function representing the tube spectrum, then Eqs. (11) and (13) can still be
used, provided that I
0
l
0
is replaced by Jl, representing the intensity of the incident
spectrum at wavelength l:
I
i
Z
l
edge. i
l
min
I
i
Jl dl
Z
l
edge. i
l
min
P
i
l
X
j
S
ij
l. l
j
" #
dl 16
This equation allows the calculation of the intensity of characteristic radiation of a given
analyte in an innitely thick specimen. Equations such as Eqs. (11), (14), and (16) are often
referred to as fundamental parameter equations because they allow the calculation of the
intensity of uorescent radiation as a function of the composition of the specimen (weight
fractions W
i
), the incident spectrum [Jl], and the conguration of the spectrometer used
(c
0
. c
00
, and O). All other variables used are fundamental constants, such as the mass-at-
tenuation coecients for a given element at a given wavelength or its uorescence yield
and so forth.
Table1 Data for Selected Absorption Edges and Characteristic Line of Pb
Absorption edge Energy (keV) Characteristic line Energy (keV)
K 88.04
L
1
15.86 La
1
(L
3
M
5
) 10.55
L
2
15.20 Lb
1
(L
2
M
4
) 12.61
L
3
13.04 Lg
1
(L
2
N
4
) 14.76
Source: Appendix I and Appendix II of Chapter 1 (this volume).
Copyright 2002 Marcel Dekker, Inc.
C. Some Observations
The integration over t in Eq. (9) is taken from zero to innity. It is obvious that the rst
layers contribute more to the intensity of l
i
than the more inward layers. Theoretically,
even at large values of t, a very minor contribution to the intensity is still to be expected.
Often, the (minimum) innite depth is dened arbitrarily as that thickness t where the
contribution of the layer t. t dt is 0.01% of that of the surface layer. In this case, it is
dened relative to the surface layer. Alternatively, it is dened as the thickness where the
contribution to the total intensity is less than 1%. In this case, it is relative to the total
intensity from a truly innitely thick specimen. The value of the innite thickness de-
pends on the value of the absorption coecients and the density of the specimen. In
practice, it may vary from a few micrometers for heavy matrices and long wavelengths to
centimeters for short wavelengths and light matrices, as in solutions.
For a given element and a xed geometry, an eciency factor Cl
0
. l
i
can be in-
troduced:
Cl
0
. l
i
m
i
l
0
P
n
j1
W
j
m
j
l
0
G
P
n
j1
W
j
m
j
l
i
17
It is obvious that the terms
P
W
j
m
j
are the origin of nonlinear calibration lines, as
variations in W
j
inuence the value of the denominator. For a pure metal, this reduces to
Cl
0
. l
i
m
i
l
0
m
i
l
0
Gm
i
l
i
18
In the general case, the wavelengths in the primary spectrum close to the absorption
edge are the most eective in exciting the analyte i. The eciency factor Cl
0
. l
i
is thus a
combination of the absorption curve of analyte i as a function of l and the spectral dis-
tribution. In a rst approximation, an eective wavelength l
e
can be introduced, which
has the same eect of excitation of element i as the total primary spectrum. The exact value
of this l
e
will be inuenced by the characteristic tube lines if they are active in exciting i.
Otherwise, l
e
can, in general, be assumed to have a value of approximately two-thirds of
the absorption edge l
edge
. Its actual value is, however, dependent on the chemical com-
position of the specimen. For instance, for Fe, the wavelength of the K edge l
edge
is
0.174 nm (7.11 keV) and an eective wavelength l
e
of 0.116 nm is obtained using this rule
of thumb. In the ZnOFe
2
O
3
system, l
e
was found to vary from 0.130 nm for 100% Fe
2
O
3
to 0.119 nm for 10% Fe
2
O
3
in ZnO. The estimated value of 0.116 nm is in good agreement
with the experimental one for Fe
2
O
3
.
Another interesting case is the analysis of a heavy element i in a light matrix. In
the summation in the denominator of Eq. (12), the terms W
i
m
i
l
0
and W
i
m
i
l
i
are
then the most important. The other terms can, in rst approximation, be neglected if
W
i
is not small. However, that means that the terms W
i
in the numerator and the
denominator cancel and the measured intensity becomes independent of W
i
; thus, the
analysis becomes impossible in this extreme case. A solution to this problem is found
by making the inuence of the terms W
i
m
i
l in the denominator less dominating by
adding a large term W
a
m
a
l. This can be done be making W
a
large (e.g., diluting) or
m
a
l large, by adding heavy absorber. Equation (17) enables one to calculate before-
hand how large this term should be to eliminate uctuations in the concentration of the
other elements.
In the derivation of the uorescence of the analyte i in the preceding paragraphs, the
following simplications were made:
Copyright 2002 Marcel Dekker, Inc.
1. First, it was assumed that the primary rays follow a linear path to the layer dt at
depth t. However, the primary rays may also be scattered. In general, the loss in
intensity of the primary beam of photons due to scattering may be neglected.
These scattering eects become more important when the primary x-rays are
more energetic and the average atomic number of the matrix decreases. This
scatter may give a higher background in the secondary spectrum, thus leading to
a poorer precision of the analysis. On the other hand, the excitation eciency
may be enhanced, as the primary rays dwell longer in the active layers, thus
having a higher probability to encounter atoms of element i. This eect may
overrule the increase in intensity of the background radiation. A case in hand is
the determination of Sn in oils which gives better results using the SnK lines at a
high x-ray tube voltage, than using the SnL lines at moderate voltages.
Incidentally, this scattering of the primary radiation makes it possible to check
the voltage over the x-ray tube. According to Braggs law, the intensity of the
primary spectrum is zero at an angle y
0
, given by
y
0
sin
1
nl
min
2d
crystal
19
where 2d
crystal
is the 2d spacing of the crystal used and n is an integer number.
l
min
(in nm) is given by
l
min
1.24
V
20
where V is the voltage on the x-ray tube, expressed in kilovolts. In practice, a
lower value for V will be found when Compton scattering dominates over
Rayleigh scattering and thus l
min
found has a value too low by the Compton
shift, which is about 0.024 nm for most spectrometers; the actual value depends
on the incidence and exit angle (see also Sec. V.C.1.)
2. The integral in Eq. (9) was taken from zero to innity; further, it was assumed
that the specimen is completely homogeneous. This, of course, is never realized
in practice, as we are dealing with discrete atoms in chemical compounds. In
powders, the dierent compounds may have a tendency to cluster. The particles
will, in general, have dierent sizes and shapes. Putting the sample into solution,
either aqueous or solid (melt) may overcome this problem.
It was stated formerly that innite thickness may vary from 20 mm to a few cen-
timeters. However, the most eective layers are much thinner. Thus, the number of dis-
crete particles actually contributing to the uorescent radiation may be rather small.
III. FACTORS INFLUENCINGTHE ACCURACYOF THE INTENSITY
MEASUREMENT
A. Introduction
The total uncertainty of the analysis consist of many errors whose source may be the
following:
The measurement of the intensity
The reproducibility of the specimen preparation
The conversion of intensity into concentration
Copyright 2002 Marcel Dekker, Inc.
The uncertainty of a single determination may be found from n determinations of the same
analysis, given a mean result x
mean
, of the all individual results x
i
. If n is suciently large,
the standard deviation s may be found from
s
P
n
i1
x
i
x
mean
2
n 1
s
21
In this total standard deviation s, random and systematic uncertainties are combined.
Random uncertainties give an indication of the precision of an analysis (the scatter of
results around the mean value), whereas systematic errors are the reason for deviations of
the mean value from the true value. An analysis may thus be precise, but not very
accurate, if systematic errors are present, whereas accurate values could be found from the
mean of widely scattered measurements if only large random uncertainties were present.
The total error of a measurement is composed of all the separate errors. If only random
errors s
1
, s
2
, and so forth are considered, the resulting standard deviation is given by
s
2
s
2
1
s
2
2
s
2
3
s
2
n
22
where s
1
, s
2
, and so on are the errors associated with, for example, intensity measurements,
specimen preparation, instrumental settings, and so on. In practice, it is often found that s
is dependent on the concentration of the analyte W
i
(Johnson, 1967; Hughes and Hurley,
1987):
s K
W
i
W
b
p
23
where W
b
is a small concentration oset (typically 0.001 in weight fraction). Thus, K,
rather than s, becomes an indication of the accuracy of the determination.
B. RandomErrors
1. Counting Statistics
If an x-ray measurement consisting of the determination of a number of counts N is re-
peated n times the results N
1
,N
2
, N
3
, . . . ,N
n
would spread about the true value N
0
. If n is
large, the distribution of the measurements would follow a Gaussian distribution,
WN
1
2pN
p exp
NN
0
2
2N
" #
24
provided N is also large. The standard deviation s of the distribution is equal to
N
mean
p
,
again if n and N are large, where N
mean
is the mean of n determinations. From the
properties of the Gaussian distribution, the following hold:
68.3% of all values N will be between N
0
s and N
0
s.
95.4% of all values N will be between N
0
2s and N
0
2s.
99.7% of all values N will be between N
0
3s and N
0
3s.
Similarly, there is a certain probability that the true result N
0
will lie between N
N
p
and N
N
p
, assuming the same distribution for N and N
0
. Measurement results are
commonly expressed as a count rate (intensity per unit time) instead of an intensity,
which gives the number of counts collected in the counting interval. This allows an easier
comparison of measurements made with dierent counting times, but the measuring
time needs to be specied in order to be able to assess the counting statistical error.
Copyright 2002 Marcel Dekker, Inc.
The determined concentration is dependent on the net count rate, which is the peak
count rate R
p
minus the background count rate R
b
. The total measuring time T equals
t
p
t
b
, where t
p
and t
b
are the measurement times for peak and background, respec-
tively. In modern equipment, there is no signicant statistical error in the measurement
of t. We can thus assume that R follows the same Gaussian distribution as N with the
same relative standard deviation e
N
. e
N
is dened as
e
N
s
N
N
25
Hence,
e
N
N
p
N
1
N
p
1
R
p
t
p e
R
26
and
s
R
e
R
R
R
p
t
p 27
It is obvious that the relative counting error decreases as t increases.
When a net count rate has to be determined the peak, R
p
and the background, R
b
,
have to be measured; there are thus two independent variables. The standard deviation of
the net intensity, s
d
, is given by
s
d
s
2
p
s
2
b
q
R
p
t
p
R
b
t
b
s
28
and the relative standard deviation e
d
by
e
d
R
p
,t
p
R
b
,t
b
p
R
p
R
b
29
When using a sequential wavelength-dispersive x-ray uorescence spectrometer and when
the count rates are rather low and time is limited, it is of interest to divide the total
counting time available over t
p
and t
b
in the best possible way. In principle, there are three
methods to split up the total counting time T:
1. Fixed time. Peak and background are measured for the same time;
t
p
t
b
T,2. The resulting standard deviation in the dierence between the
peak and the background count rate is then given by
s
d
2
T
r
R
p
R
b
q
30
2. Fixed count. The same number of counts, N, is collected on the peak and on the
background: N
p
N
b
or
t
p
t
b
R
b
R
p
31
The resulting standard deviation for the dierence is given by
s
d
1
T
r
R
p
R
b
R
p
R
b
R
b
R
p
s
32
Copyright 2002 Marcel Dekker, Inc.
3. Fixed time optimal. The optimum division of the total measurement time T over
t
p
and t
b
can be found by dierentiating Eq. (29) with respect to t
p
, where t
b
Tt
p
. It is found that for optimal results, the ratio of the measuring times is
t
p
t
b
R
b
R
p
s
33
The resulting standard deviation is given by
s
d
1
T
r
R
p
p
R
b
p
34
and the relative standard deviation by
e
d
1
T
p
1
R
p
p
R
b
p 35
It can easily be demonstrated that
s
FTO
s
FT
s
FC
36
where s
FTO
refers to the method of xed time optimized, s
FT
is for xed time, and s
FC
is
for xed counts. When R
p
is very large compared with R
b
, s
FTO
is close to s
FT
. Thus, the
xed time method is often used in practice because R
p
and R
b
are not known beforehand.
If there is a great dierence between peak and background count rate, then usually all the
available time is spend counting the peak. Thus, the background is not measured. A
constant value for the background can be assumed or it can be ignored altogether.
Compared with method of xed time, the following observation can be made: If, R
b
is not
measured, the standard deviation in the net count rate (calculated as the dierence
between R
p
and R
b
) is given by
s
d
R
p
T
s
2
b
r
37
This approximation is allowed if this value is smaller than
2,T
p
R
p
R
b
p
. This is the
case if s
b
5
R
b
,T
p
. This applies if the background is more or less constant between
specimens and a xed value for the background can be deducted. If the background is
altogether ignored, s
b
R
b
. This is only allowed when R
b
is much smaller than
R
p
,T
p
.
It follows that when one aims for optimum instrumental conditions, the expression
R
p
p
R
b
p
is a good quality function. This parameter is often used as a gure of merit
(FOM). Obviously, the highest value of
R
p
p
R
b
p
gives the best result. When it can be
assumed that
R
p
p
R
b
p
approximates R
p
, for low intensities and high background,
optimizing this term, equals optimizing
M
R
b
p 38
where M (the slope of the calibration line in counts per second per percent) is proportional
to R
p
R
b
.
If the ratio of two count rates R
1
and R
2
has to be determined, the methods of xed
time and xed count give the same result, whereas the method of optimum division, where
t
1
t
2
R
2
R
1
r
39
always gives the best result.
Copyright 2002 Marcel Dekker, Inc.
The total counting uncertainty is the combination of instrumental error and
counting statistics. When the count rate is very high, the counting uncertainty is small and
it may be worthwhile to apply a ratio method to reduce possible instrumental un-
certainties. However, if the count rate is low, it is better to spend all the available time in
analyzing the specimen to reduce the counting error.
2. Instrumental Errors
If the instrumental and counting uncertainties are random and independent variables, then
e
tot
e
2
instr
e
2
count
q
40
e
instr
e
2
tot
e
2
count
q
41
Although the counting error is inuenced by the instrumental error, Eq. (41) is still a good
approximation. s
tot
can be found from a series of repeated results of one measurement;
thus, s
count
is known. To check the instrumental instability, all the functions should be
measured separately. A radioactive source
55
Fe, for example, can be used to check the
detector and electronic circuitry. The x-ray tube can be checked by repeating the mea-
surements with specimen and goniometer in a xed position, eliminating errors stemming
from the mechanics of the spectrometer.
In another series of experiments, recycling between dierent angles checks the re-
producibility of the goniometer, whereas repositioning the specimen between measure-
ments checks the specimen holder, the reproducibility of the specimen loading mechanics,
and so forth. A comprehensive series of tests for wavelength-dispersive x-ray uorescence
spectrometers is described in the Australian Standard 2563-1982 (1982). Sometimes the
error s
tot
found is smaller than expected, or even smaller than s
count
. This may indicate that
an unexpected systematic error is involved, or there may be an uncorrected deadtime in the
equipments, (i.e., the true counting rate is higher than measured, which means that the
relative error is smaller).
3. Detection Limit
A characteristic line intensity decreases with decreasing concentration of the analyte and
nally disappears in the background noise. The true background intensity may be con-
stant, but the results of the measurements uctuate around a mean value R
bmean
. To be
signicantly dierent from the background, a signal R
p
must, although it is larger than
R
bmean
, be distinguished from the spread in R
b
. In other words, if we measure a signal R
p
larger than R
b
and we assume the analyte is present, what is the probability that our
assumption is correct? If the results of the measurements are random and follow a
Gaussian distribution, then this probability is determined by s
Rb
. If the measurement R
p
is
higher than R
b
2s
Rb
, then the probability that our assumption is correct is approxi-
mately 95% if a higher certainty is required (e.g., 99.7%), then R
p
should be larger than
R
b
3s
Rb
. Thus, the net intensity is 3s
Rb
and the detection limit, DL, would be
DL
3s
Rb
M
42
where M is the sensitivity in counts per second per percent. So, the detection limit in the
above equation is the concentration corresponding to a net peak intensity of 3s
Rb
.
However, in x-ray spectrometry, the background signal is specimen dependent and cannot
Copyright 2002 Marcel Dekker, Inc.
be measured independently, as in radioactivity measurements. Hence, R
b
has to be mea-
sured in an o-peak location in the spectrum. The result R
b
found in measuring time t s is
assumed to be R
bmean
and s
Rb
is assumed to be
R
b
,t
p
. Thus, two measurements have to
be made: R
p
and R
b
, each in time t s. The detection limit thus becomes
DL
3
2
p
M
R
b
T
r
43
where T 2t. If we are satised with a 95% probability that our assumption is correct,
then
DL
2
2
p
M
R
b
T
r
44
which is roughly equal to
DL
3
M
R
b
T
r
45
It is obvious that the detection limit decreases if the counting time increases. However, the
total error in R
b
contains the instrumental error as well. Thus, there is no sense to increase
the counting time when the instrumental error dominates.
Ingham and Vrebos (1994) have shown that the detection limit can be improved by
carefully selecting a primary lter. If the application of such a primary beam lter re-
duces the intensity of the (scattered) continuum from the tube more than it aects the
sensitivity, the detection limit is improved. The loss in sensitivity M needs to be more
than compensated for by the reduction in background intensity; as from Eq. (45), the
detection limit is proportional to the square root of R
b
and inversely proportional to the
sensitivity M.
4. Variationin X-ray Spectrum
The uorescent intensity of the analyte I
i
is, in rst approximation, dependent on the
primary spectrum according to
I
i
KiV
0
V
c
p
46
where K is a constant, i is the current of the x-ray tube, V
0
is the working voltage, V
c
the
excitation voltage, and p varies between 1 and 2, depending on the ratio of excitation by
characteristic tube rays and white continuum. In modern instruments, the tube voltage is
not dependent on the mains cycle, but they run on constant potential, which still may
uctuate. If the working voltage V
0
, or the region or line of highest excitation probability
is rather close to V
c
, then small uctuations in V
0
will introduce a considerable error in I
i
.
For instance, if V
0
1.5 V
c
, then a 1% error in V
0
gives an error in I
i
of 6% when p 2.
It is therefore better to run the tube at three to ve times the excitation voltage of
the analyte. A too high voltage, however, might introduce an unproportionally high
background.
5. Other Instrument Errors
Other possible random instrumental errors include positioning the specimen and setting
the goniometer in wavelength-dispersive x-ray spectrometers. These errors have to be
checked by repeated measurements of one specimen in a systematic way:
Copyright 2002 Marcel Dekker, Inc.
Repeated counting with stationary specimen and goniometer
Repeated counting with stationary specimen and repositioning goniometer
Repeated counting with stationary goniometer and repositioning specimen
Repeated counting with stationary goniometer and reloading specimen holder.
Some diracting crystals have a rather high coecient of thermal expansion;
therefore, their d value may uctuate with uctuations in temperature. This results in a
setting of the goniometer, slightly o-peak, which introduces a change in measured in-
tensity. Therefore, most modern equipment are operating at a stabilized temperature.
6. Particle Statistics
Only a limited volume of the specimen can actually contribute to the uorescent radiation.
As long as this active volume is the same in standards and actual specimens and the atomic
distribution is completely homogeneous, this poses no problem. However, the atoms are
bound into chemical compounds, forming nite particles with dierent chemical compo-
sitions. The analyte may only occur in particles with a certain chemical composition and
not in other particles. Then, only these specic particles can contribute to the uorescent
radiation of the analyte i. Therefore, the count rate R
i
measured depends on the number of
those particles present in the active volume, where, evidently, the rst layers contribute
most of the uorescent radiation.
Table 2 gives an indication of the penetration depth of radiation of various wa-
velengths into matrices with varying absorption power. It is evident that for most solid
specimens, the uorescent radiation originates within 20 mm or less from the surface. To
get an idea of how many particles can actually contribute to the uorescent intensity of
analyte i, let us assume that the irradiated area of 10 cm
2
is covered with cubic particles
of 100 mm dimension in a random fashion. Assuming a lling factor of 0.8 and assuming
that the analyte i is only present in 10% of the particles, then 10 10
8
10
4
10
1
8000
p
90 particles, or a relative
standard deviation of approximately 1.1%. If the concentration of analyte is only 1%,
then this relative standard deviation would be roughly 3.3%. In practice, these errors
might even be larger, as the radiation of the specimen is not homogeneous because the
primary spectrum originates in a rather small anode and passes through a large window
and is, thus, conically shaped. Spinning the specimen in its own plane during the ana-
lysis will reduce this error. Furthermore, the rst layers are the most eective, having
Table 2 Innite Thickness (in mm for Certain Analytical Lines as a Function of the Matrix)
Analytical line Fe base Mg base H
2
O solution Borate Borate La
2
O
3
10%
SnKa 300 10,000 100,000 70,000 10,000
MoKa 100 3,700 30,000 30,000 2,600
NiKa 12 340 2,400 2,000 300
CrKa 33 120 900 800 250
AlKa 1.5 4 10 15 5
NaKa 0.7 20 9 6 5
CKa 0.3 0.3 4 1 0.4
Note: Both the incidence and exit angles are 45
57
where m
s
l
0
and m
s
l
i
are the mass-attenuation coecients of the specimen considered
for wavelengths l
0
and l
i
, respectively, and G is the geometrical factor for both compo-
sitions. The relative dierence between these expressions should not exceed a few percent,
otherwise the matrix eects become too important to ignore. If the specimens include ux
or a binding agent, then these must also be considered in Eq. (57).
With increasing range of concentrations, deviations from linearity will be observed
due to variations in matrix eects between specimens and standards. The analyst must
then resort to other methods to obtain accurate results.
Matrix eects are studied most easily by considering binary systems (i.e., specimens
with only two elements or compounds).
In the case of absorption (both primary and secondary absorption must be con-
sidered), three cases can be distinguished (see Fig. 6):
1. A simple, linear relationship between relative intensity R and weight fraction W
(Curve 1 in Fig. 6). In this case, there is no matrix eect: The analyte and the
matrix have very similar (in principle, the same) attenuation coecients for the
incident and the characteristic radiations.
2. Curve 2 in Figure 6 is obtained when the matrix has a higher attenuation
coecient for the analytes characteristic radiation than the analyte itself: The
characteristic radiation is primarily absorbed by the matrix element. This is
usually called positive absorption.
3. Curve 3 in Figure 6 is obtained when the matrix absorbs less than the analyte
itself: The matrix has a smaller value for the attenuation coecient for the
characteristic radiation than the analyte itself. This can be the case, for example,
if the matrix has a much lower atomic number than the analyte, such as Mo in
Al. This eect is called negative absorption.
Enhancement will generally lead to a calibration graph like curve 4 in Figure 6. The
eect of enhancement is usually smaller than that of positive or negative absorption, as
indicated by the position, relative to curve 1, of curves 2 and 3 and curve 4.
It can be shown that the behavior of the calibration curves can be explained in terms
of attenuation coecients only if absorption is the only matrix eect. Furthermore,
if monochromatic excitation is used, a single constant (calculated from attenuation
Copyright 2002 Marcel Dekker, Inc.
coecients) suces to express the eect of one element on the intensity of another. In the
following section, various methods to deal with matrix eects will be discussed.
C. Elimination or Evaluation of theTotal Matrix Effect: Compensation Methods
1. Scattered Radiation: Compton Scatter (in cooperationwith Mark N. Ingham,
British Geological Survey, Keyworth, UK)
If the variation in matrix eects is mainly due to absorption, scattered x-rays can be used
to obtain an estimate of the absorption coecient of the specimen at a certain wavelength
l
s
. The intensity of the scattered radiation can be shown to be inversely proportional to the
mass-attenuation coecient m
s
of the specimen:
I
s
l
s
1
m
s
l
s
58
where l
s
is the wavelength of the scattered radiation and m
s
l
s
is the mass-attenuation
coecient of the specimen for wavelength l
s
. This is illustrated in Figure 7. The intensity
of the uorescent radiation, I
i
. is also inversely proportional to the mass-attenuation
coecient [Eq. (57)], but at a dierent wavelength:
I
i
W
i
m
s
l
0
Gm
s
l
i
W
i
m
s
59
where
m
s
m
s
l
0
Gm
s
l
i
60
Figure 6 Calibration curves for binaries. Curve 1: no net matrix-effect; curve 2: net absorption of
the analytes radiation by the matrix (positive absorption); curve 3: net absorption of the analytes
radiation by the analyte (negative absorption); curve 4: enhancement of the analytes radiation by
the matrix.
Copyright 2002 Marcel Dekker, Inc.
Mass-attenuation coecients at two dierent wavelengths are virtually proportional,
independent of matrix composition, provided there are no signicant absorption edges
between the two wavelengths considered (Hower, 1959). Hence, the ratio I
i
,I
s
is pro-
portional to the concentration of the analyte.
Both coherently and incoherently scattered primary radiation, such as tube lines
for tube excited x-ray uorescence (XRF), as well as the scattered continuous radiation
can be used. The method corrects also, to some degree, for surface nish, grain size
eects, and variations in tube voltage and current, but it does not correct for en-
hancement, thus limiting its use to analytes that are inuences by absorption only.
Furthermore, no absorption edges of major elements may be situated between the two
wavelengths considered. If that is the case, the ratio between the mass-attenuation
coecients for the wavelengths considered depends to a large degree on the concen-
tration of the major element(s) and the ratio is no longer constant between specimens.
This reduces the range of analytes that can be covered using the scattered tube lines. The
use of scattered continuum radiation close to the analyte peak may be disadvantageous
due to limited intensity, leading to either long measurement times or poor counting
statistics.
The intensity of the scattered Compton radiation is higher for specimens mainly
consisting of low-atomic-number elements than for specimens with a higher average
atomic number. This is illustrated in Figure 8, where the scattered radiation is plotted
as a function of wavelength for two specimens: iron and magnesium. In this experi-
ment, the spectrometer was equipped with a Rh anode x-ray tube. The two sharp
peaks that can be observed are the RhKb (at 0.0546 nm) and the RhKa (at 0.0613 nm),
respectively. The other two, much broader peaks are the Compton-scattered RhKb and
the RhKa. The Compton-scattered peaks are much broader than Rayleigh-scattered
Figure 7 The intensity of Compton-scattered radiation is inversely proportional to the mass-
attenuation coefficient (mac) of the specimen. Conditions: Rh tube at 60 kV.
Copyright 2002 Marcel Dekker, Inc.
characteristic lines. They are also shifted toward longer wavelengths by an amount Dl,
which is given by
Dl 0.002431 cos c 61
where c is the angle through which the radiation is scattered: c c
0
c
00
and Dl is ex-
pressed in nanometers.
In the spectrometer used for the recording of the spectra in Figure 8, c is 100
.
The Compton shift, Dl, is thus 0.0029 nm for this conguration. The maxima of the
Compton peaks (at 0.0576 nm and 0.0644 nm) in Figure 8 are in good agreement with
the theoretical values: 0.0546 nm0.0029 nm0.0575 nm and 0.0613 nm0.0029 nm
0.0642 nm, respectively.
The intensity of the Compton-scattered radiation is higher for shorter wavelengths
and for specimens consisting of elements with low atomic numbers. For a given wave-
length (e.g., the characteristic radiation of a tube line), the intensity of Compton scatter
decreases as the specimen consists of more and more elements with higher atomic
numbers. For specimens made up of oxides, the scattered intensity is usually so intense
that it can be measured with sucient precision in a relatively short time. On the other
hand, for specimens made up predominantly of heavier elements, such as steel and even
more so for brasses and solders, the intensity of the scattered radiation is very low (see
Fig. 8) and counting the statistical error can preclude precise analysis in a reasonable
amount of time. The most common application where this approach (or a variant
thereof) is used is in the determination of trace elements in specimens of geological
origin. This is illustrated in Figure 9 for the determination of Sr in specimens of widely
Figure 8 Intensity of scattered radiation as a function of wavelength. Two different specimens
have been used: iron and magnesium. A Rh anode tube was used in this experiment. Note the large
difference in intensity between the specimens. For the specimen consisting of the element with the
higher atomic number (Fe), the intensity of the Compton (incoherent)-scattered radiation is lower
than the intensity of the Rayleigh (coherent)-scattered radiation, compared to the specimen with the
lower-atomic-number element (Mg).
Copyright 2002 Marcel Dekker, Inc.
varying geological origin. In Figure 9a, the net count rate for SrKa is plotted against the
concentration for a large number of specimens. There is a considerable spread around
the calibration line established. The scatter is greatly reduced when the net count rate of
Figure 9 (a) Net count rate of SrKa as a function of Sr concentration for a large number of
specimens of varying geological origin. There is considerable spread around the calibration line.
(b) The ratio of the count rates of the SrKa radiation and the Compton-scattered tube line is plotted
against the concentration of Sr. The spread of the data points around the calibration line is now
much reduced compared to (a); this is especially the case for the point labeled A.
Copyright 2002 Marcel Dekker, Inc.
SrKa is divided by the count rate of the RhKa Compton-scattered tube radiation, as
indicated in Figure 9b.
Feather and Willis (1976) have shown that the intensity of the Compton peak can
also be used as an estimate for the background under characteristic lines; this eliminates
measuring the intensity of the background near the peak. With this method, not only is the
measurement time per specimen reduced (using WDS) but the dicult task of nding
interference-free background positions is no longer required.
2. Internal Standard
In this method, an element is added to each specimen in a xed proportion to the original
sample. This addition has to be made to the standard samples as well as to the unknowns.
The characteristic radiation of the element added should be similar to the characteristic
radiation of the analyte in terms of absorption and enhancement properties in the matrix
considered. Such an element is called an added internal standard, or internal stan-
dard for short. In practice, the method works equally well if a pure element or a pure
compound is added, or if a solution with the internal standard element is used. If a so-
lution is used, care must be taken that the solution itself does not contain any elements
that are to be analyzed. The composition of the solution used as the additive must be
constant, otherwise it might aect the matrix eect. The intensity of the internal standard
is aected by matrix eects in much the same way as the intensity of the analyte, provided
there are no absorption edges (leading to dierence in absorption) or characteristic lines
including scattered tube lines (leading to dierence due to enhancement) between the two
wavelengths considered. Because
W
i
K
i
I
i
M
i
62
for the analyte i and
W
s
K
s
I
s
M
s
63
for the internal standard s, the following ratio can be obtained by dividing Equation (62)
by Equation (63):
I
i
I
s
K
is
W
i
64
where
K
is
K
s
M
s
K
i
M
i
W
s
65
Because the same amount of internal standard is added to all specimens, W
s
is essentially a
constant and can be included in the constant K
is
. It should be noted that M
i
(and M
s
) is
not a constant over the concentration range of interest (otherwise linear calibration would
suce) but depends on the matrix elements. However, if both M
i
and M
s
vary in a similar
manner with the matrix elements, the ratio M
i
,M
s
is less sensitive to variation in the
matrix eect and, in practice, can be considered a constant. In practice, the constant K
is
is
determined using linear regression. The main advantage of the internal standard method
over the scattered-radiation method is its ability to correct eectively for enhancement as
well as for absorption. It also correctsat least partiallyfor variations in density of
pressed specimens. The requirement that the intensity of the characteristic radiation of
both the analyte and the internal standard element vary in the same manner with the
Copyright 2002 Marcel Dekker, Inc.
matrix eects imposes that there should be no absorption edges and no characteristic
radiation from other elements between the measured line of the analyte and that of the
internal standard. Furthermore, ideally, the analyte should not enhance the internal
standard or vice versa. If Ka radiation is measured and if the atomic number of the analyte
is Z (with Z23), then, very often, the elements with atomic number Z1 or Z1 are
very good candidates. This assures that there are no K absorption edges and no K emission
lines of other elements between the two elements considered. The element with atomic
number Z is not enhanced by Ka radiation from an element with Z1, but only by the
much weaker Kb radiation, whereas for elements with atomic number Z2 and higher,
both Ka and Kb contribute to enhancement. In practice, some enhancement between the
internal standard element and the analyte or vice versa is allowed, as the concentration of
the internal standard is constant and the ratio is based on intensities. The absence of major
elements in the specimens with L absorption edges and emission lines, however, must be
checked for. The situations that must be avoided are (1) the case where a major line of a
major matrix element is between the absorption edges of the analyte and the internal
standard and (2) the case where a major absorption edge of a major matrix element is
situated between the measured characteristic lines of the analyte and the internal standard.
In the rst case, the matrix element would enhance either the analyte or the internal
standard element, but not both; in the second case, the matrix element absorbs strongly
either the radiation from the analyte or the internal standard, but not both. In both of
these cases, varying concentrations of the matrix element will lead to variable and dierent
eects on the intensities of the analyte and of the internal standard, and the ratio used in
Eq. (64) will not compensate for such events.
The method, however, has some important limitations:
The specimen preparation is made more complicated and is more susceptible to
errors.
The addition of reagents and the requirement of homogeneity of the specimen tends
to limit the practical application of the method to the analysis of liquids and
fused specimens, although it sometimes nds application in the analysis of
pressed powders.
Although the rule Z1 or Z1 can serve as a rule of thumb, it is quite clear that for
samples where many elements are to be quantied, a suitable internal standard
cannot be found for every analyte element. Sometimes, more elements are used
in one internal standard solution to provide suitable internal standards for
more analytes.
Also, the fact that the internal standard method is easier to apply to liquids can
generate some problems. Heavier elements (e.g., Mo) are more dicult to
determine using this method, because liquid specimens are generally not of
innite thickness for the K wavelengths of these heavier elements. In such cases,
the L line can be used, with an appropriate internal standard. The method will,
however, also provide some compensation for the eects of noninnite
thickness, especially if the wavelength of the internal standard selected is very
similar to that of the analyte line.
Theoretically, L lines of a given element can be used as internal standards for K lines
of other elements and vice versa if these wavelengths are reasonably close to each other
and neither interfering lines nor edges occur between them. In principle, the method allows
the determination of one or two elements in a specimen without requiring analysis (or
knowledge) of the complete matrix.
Copyright 2002 Marcel Dekker, Inc.
The range of concentration over which this method is suitable can be quite large, up
to 1020 % in favorable situations, but the internal standard technique is most eective at
low concentrations (or high dilutions). The method nds, for instance, application in the
determination of Ni and V in petroleum products, where MnKa is used as an internal
standard (ISO, 1995). The internal standard method allows an accurate determination of
these elements in a much wider variety of petroleum products than the method based on
linear calibration.
3. Standard Addition Methods
Another method of analysis involves the addition of known quantities of the analyte to the
specimen and is referred to as the standard addition method.
If the analyte element is present at low levels, and no suitable standards are available
(e.g., the matrix is unknown), standard addition and,or dilution may prove to be an al-
ternative, especially if the analyst is interested in only one analyte element. The principle is
the following: Adding a known amount of the analyte i DW
i
to the unknown specimen
will give an increased intensity I
i
DI
i
. Assuming a linear calibration, the following
equations apply:
W
i
K
i
I
i
62
for the original specimen and
W
i
DW
i
K
i
I
i
DI
i
66
for the specimen with the addition. Thus, the method assumes that linear calibration is
adequate throughout the range of addition, because it assumes that an increase in the
concentration of the analyte by an amount DW
i
will increase the intensity by K
i
DI
i
.
Figure 10 Standard addition method. The net intensity is plotted versus weight fraction of the
element added to the sample and a best-fit line is determined. The intercept of that line with the
concentration axis is W
i
.
Copyright 2002 Marcel Dekker, Inc.
These equations can be solved for W
i
. To check the linearity of the calibration, the process
can be repeated by adding dierent amounts of the analyte to the specimen and plotting
the intensity measured versus the concentrations added (Fig. 10). The intercept of the line
on the concentration axis equals W
i
. Note that the concentration in the unknown is
actually found through extrapolation of the linear calibration toward zero intensity. The
intensities used for calibration must be corrected for background and line overlap. If this
correction is not performed (or not performed accurately), the value of the concentration
determined will be overestimated by an amount proportional to the intensity of the
background (or line overlap) and inversely proportional to the sensitivity.
The method is suitable mainly for determination of trace and minor concentration
levels, as the amount DW
i
added to the sample must be in proportion to the amount W
i
in
the sample itself. The extrapolation error can be quite large if the slope of the line is not
known accurately. Adding signicant amounts of additives to the sample might, however,
lead to nonlinearity, as it will alter the matrix eect.
Compounds and solutions can be used for the standard addition. If the analyte i in
the original sample is in a dierent phase than in the additive, care must be taken in the
calculation of the concentration sought. The relevant stoichiometric or gravimetric factors
must be included. This also applies if the analyte is present under elemental or ionic form
in, for example, the original sample and in a compound phase in the additive, or vice versa.
Another way to alter the concentration of the analyte is by diluting the liquid or solid
solution of the sample. By diluting several times by known amounts, a line can be es-
tablished. By repeating this procedure with a standard solution containing a known
amount of i, the unknown concentration can be found.
4. Dilution Methods
Dilution methods can also eliminate or reduce the variation of the matrix eect, rather than
compensating for such variation. The dilution method can be explained using Eq. (17):
C l
0
. l
i
m
i
l
0
P
n
j1
W
j
m
j
l
0
G
P
n
j1
W
j
m
j
l
i
17
which can be rewritten as
C l
0
. l
i
m
i
l
0
m
s
l
0
Gm
s
l
i
67
where m
s
l
0
and m
s
l
i
are the mass-attenuation coecients of the specimen for the pri-
mary wavelength l
0
and analyte wavelength l
i
, respectively. Apparently, deviations from
linearity are due to variations in m
s
l
0
and,or m
s
l
i
. Enhancement is ignored at this
stage. If one adds, to the sample, D grams of a diluent (d ) for each gram of sample, the
denominator or Eq. (67) becomes
1
1 D
m
s
l
0
Gm
s
l
i
D
1 D
m
d
l
0
Gm
d
l
i
68
If the term D,1 Dm
d
l
0
Gm
d
l
i
is much larger than 1,1 Dm
s
l
0
Gm
s
l
i
,
the factor Cl
0
. l
i
becomes essentially a constant and variations due to varying matrix
eects between samples become negligible. This can be done in two ways.
(a) Making D,1 D large by diluting each sample by adding a large, known
amount of a diluent.
Copyright 2002 Marcel Dekker, Inc.
(b) Adding a smaller quantity of diluent than in the previous case, but with a much
larger value for m
d
l
0
Gm
d
l
i
. This is called the technique of the heavy
absorber.
Both of these procedures, however, require the addition of reagents to the sample. This
can easily be done for dissolved samples, either in liquids or fused samples, but it is more
dicult for powdered samples (homogeneity!).
These methods do not eliminate the matrix eects completely, but reduce their
inuence. On the other hand, they also reduce the line intensity of the analyte; thus, a
compromise must be sought.
Dilution methods also have the advantage of reducing the enhancement eect if one
uses a nonuorescing diluent (e.g., H
2
O or Li
2
B
4
O
7
). In this case, the eect is reduced by
the fact that the concentrations of both the enhancing element and the analyte are reduced.
If the diluent contains elements whose characteristic x-rays can excite the analyte, as well
as some other matrix elements, then the contribution of those unknown quantities of
matrix elements to the total enhancement is reduced: The enhancement of the analyte by
the diluent would then be determining and can be considered to be constant.
This method allows the determination of all measurable elements in the sample, as
opposed to the standard addition method, where an addition must be made for each
element of interest.
D. Mathematical Methods
1. General
The term mathematical methods refers to those methods that calculate rather than
eliminate or measure the matrix eect.
Mathematical methods are independent of the specimen preparation in the sense that
specimen preparation is taken into account if the composition of the specimen presented to
the spectrometer has been changed (e.g., by fusion), but mathematical methods do not
prescribe the specimen preparation method as is done, for example, by the standard ad-
dition method. The actual calculation method used to convert intensities to concentrations
does not aect the choice of the specimen preparation method. The aim of the specimen
preparation is limited to the presentation to the spectrometer of a specimen that is
homogeneous (with respect to the XRF technique) and that has a well-dened, at surface
representative for the bulk of the specimen. Mathematical methods usually require
knowledge of all elements in the standard specimens and allow determination of all mea-
surable elements in the unknowns. In practice, trace elements can be neglected in the cal-
culations for the analytes present at higher concentrations, as these trace compounds are
neither subject to an important (and variable) matrix eect nor do they contribute sig-
nicantly to the matrix eect of other elements. Their concentrations are often found by
straightforward linear regression. The mathematical methods are divided in two main ca-
tegories: the fundamental parameter method and the methods using inuence coecients.
2. The Fundamental Parameter Method
a. Introduction
The fundamental parameter method is based on the theory that enables one to calculate the
intensity of uorescent radiation, originating from a specimen of known composition. The
equations used usually consider both primary and secondary ourescence (enhancement).
Copyright 2002 Marcel Dekker, Inc.
Higher-order eects and eects due to scattered radiation are usually neglected. Formulas to
calculate intensities of uorescent radiation were proposed very shortly after the introduc-
tion of the commercial XRF spectrometers, in the early 1950s (Gillam and Heal, 1952). The
equation describing the intensity of the uorescent radiation as a function of specimen
composition, spectrometer conguration, and incident spectrumwas derivedearlier (see Sec.
II.B) and is repeated here:
I
i
Jl
Z
l
edge.i
l
min
P
i
l
X
j
S
ij
l. l
j
" #
dl 69
where P
i
l is the contribution of the primary uorescence caused by incident photons
with wavelength l, and S
ij
l. l
j
is the contribution of the secondary uorescence (en-
hancement) by characteristics photons l
j
which have been excited by primary photons l.
The summation in Eq. (16) or (69) is over all elements j that have characteristic lines that
can excite the analyte i. For each of these elements j, all characteristic lines must be
considered. This is quite simple if none of the L lines or M lines of element j can excite the
analyte. In that case, only the Ka and Kb lines are to be considered. If the L lines of an
element j are energetic enough for enhancement of the analyte, the sheer number of L lines
(e.g., W has more than 20 characteristic L lines that can be considered) would make the
calculation very time-consuming. Therefore, most programs consider only three to ve L
lines for each element. A similar reasoning holds for the M lines.
Application of these formulas was originally limited to the prediction of intensities
for specimens with given composition. The application of those formulas for analysis was
not pursued until the 1960s. The method of analyzing specimens by fundamental para-
meter equations has been developed independently around the same time by Criss and
Birks (1968) and by Shiraiwa and Fujino (1966). Due to the large amount of calculations
involved (especially the integration of the incident spectrum and the calculation of the
contribution of enhancement), these programs initially ran on mainframes and mini
computers. A PC version of the same program was proposed by Criss in 1980 (Criss,
1980a).
The application of a fundamental parameter method for analyzing specimens con-
sists of two steps: calibration and analysis. Both steps will be discussed in more detail in
the following subsections.
b. Calibration
The fundamental parameter equation is used to predict the intensity of characteristic lines
for a composition identical to that of the standard used. If more than one standard spe-
cimen is used, the calculations are repeated for each of the standards. The calculations are
performed using the appropriate geometry (i.e., incidence and take-o angles are taken in
agreement with those of the spectrometer used) and the parameters determining the tube
spectrum (such as the anode material, voltage, thickness of beryllium window, and so on)
correspond to the ones used in the spectrometer for the measurements. The intensities
predicted are (almost always) net intensities, void of background, line overlap, crystal
uorescence, and so on. Hence, the measured intensities must be corrected for such
spectral artifacts. These theoretically predicted intensities are then linked to the actually
measured ones. If only one standard specimen is used, the ratio between the measured
intensity and the calculated intensity is calculated. If more than one standard is used, the
net intensities obtained from the measurements are plotted versus the calculated intensities
and a straight line can be determined for each characteristic line measured. The slope of
Copyright 2002 Marcel Dekker, Inc.
such a line is the proportionality factor between predicted (calculated) and measured in-
tensities. In general, this relationship will be determined more accurately if more standards
are used.
A special case of calibration ensues when one uses pure elements as standards. Di-
viding each of the measured (net) intensities of the corresponding pure element gives the
relative intensity. This relative intensity can be calculated directly by some fundamental
parameter programs. In fact, some programs use equations that express the intensity of
characteristic radiation directly in terms of relative intensity. The relative intensity in this
respect is thus dened as the intensity of the sample, divided by the intensity of the cor-
responding pure element (or compound, if the concentration of the analyte is dened in
compound concentration), under identical conditions for excitation and detection.
Basically, the calibration function, which is determined using the measured in-
tensities of the standards and the calculated intensities, accounts for instrument-related
factors only; matrix eects are accounted for by using the physical theory as described in
the fundamental parameter equation.
The instrument-related parameters for a wavelength-dispersive spectrometer are as
follows:
Collimation
Crystal reectivity
Eciency of detector(s)
Fraction of emergent beam, allowed into the detectors, after Bragg reection (is also
dependent on Bragg angle)
For energy-dispersive spectrometers, the instrument-related parameters are collimation
and detector eciency. The eect of the windows of the detector, the dead layer, and so
forth can also be taken into account.
c. Analysis
Step 1. For every unknown specimen, a rst estimate of the composition is made.
There are several ways to obtain such a rst estimate. They vary from using a simple, xed
composition (e.g., equal to 100%, divided by the number of elements considered; in the
rst estimate, all the concentrations of each of the elements are thus taken equal to one
another) to the composition derived from the measured intensities in combination with the
calibration curves. Using the calibration data, it is possible to estimate for each element
the intensity that would be obtained if the pure elements were measured. These numbers
are then used to divide the intensity of the corresponding element, measured in the un-
known specimen. The resulting fractions are scaled to 100% and used as the rst estimate.
Step 2. For this estimate of composition, the theoretical intensities are calculated.
These are converted to measured intensities, using the calibration data, so that these two
sets of intensities for the same specimen can be compared.
Step 3. The next estimate of composition is obtained based on the dierence between
the measured and calculated intensities. Again, there are dierent methods available:
1. The simplest method is based on linear interpolation. If, for a given element, the
measured intensity is 10% higher than the calculated intensity, the concentration
of that element is increased by 10%.
2. Rather than a linear relationship, some authors (Criss and Birks, 1968) use an
interpolation based on three points. This is done because the relationship
between concentration and intensity is usually non linear over a wider range.
If the specimen is a pseudobinary, hyperbolic relationships have proven to be
Copyright 2002 Marcel Dekker, Inc.
better approximations. For more complex specimens, it still works out quite
well, because the concentrations of the other elements are considered xed at this
stage. The hyperbolic equation requires a minimum of three points for its
parameters to be determined. The points requiring the least additional
calculation time are the following:
The origin (net intensity zero, at concentration zero).
The pure element, W100%; the corresponding intensity has already been
calculated.
The point Wcurrent estimate, the intensity has already been calculated.
These three points allow an hyperbolic relation to be established around the
current estimated composition. From this curve and the measured
intensity, the new concentration estimate is derived. This approach is
repeated for every analyte element. This method usually provides a faster
convergence than using the simple straight line.
3. It is also possible to use gradient methods to determine the next composition
estimate. The formula for the rst derivative, with respect to concentration, of
the fundamental parameter equation have been published (Shiraiwa and Fujino,
1968), but the calculation is cumbersome and time-consuming. Also, the
derivatives can be obtained by a nite-dierence method, where the eect of a
small change in composition on the intensity is observed.
Step 4. The process, starting at Step 2, is now repeated until convergence is obtained.
Dierent convergence criteria exist. The calculation can be terminated if one of the fol-
lowing criteria is satised for all the elements (compounds) concerned:
1. The intensities, calculated in Step 2, do not change from one step to another, by
more than a present level.
2. The intensities, calculated in Step 2 agree, to within a preset level, with the
measured intensities
3. The compositions, calculated in Step 3, do not change from one step to another
by more than a present level (e.g., 0.0005 or 0.0001 by weight fraction).
One or more of these criteria might be incorporated in the program. These criteria,
however, are no guarantee that the nal result is accurate to within the level, specied in
the convergence criteria. Furthermore, especially with convergence criteria based on
concentrations (such as criteria 3), it must be realized that a convergence criterion of
0.0005 is unacceptable when determining elements at levels below 0.0005.
d. Extensions to the Method
More complex scenarios are possible. The most common one includes the calculation
of a set of inuence coecients (based on theoretical calculations) to obtain a com-
position quickly, close to the nal result. Next, the fundamental parameter method is
applied (Rousseau, 1984a). This should yield faster convergence in terms of compu-
tation time, because the calculation by inuence factors of the preliminary composition
is very fast. This method reduces the number of evaluations of the fundamental param-
eter equation.
Also, the dierent programs available dier quite markedly in their treatment of
the intensities of the standards measured. Some programs use a weighting of the stan-
dards, stressing the standard(s) closest (in terms of intensity) to the unknown (Criss,
Copyright 2002 Marcel Dekker, Inc.
1980a). Such programs use a dierent calibration for each unknown specimen. The
unknown specimen dictates which standards will be given a high weighting factor and
which standards will be used with less weighting. Other programs use all standards with
equal weighting.
e. Typical Results
Early results of the fundamental parameter method on stainless steels are given by Criss
and Birks (1968). The average relative dierences between x-ray results and certied values
were about 34%. Later, Criss et al. (1978) reported accuracies of about 1.5% relative for
stainless steels, using a more accurate fundamental parameter program. Typical results for
tool steel alloys are given in Table 3. Often, the fundamental parameter method is con-
sidered to be less accurate than an inuence coecient algorithm. This is caused primarily
by the fact that the fundamental parameter method has extensively been used and
described as a method that allows quantitative analysis with only a few standards. This
is obviously an advantage, but it does not imply that fundamental parameter methods
cannot be used in combination with many standards similar to the unknown. As a matter
of fact, on several occasions and with a variety of matrices, the authors have obtained
results of analysis with an accuracy similar to that of inuence coecient algorithms when
using the same standards in both cases.
f. Factors Aecting Accuracy
The accuracy of the nal results is determined by the following:
The measurement
The specimen preparation
The physical constants used in the fundamental parameter equation
The limited description of the physical processes that are considered in the
fundamental parameter equations
The standard and the calibration
In the following discussion, the eect of measurements and specimen preparation
will not be considered.
Table 3 Analysis of Seven Tool Steels with a Fundamental Parameter Program [XRF11, from
CRISS SOFTWARE, Largo, MD (Criss, 1980)]
Element Minimum conc. (%) Maximum conc. (%) Standard deviation (%)
W 1.8 20.4 0.52
Co 0.0 10.0 0.20
Mn 0.21 0.41 0.01
Cr 2.9 5.0 0.013
Mo 0.2 9.4 0.04
S 0.015 0.029 0.003
P 0.022 0.029 0.003
Si 0.14 0.27 0.03
C 0.65 1.02 0.16
Note: One standard has been used. The minimum and maximum concentrations refer to the minimum and
maximum concentrations in the set of analyzed specimens, respectively. The standard deviation is calculated from
the difference between concentration values found and certied.
Source: Data courtesy of Philips Analytical, Almelo, The Netherlands.
Copyright 2002 Marcel Dekker, Inc.
g. Physical Constants
The physical constants used in the fundamental parameter equations are as follows:
Incidence and exit angles
Spectrum of incident beam
Mass-attenuation coecients
Flourescence yields
Absorption jump ratios
Ratios of intensity of dierent lines within a given series (e.g., Ka,Kb ratio)
Wavelengths (or energies) of absorption edges and emission lines
Incidence and exit angles. The incidence angle in most wavelength-dispersive (WD) and
energy-dispersive (ED) spectrometers is, infact, denedbyarelativelywide cone withadierent
intensity at the boundaries compared to the center. This incident cone is neglected and the
incident radiation is considered parallel, along a single, xed direction. A similar observation
holds for the exit angle. The eect is far less pronounced if diraction from a plane crystal
surface is used for dispersion, as is done, for example, in most sequential WD spectrometers.
This has been studied to some extent by Mu ller (1972). To our knowledge, none of the fun-
damental parameter programs available takes this eect intoaccount. Its inuence, however, is,
to some extent, compensated for by calibration with standard specimens.
Spectrum of incident beam. The spectrum of the incident beam from an x-ray tube
spectrum requires more attention. Parts of the primary spectrum might excite an element B
that, in turn, excites element A very eciently. In such cases, this enhancement may make
the intensity of element A sensitive to small errors in the tube spectrum representation,
which would not be compensated for if the pure A was used for calibration. This can arise,
for example, in the analysis of silicazirconia specimens with a Rh tube (Criss, 1980b).
Pure silica is relatively insensitive to the intensity of the characteristic K lines of Rh. In
combination with Zr, however, the situation is dierent. Indeed, the RhK lines are strongly
absorbed by Zr. Zr then emits K and L lines that enhance Si. As a result, the Si intensity is
more sensitive to the RhK lines in SiO
2
ZrO
2
mixtures than it is in pure SiO
2
. Tube spectra
have been calculated using, for example, the algorithm of Pella et al. (1985).
Mass-attenuation coecients. There are several compilation of mass-attenuation
coecients, published in the literature. A continuing eort to compile the most compre-
hensive table has been undertaken by the National Institute of Standards and Technology
(formerly National Bureau of Standards), Gaitherburg, MD.
When selecting a table of mass-attenuation coecients for use in a fundamental
parameter program, the following question must be addressed: Does the table cover all the
analytical needs? (In practice, does it cover the complete range of interest from the longest
wavelength considered, down to the excitation potential of the tube?)
The analyst should be aware that the use of formulas to generate mass-attenuation
coecients can lead to values that can be signicantly dierent from the corresponding
table values.
Presently, for applications in XRF, the complications of McMaster et al. (1969),
Heinrich (1966), Leroux and Thinh (1977), or Veigele (1974) are most often used. A short
discussion on the agreement between some of these compilations has been presented by
Vrebos and Pella (1988). A more recent compilation has been published by de Boer (1989).
Fluorescence yields. A comprehensive reference to uorescence yields, including
CosterKronig transitions, can be found in the work of Bambynek et al. (1972). (see
also Chapter 1, and Appendix VI).
Copyright 2002 Marcel Dekker, Inc.
Absorption jump ratios. These can be derived from the tables of attenuation coef-
cients.
Ratios of dierent uorescent lines within a family. Data for the K spectra can be
found in the work by Venugopalo Rao et al. (1972) (see also Chapter 1).
Wavelengths of absorption edges and emission lines. A comprehensive table was
published by Bearden (1967) and is also presented in the appendices to Chapter 1. Because
attenuation coecients are wavelength dependent, an error in a wavelength of any char-
acteristic line will automatically lead to a bias in the corresponding attenuation coecients.
h. Limited Physical Processes Considered
The fundamental parameter equation [Eq. (16)] does not consider all physical processes in
the specimens. Three of the most obvious that are missing are described here.
Tertiary uorescence. Although the formula for tertiary uorescence has been de-
rived by, for example, Shiraiwa and Fujino (1966) and Pollai and Ebel (1971), it is not
included in most fundamental parameter programs. Usually, the tertiary uorescence eect
is considered small enough to be negligible. Shiraiwa and Fujino (1967, 1974) have pre-
sented data showing a maximum contribution of tertiary uorescence of about 3% relative
to the total intensity of Cr in FeCrNi specimens. Therefore, even in FeCrNi specimens
whose characteristic lines and absorption edges are ideally positioned relative to one an-
other to favor enhancement, the eect of tertiary uorescence is quite limited. Higher-order
enhancement is also possible, but it is even less pronounced than tertiary uorescence.
Scatter. Other processes not considered in most of the fundamental parameter
methods are coherent and incoherent scatter of both the primary spectrum and the
uorescent lines. This is usually justied by pointing out that the photoelectric eect is, by
far, the major contribution to the total absorption. It is believed that the contribution by
scattered photons to the excitation of characteristic photons is negligible. However, in
some cases the scattered primary spectrum may have a considerable inuence, as illu-
strated earlier in this chapter. The equations describing the contribution of scatter to
uorescent intensity have been derived by Pollai et al. (1971). These equations have ob-
viously many similarities to those for secondary uorescence.
Photoelectrons. The processes that are probably the must unknown in the funda-
mental parameter method are related to the contributions of the photoelectrons and of the
Auger electrons that are produced as a result of absorption of the primary and uorescent
x-ray photons. These electrons have sucient energy to excite other atoms and thus create
additional uorescence. This is especially important in the case of low-atomic-number
elements, as has been described by Mantler (1993) and has been illustrated, for example,
by Kaufmann et al. (1994).
i. Standards and Calibration
The use of good standards (similar to the unknown) will almost always lead to more
accurate results, compared to a situation where the standards used have a widely dierent
composition from that of the unknown. This is because most of the uncertainties, caused
by inaccuracies in the physical constants, cancel. The degree of similarity between stan-
dards and unknown has an important eect on the accuracy of the analysis.
3. Influence Coefficient Algorithms
Another class of mathematical methods calculates the matrix eect by means of coe-
cients, rather than by evaluating the fundamental parameter equation for each unknown.
It will be shown that these coecients can also be calculated from theory, using funda-
Copyright 2002 Marcel Dekker, Inc.
mental parameters. Many of such inuence coecient algorithms have been proposed and
they have been divided and subdivided in dierent ways (Lachance, 1979). It is not the
intention to discuss all of the algorithms here; only a few selected ones will be discussed.
This selection is based on the popularity of the methods and,or on some interesting
characteristics of their underlying theory. Some of these algorithms use only one single
coecient per interfering element; others use more than one. The distinction used here,
however, depends on whether the inuence coecients is considered to be a constant for a
given application or whether the value of the coecient varies with composition. The
latter methods will be discussed in Sec. V.5. Only two algorithms that use constant in-
uence coecients will be discussed here: the LachanceTraill and the de Jongh algo-
rithms. The practical application of the resulting equations (i.e., calibration and analysis)
will be treated separately in Sec. V.7. One has to emphasize that the approach based on
constant inuence coecients has many common aspects with all the inuence coecient
algorithms discussed in Secs. V.4 and V.5.
All of the inuence coecient models express the total matrix eect M
i
for a binary
mixture ij as follows:
M
i
1 m
ij
W
j
70
where m
ij
indicates the true binary inuence coecient describing the matrix eect of j on
the analyte i in binaries ij. More generally,
M
i
1
X
n
j1
j6e
m
ij
W
j
2
6
6
4
3
7
7
5
71
for a multielement specimen, with n being the total number of elements or compounds. In
most of the inuence coecient algorithms, one element is eliminated from the summation
{i.e., one inuence coecient is used when dealing with binaries, as in Eq. (70), and n1
coecients deal with a specimen consisting of n compounds [Eq. (71)]}. In Eq. (71) this is
explicitly indicated by the j 6 e under the summation sign. The eliminated compound e can
be any of the ones present in the specimens; however, most authors eliminate the analyte.
The expression for M
i
is then used as follows:
W
i
R
i
1
X
n
j1
j6e
m
ij
W
j
2
6
6
4
3
7
7
5
72
which links the relative intensity and the inuence coecients to the composition of the
specimen. The relative intensity R
i
for a given analyte is dened as the ratio of the net
measured intensity I
i
in the specimen and the intensity that would have been measured on
the pure analyte I
i
under identical conditions:
R
i
I
i
I
i
73
In practice, the relative intensity is often derived indirectly from measurements on stan-
dards, and the pure element (or compound) is not required. Equation (56) can be rewritten
in terms of R
i
:
W
i
R
i
M
i
74
Copyright 2002 Marcel Dekker, Inc.
By denition,
M
i
W
i
R
i
75
The ratio of the weight fraction of the analyte W
i
and its relative intensity R
i
is the
matrix eect M
i
. When the analyte radiation is absorbed (or when the absorption eects
are dominating over enhancement), M
i
is larger than 1. On the other hand, when en-
hancement is dominant, M
i
is smaller than 1. Also, M
i
is smaller than 1 in the absence
of enhancement, but when the absorption by the analyte is signicantly higher than that
of the matrix elements, as indicated by curve 3 in Figure 6. One of the consequences of
Eq. (74) is that for the pure analyte W
i
1, R
i
1 and, thus, M
i
is also equal to 1. This
implies that the matrix eect as introduced here should be viewed as relative to the pure
element and not in absolute terms. Even in the pure element specimen, x-ray photons are
subject to matrix eect. It is possible, even in the pure analyte, to be subjected to en-
hancement eects. This is, for example, the case when L lines are analyzed if the K lines
of the same analyte are also excited. Furthermore, the L lines of elements with large
atomic numbers can be uoresced by other L lines, as indicated in Table 1. However, it
is customary to refer to the situation as a situation without matrix eect. Also, if
M
i
1, Eq. (74) reduces to
W
i
R
i
76
In other words, in the absence of matrix eect, the concentration of the analyte is equal to
the relative intensity. For a specimen containing, for example, 25% (by weight) of the
analyte, an intensity will then be measured that is 25% of that of the pure analyte.
Comparing Eqs. (56) and (74) and considering the denition of the relative intensity
given by Eq. (73) yields
K
i
1
I
i
in other words, the sensitivity is the reciprocal of the intensity of the pure analyte. So, the
intensity of the pure analyte can also be obtained without making measurements for a
specimen of the pure element. It can be obtained from the slope of the calibration line or
even from the measurement on a single specimen:
I
i
I
i
M
i
W
i
77
where M
i
is calculated and I
i
is measured. M
i
can be calculated using Eq. (75), where R
i
is
calculated from theory for the standard specimen of known composition. Therefore, al-
though many inuence coecient algorithms are presented using the format of Eq. (72),
involving the relative intensity, there is no real need to perform measurements of the pure
analyte, as Eq. (72) can be written as
W
i
K
i
I
i
1
X
n
j1
j6e
m
ij
W
ij
2
6
6
4
3
7
7
5
78
where K
i
is then determined during the calibration phase. Furthermore, if the background
is not subtracted, Eq. (78) can be written as
Copyright 2002 Marcel Dekker, Inc.
W
i
B
i
K
i
I
i
1
X
n
j1
j6e
m
ij
W
ij
2
6
6
4
3
7
7
5
79
where B
i
is the background expressed as a concentration equivalent. The constants K
i
and
B
i
are then determined during regression analysis.
4. Algorithmswith Constant Coefficients
a. The LachanceTraill Algorithm
Formulation. In 1966, Lachance and Traill proposed a correction algorithm based
on inuence coecients (Lachance and Traill, 1966). The equations are, for a ternary
consisting of the elements (or compounds) A, B, and C:
W
A
R
A
1 a
AB
W
B
a
AC
W
C
80a
W
B
R
B
1 a
BA
W
A
a
BC
W
C
80b
W
C
R
C
1 a
CA
W
A
a
CB
W
B
80c
R
A
, R
B
, R
C
are the relative intensities of A, B, and C, respectively. The coecients a
AB
,
a
AC
, a
CA
, and so forth are called inuence coecients. A more general notation of the
LachanceTraill algorithm is, for analyte i,
W
i
R
i
1
X
n
j1
j6i
a
ij
W
j
2
6
6
4
3
7
7
5
81
where the summation covers all n elements (or compounds) in the specimen, except the
analyte itself. Hence, there are n1 terms in the summation. This is common to all cur-
rently used algorithms. Most of the algorithms, developed earlier [such as Shermans
(1953) and Beattie and Brisseys (1954)], used to have n terms, rather than n1, for spe-
cimens with n elements.
Equations (80a), (80b), and (80c) are linear equations in the concentrations of the
elements W
A
, W
B
, and W
C
, respectively. Note that there are only two coecients for each
analyte element. Consider, for example, the rst equation of the set 80, namely Eq. (80a):
Element A is the analyte and its concentration is equal to the relative intensity R
A
, mul-
tiplied by the matrix correction factor (1 a
AB
W
B
a
AC
W
C
). This matrix correction
factor has only two coecients: one (a
AB
) to describe the eect of element B on the in-
tensity of A and, similarly, one to describe the eect of element C on the intensity of A.
The value of the coecient a
AA
, which would correct for the eect of A on its own in-
tensity (sometimesbut incorrectlyreferred to as self-absorption) is zero. Similarly, a
BB
and a
CC
are also zero. The eect of A on A, however, is taken into account, as will be
shown in the next subsection.
Calculation of the coecients. Lachance and Traill also showed that the inuence
coecients, c
i,
can be calculated for monochromatic excitation by photons with wave-
length `
0
(assuming absorption only) from the expression
a
ij
m
j
l
0
cscc
0
m
j
l
i
cscc
00
m
i
l
0
cscc
0
m
i
l
i
cscc
00
1 82
Copyright 2002 Marcel Dekker, Inc.
When secondary uorescence (enhancement) is involved, the coecients are calculated in
the same way. Thus, enhancement is being treated as negative absorption. This assump-
tion is not valid when enhancement is quite severe. Dierences in primary absorption may
easily be confused with enhancement (see Fig. 6). From Eq. (82), it follows clearly that a
ii
is always zero. Therefore, it is not included in the summation of Eq. (81). Also, from
Eq. (82), it follows that the value of the coecients for the LachanceTraill algorithm
cannot be less than 1. It must be stressed that Eq. (82) is only valid strictly for mono-
chromatic excitation and for those analytes that are only subject to absorption (no en-
hancement). In this case, the inuence coecient is concentration independent: It is a
constant, even for the complete concentration range from 0% to 100%. It does, however,
depend on parameters, such as the wavelength of the primary photons and the incidence
and exit angles. In all other cases (polychromatic excitation and,or enhancement),
Eq. (82), strictu sensu, cannot be used. A polychromatic beam (from, e.g., an x-ray tube)
can be replaced by a monochromatic one, by resorting to the eective wavelength. The
eective wavelength, however, is composition dependent (see Sec. II.C). The value of the
coecients, calculated using Eq. (82), is also dependent on composition, although W
i
nor
W
j
gure explicitly in Eq. (82). If enhancement is dominant, another method must be
applied to calculate the coecients.
Inuence coecients for the algorithm of, for example, LachanceTraill can also be
calculated based on actual measurements. Rewriting
W
i
R
i
1 a
ij
W
j
83
to
a
ij
W
i
,R
i
1
W
j
84
yields an expression that can be used to obtain a
ij
, based on the composition of the
binary and the relative intensity R
i
. The drawbacks associated with this method are as
follows:
1. The calculation of R
i
requires the measurements of the intensity on the pure i
(element or compound). This could lead to large errors if the intensity of i in the
binary is much lower than that of the pure, due to, for example, nonlinearity of
the detectors.
2. The pure elements (or compounds) are not always easily available or could be
unsuitable to present to the spectrometer (e.g., pure Na or Tl).
3. Equation (84) is very prone to error propagation when W
i
is close to 1. The
numerator is then a dierence between two quantities of similar magnitude, and
the denominator is then close to zero, magnifying the errors.
4. Also, the availability of suitable binary specimens can present problems: Some
alloys tend to segregate and homogeneous specimens are then dicult to
obtain.
The coecients a
ij
can also be calculated from theory: Calculate R
i
for the binary
with composition W
i
. W
j
rather than obtain it from measurements and substitute in Eq.
(84). This method eliminates drawbacks 1, 2, and 4. However a better methodwithout
the problem associated to error propagationis to use Eq. (75) directly with the values for
R
i
and W
i
. Lachance (1988) has also presented methods to calculate the values of the
coecients from theory.
Copyright 2002 Marcel Dekker, Inc.
The LachanceTraill algorithm assumes the following:
1. The inuence coecients can be treated as constants, independent of
concentration; this limits the concentration range in cases where the matrix
eects change considerably with composition.
2. The inuence coecients are invariant to the presence and nature of other
matrix elements. So a
FeCr
, determined for use in FeCrNi ternary specimens, is
the same as a
FeCr
in FeCrMoWTa or FeCr specimens.
b. The de Jongh Algorithm
Formulation. In 1973, de Jongh proposed an inuence coecient algorithm (de
Jongh, 1973), based on fundamental parameter calculations. The general formulation of
his equation is
W
i
E
i
R
i
"
1
X
n
j1
j6e
a
ij
W
j
#
85
where E
i
is a proportionality constant (and is usually determined during the calibration).
The summation covers n1 elements (as is the case with the algorithm of Lachance and
Traill), but the eliminated element, e, is the same for all equations. If, for a ternary spe-
cimen, element C is eliminated, the following equations are obtained:
W
A
E
A
R
A
1 a
AA
W
A
a
AB
W
B
86a
W
B
E
B
R
B
1 a
BA
W
A
a
BB
W
B
86b
W
C
E
C
R
C
1 a
CA
W
A
a
CB
W
B
86c
Note that in order to obtain the concentration of elements A and B, the concentration of
C, W
C
, is not required. This is dierent from Lachance and Traills algorithm: In order to
calculate the concentration of A, using Eq. (80a), the concentrations of both B and C are
required. If the user is not really interested in element C (e.g., element C is iron in stainless
steels), Eq. (86c) need not be considered and the analysis of the ternary specimen can be
done by measuring R
A
and R
B
and solving Eqs. (86a) and (86b).
Calculation of the coecients. de Jongh also presented a method to calculate the
coecients from theory. The basis is an approximation of \
i
,1
i
by a Taylor series
around an average composition:
W
i
R
i
E
i
d
i1
DW
1
d
i2
DW
2
d
in
DW
n
87
where E
i
is a constant given by
E
i
W
i
R
i
average
88
and
DW
i
W
i
W
i.average
89
d
ij
are the partial derivatives of W
i
,R
i
with respect to concentration:
d
ij
0W
i
,R
i
0W
j
90
Copyright 2002 Marcel Dekker, Inc.
In practice, these derivatives are calculated as nite dierences. W
i
,R
i
is calculated
for a specimen with the average composition W
1.average
,W
2.average
, . . . ,W
n.average
(symbol
W
i
,R
i
average
. Then, the concentration of each element j in turn is increased by a small
amount [e.g., 0.1% (0.001 in weight fraction)] and W
i
,R
i
is calculated for that composi-
tion (symbol [W
i
,R
i
W
j
0.001
). Substituting in Eq. (90) yields
d
ij
0W
i
,R
i
0W
j
W
i
,R
i
W
j
0.001
W
i
,R
i
average
0.001
91
This process is repeated for each of the elements j to calculate all the coecients for
analyte i. This is also repeated for the other analyte elements. The coecients, calculated
from Eq. (91), can be used in Eq. (87) for analysis. Equation (87), however, has n factors
rather than n1 and uses DW
j
rather than W
j
. Using the fact that
X
n
j1
DW
j
0 92
or
DW
e
DW
1
DW
2
DW
n
93
one element e can be eliminated. The resulting equation is similar to Eq. (87), but has only
n1 terms:
W
i
R
i
E
i
b
i1
DW
1
b
i2
DW
2
b
in
DW
n
94
with
b
i1
d
i1
d
ie
95
for all b
ij
except b
ie
, which is equal to zero. Equation (94) has n1 terms, but they are still
in DW, rather than W. Transformation of DW to W is done by substituting Eq. (89) in
Eq. (94):
W
i
R
i
E
i
X
n
j6e
b
ij
W
j.average
X
n
j6e
b
ij
W
j
96
which can be rearranged to Eq. (85) with
a
ij
b
ij
E
i
P
n
j6e
b
ij
W
j.average
97
Combining Eq. (85) with Eqs. (75) and (88) yields
W
i
M
i.average
R
i
1
X
n
j1
j6e
a
ij
W
j
2
6
6
4
3
7
7
5
98
indicating that the weight fraction of the analyte is calculated from its relative intensity, a
matrix correction term, and the matrix correction term for the average composition (which
for a given composition is a constant). The value for M
i,average
can be calculated using the
inuence coecients calculated and taking the composition (W
i
, W
j
) equal to the average
Copyright 2002 Marcel Dekker, Inc.
composition. This equation is very similar at rst sight to the LachanceTraill equation
[Eq. (81)], except for the term M
i,average
.
Typical results. Tables with de Jonghs coecients have been used for a wide
variety of materials, including high-temperature alloys, brass, solders, cements, glasses,
and so forth. An example for stainless steels is given in Table 4. Results are shown for
Mn, Ni, and Cr only. For these analytes, the matrix eects are the most important.
Other elements, such as Si, P, S, and C, are also present at trace level. The coecients are
calculated at a given composition [see Eq. (91)]. The practical range of concentration over
which these coecients yield accurate results varies from 5% to 15% in alloys to the whole
range from 0% to 100% in fused oxide specimens.
Comparison between the algorithms of de Jongh and LachanceTraill. The following
points can be noted:
1. The basis of the de Jongh algorithm is a Taylor series expansion, around an
average (or reference) composition. The values of the coecients calculated
depend on this composition.
2. De Jongh can eliminate any element; Lachance and Traill eliminate the analyte
itself: a
ij
is zero. De Jongh eliminates (i.e., xes the coecient to zero) the same
element for all analytes. Eliminating the base material (e.g., iron in steels) or the
loss on ignition (for beads) generally leads to smaller numerical values for the
coecients and avoids the necessity to determine all elements.
3. De Jonghs coecients are calculated at a given reference composition. They are
composition dependent and take into account all elements present. A coecient
a
ij
represents the eect of element j on the element i in the presence of all other
elements: They are multielement coecients rather than binary coecients. This
is seen in Table 5, where the values of the coecients a
CrCr
and a
CrNi
are shown
for several dierent specimens, but with identical concentrations for the analyte
Table4 Analysis of Stainless Steels with Theoretical Inuence Coefcients (de Jongh)
Element Min.
a
(%) Max.
a
(%) Std. dev.
b
(%)
Mn 0.64 1.47 0.015
Cr 12.40 25.83 0.06
Ni 6.16 20.70 0.06
a
Min., Max.: minimum and maximum concentration in the set of analyzed specimens, respectively.
b
Std. dev.: standard deviation, calculated from the difference between concentration values found
and certied.
Source: Data courtesy of Philips Analytical, Almelo, The Netherlands.
Table 5 Values for Inuence Coefcients a
CrCr
and a
CrNi
, Calculated According to the Algorithm
of de Jongh, for Selected Specimens
W
Cr
W
Fe
W
Ni
W
Ti
W
W
a
CrCr
a
CrNi
0.18 Bal. 0.08 0 0 2.06 0.750
0.18 Bal. 0.08 0 0.08 2.30 0.817
0.18 Bal. 0.08 0.08 0 2.49 0.868
Note the variation of the values of the coefcients, depending on the presence of Ti or W.
Copyright 2002 Marcel Dekker, Inc.
and the interferent. In all cases, Fe has been eliminated and W and Ti are either
present at 0.08 or not.
4. The calculation of the coecients is based on theory, treating both absorption
and enhancement eects. Hence, the coecients are susceptible to the errors,
described earlier (Sec. V.D.2). However, the calculation of the coecients
involves a division of the matrix correction terms for the slightly aected
composition by the corresponding term of the reference composition [Eq. (97)].
This compensates to a large degree for some of the biases introduced by the
fundamental parameters.
5. For a specimen, containing n elements, there are n equations (one for each of the
elements) if one uses the LachanceTraill algorithm. Each of these equations has
n1 inuence coecients; the coecient of the analyte in each of the equations
has been set to 0. For the same specimen, de Jongh only requires n1 equations,
using n1 coecients per equation. One element has been eliminated
throughout. This element (or compound) is usually one that is of no or little
interest to the analyst (e.g., iron in steels or the loss on ignition in fused beads)
(de Jongh, 1979). The nth equation (for the eliminated element) can also be
written: Its form is identical to the others, it has also n1 terms, and the
coecients can be calculated following exactly the same procedure as for the
other coecients.
6. The coecients in the LachanceTraill equation have been calculated
empirically, using measured data from many more standards than used in the
de Jonghs algorithm, which always used theoretical coecients. These
coecients could be obtained from Philips, eliminating the need for a large
computer at each users site. With increased computer capabilities at the
disposition of every analyst, the calculation of the coecients from theory has
now been feasible for some years. It is possible to calculate the coecients for
the LachanceTraill algorithm from theory as well (Lachance, 1988).
5. AlgorithmswithVariable Coefficients
a. Introduction
Both algorithms discussed so far use a single, constant coecient for each (except one)
interfering element: The expression from Lachance and Traill uses coecients, expressing
the eect of one element on the characteristic intensity of the analyte, relative to the
analyte, ignoring all other elements. Such coecients are therefore referred to as binary
coecients. The algorithm of de Jongh calculates multielement inuence coecients ef-
fectively. Such coecients predict the eect of one element on the intensity of another in a
given matrix.
This distinction between binary and multielement coecients can clearly be seen in,
for example, a ternary specimen. Algorithms based on binary coecients add interelement
eects from each of the constituent elements. They calculate the matrix eect using in-
uence coecients that have been calculated for binaries. Assume a NiFeCr specimen.
The total matrix eect of Cr is accounted for using a coecient expressing the inuence of
Ni on Cr (a
CrNi
) and a similar coecient for Fe on Cr (a
CrFe
). Both of these coecients are
calculated for the corresponding binaries (NiCr and FeCr, respectively). In a ternary
specimen (e.g., NiFeCr), however, the eect of, for example, Ni on Cr is aected by the
presence of Fe (and Ni similarly aects the eect of Fe on Cr). This eect is called the
crossed eect and will be discussed in a following section.
Copyright 2002 Marcel Dekker, Inc.
The original expression presented by Lachance and Traill (1966) to calculate the
inuence coecients required the use of monochromatic excitation. An equivalent
wavelength was used to calculate the coecients when the polychromatic excitation is
applied. The equivalent wavelength, however, has been shown to vary with composition,
as treated earlier in this chapter. The theoretical coecients, calculated according to
de Jongh, are also composition dependent, as the reference composition is used explicitly
in the calculations.
This variation is due to the fact that the composition of the matrix varies con-
siderably if analysis is required over a wider range of concentrations. This has been re-
cognized early in the development of inuence coecient algorithms, and many dierent
algorithms with variable coecients have been proposed. A variable inuence coecient
in this respect is an inuence coecient that varies explicitly with concentration of one or
more components in the specimen. Some of these will be discussed in the subsequent
subsections.
b. The ClaisseQuintin Algorithm
Formulation. Claisse and Quintin (1967) extended Lachance and Traills algorithm
by considering a polychromatic primary beam. The resulting equation for W
A
can be
expressed as
W
A
R
A
1
X
n
j6A
a
Aj
W
j
X
n
j6A
a
Ajj
W
2
j
X
n
j6A
X
n
k6A.kj
a
A
j
k
W
j
W
k
" #
99
where the summation over j has n1 terms (all n elements, except i), and the summation
over k has (n2),2 terms (all n elements, except the analyte i and element j; furthermore, if
a
A
j
k
is used, then a
Akj
is not). For a binary specimen, Eq. (99) reduces to
W
A
R
A
1 a
0
AB
W
B
100
with
a
0
AB
a
AB
a
ABB
W
B
101
clearly showing that the inuence coecient a
0
AB
varies linearly with composition (i.e.,
W
B
). For binaries, W
A
1 W
B
; hence, Eq. (101) can also be rearranged to
a
0
AB
a
AB
a
ABA
W
A
102
Equations (101) and (102) are, at least theoretically, identical. It has been shown, however,
that Eq. (102) is preferable to Eq. (101) if specimens with more than two elements (or
compounds) are analyzed (Lachance and Claisse, 1980). This will be discussed in more
detail in Sec. V.D.6. Note that the value of a
AB
in Eq. (101) is dierent from its value in
Eq. (102).
Cross-product coecients. For a ternary specimen, the ClaisseQuintin algorithm
can be written
W
A
R
A
1 a
AB
W
B
a
ABB
W
2
B
a
AC
W
C
a
ACC
W
2
C
a
ABC
W
B
W
C
103
The terms
a
AB
W
B
a
ABB
W
2
B
and
a
AC
W
C
a
ACC
W
2
C
Copyright 2002 Marcel Dekker, Inc.
are the matrix corrections, due to B and C, respectively. The term a
ABC
W
B
W
C
corrects for
the simultaneous presence of both B and C and is referred to as a cross-product coecient.
Calculation of the coecients. Claisse and Quintin (1967) also published methods to
calculate the coecients frommeasurements on binary and ternary mixtures or fromtheory.
These methods, however, are now generally superseded by theoretical calculations, such as
discussed in Sec. V.C.4.c and V.C.4.d. Rousseau (1984b) has presented a calculation method
for the coecients in the ClaisseQuintin algorithm, and Wadleigh (1987) has commented
upon this approach.
c. The RasberryHeinrich Algorithm
Following a systematic study of the FeCrNi ternary system, Rasberry and Heinrich
(1974) concluded that the two phenomenaabsorption and enhancementare to be de-
scribed by two dierent equations. They introduced the following algorithm:
W
i
R
i
1
X
n
j6i
A
ij
W
j
X
n
k6i
B
ik
1 W
i
W
k
" #
104
where only one coecient is used for each interfering element. The coecients A
ij
are used
for cases where absorption is the dominant eect. In this case, the coecient B
ik
is taken
equal to zero. If, for a given analyte, all B
ik
coecients are zero, Eq. (104) reduces to the
LachanceTraill expression. When enhancement by element k dominates, a B
ik
coecient
is used. The corresponding A
ij
coecient is then taken equal to zero. Hence, the total
number of terms in both summations is n 1.
The correction factor for enhancement by element k can be rewritten as
a
ik
B
ik
1 W
i
105
showing that a
ik
varies with concentration in a nonlinear fashion. The algorithm is very
popular when analyzing stainless steels and steels in general.
Among the disadvantages of the RasberryHeinrich algorithm are the following:
1. It is not always clear which interfering elements should be assigned a B
coecient and which one an A. In PbSn alloys, the SnLa line is uoresced
(enhanced) by both SnK and PbL lines. Yet, the calibration curve for SnLa
clearly shows that absorption is dominant (Fig. 11).
2. Furthermore, Eq. (105) suggests that the value of B
ik
at W
i
0 is twice the value
at the other end of the calibration range when W
i
1. This is not generally
valid. Mainardi and co-workers (1982) have therefore suggested replacing the 1
in the denominator by an additional coecient.
3. Rasberry and Heinrich did not publish a method for calculating the coecients
from theory.
Some of the disadvantages of calculating empirical coecients have been discussed
in Sec. V.D.4.a. For these reasons, the RasberryHeinrich algorithm is not generally
applicable. However, the concept of a hyperbolically varying inuence coecient has been
incorporated in the three-coecient algorithm of Lachance.
d. The Three-Coecient Algorithm of Lachance
Formulation. In 1981, Lachance (1981) proposed a new approximation to the
binary inuence coecient c
1
i,
given by
Copyright 2002 Marcel Dekker, Inc.
m
ij
a
B
ij
a
ij1
a
ij2
W
m
1 a
ij3
1 W
m
106
with
W
m
1 W
i
107
W
m
is the concentration of all matrix elements. It has been shown by Lachance and Claisse
(1980), as well as by Tertian (1976), that variable binary coecients must be expressed in
terms of W
m
(or 1W
i
). For binary specimens, Eq. (106) can be rewritten using W
j
for W
m
and W
i
for (1W
m
):
m
ij
a
B
ij
a
ij1
a
ij2
W
j
1 a
ij3
W
i
108
For specimens with more than two compounds, however, the dierence between Eqs. (106)
and (108) becomes clear. The value for the inuence coecient m
ij
is approximated over
the complete concentration range for the binary by the function in Eq. (106), which relies
on three coecients only. The excellent agreement between the true inuence coecient
m
ij
and the approximation of Eq. (106) is shown in Figure 12 for Fe in FeNi (severe
enhancement) and for Fe in FeCr (pronounced absorption).
For multielement specimens, cross-product coecients a
ijk
are used to correct for the
crossed eect, similar to Eq. (99). The general equation for a multielement specimen is
W
i
R
i
1
X
n
j6i
a
ij1
a
ij2
W
m
1 a
ij3
1 W
m
W
j
X
n
j6i
X
n
k6i.kj
a
ijk
W
j
W
k
" #
109
Figure11 Calibration curves for PbLa (solid line), SnKa (dashed line), and SnLa (dotted line) in
PbSn binaries. The SnLa is apparently dominated by absorption, rather than by enhancement of
SnK and PbL lines.
Copyright 2002 Marcel Dekker, Inc.
where the summation over j has n 1 terms (all n elements, except i) and the summation
over k has (n 2),2 terms (all n elements, except the analyte i and element j; furthermore if
a
ijk
is used, then a
ikj
is not).
Vrebos and Helsen (1986) have published some data on this algorithm, clearly
showing the accuracy of the algorithm, using theoretically calculated intensities. The use of
Figure 12 The binary influence coefficient m
FeNi
in FeNi binary systems (j, enhancement, top)
and m
FeCr
in FeCr (j, absorption, bottom) and the approximation by the hyperbolic three-
coefficient algorithm for Lachance (COLA). Note the excellent agreement in both cases. Conditions:
W tube at 45 kV, in a spectrometer with an incidence angle of 63
and 33
take-off angle.
Copyright 2002 Marcel Dekker, Inc.
theoretically calculated intensities has the advantage that it avoids errors due to specimen
preparation and measurement errors associated with actual measured data. Pella and co-
workers (1986) have presented a comparison of the algorithm with several others and with
a fundamental parameter method using experimental data.
Calculation of the coecients. The coecients a
ij1
, a
ij2
, and a
ij3
are calculated using
fundamental parameters at three binaries i j. The cross-product coecients are calculated
from a ternary. The compositions of the specimens concerned are listed in Table 6. The
specimens referred to in Table 6 are hypothetical specimens. The intensities are calcu-
lated from fundamental parameters and require no actual measurements on real specimens.
Step 1. Calculate the relative intensity R
i
for the rst composition in Table 6. If the
analysis of interest has more than three elements, then the system is divided in combi-
nations of three elements i, j, k at a time. The analyte is element i, and j and k are two
interfering elements. If the system considered is with compound phases, such as oxides,
then the compositions in Table 6 are assumed to be for the oxides.
Step 2. Using Eq. (84), the corresponding inuence coecient a
B
ij
can be calculated.
Step 3. For this composition, W
m
1 W
i
W
j
0.001, which is small enough to
be considered zero. Hence, Eq. (106) reduces to
a
B
ij
a
ij1
110
a
B
ij
has been calculated in Step 2, so a
ij1
can be computed.
Step 4. Calculate the intensity for the second composition of Table 6 and use Eq. (84)
to calculate a
B
ij
. In most cases, this value will be dierent from the one found in Step 2
because the compositions involved are dierent.
Step 5. 1 W
m
W
i
0.001 is small enough to be considered zero; hence, Eq. (106)
reduces to
a
B
ij
a
ij1
a
ij2
111
a
ij1
and a
B
ij
are known so a
ij2
can be calculated.
Step 6. Calculate the intensity for the third composition of Table 6 and use Eq. (84)
to calculate a
B
ij
. In most cases, this value will be dierent from the one found in Step 2 or 4
because the compositions involved are dierent.
Step 7. Using W
m
1 W
i
0.5 W
i
, Eq. (106) reduces to
a
B
ij
a
ij1
a
ij2
0.5
1 a
ij3
0.5
112
Table 6 Composition of the Specimens Used for the Calculations of the
Coefcients for Lachances Three-Coefcient Algorithm, in Weight Fraction
Specimen No. W
i
W
j
W
k
1 0.999 0.001 0.0
2 0.001 0.999 0.0
3 0.5 0.5 0.0
4 0.999 0.0 0.001
5 0.001 0.0 0.999
6 0.5 0.0 0.5
7 0.30 0.35 0.35
Copyright 2002 Marcel Dekker, Inc.
which can be rearranged to
a
ij3
a
ij2
a
B
ij
a
ij1
2 113
all coecients on the right-hand side are known, so a
ij3
can be calculated.
Step 8. Repeat Steps 17 for W
i
and W
k
, to compute the coecients a
ik1
, a
ik2
,
and a
ik3
.
Step 9. Calculate the intensity R
i
for the ternary (composition 7 in Table 6). Cal-
culate a
B
ij
and a
B
ik
, using Eq. (106) and the coecients determined earlier.
Step 10. Eq. (109) combined with Eq. (106) for a ternary specimen ijk reduces to
W
i
R
i
1 a
B
ij
W
j
a
B
ik
W
k
a
ijk
W
j
W
k
114
which can be rearranged to solve for a
ijk
:
a
ijk
W
i
,R
i
1 a
B
ij
W
j
a
B
ik
W
k
W
j
W
k
115
All variables on the right-hand side of Equation (115) are known, so a
ijk
can be calculated.
Step 11. Repeat for other interfering elements ( j and k) and repeat for other analytes i.
Tao et al. (1985) published a complete computer program illustrating the method
and allowing the calculation of the coecients and analysis of unknowns. This program
suers from an oversimplication in that only the measured analytical lines are considered
for enhancement. This would generate erroneous values for the coecients in cases such as
CuZn alloys, where the ZnKa line cannot uoresce the K shell of Cu, but the ZnKb can
do so. If the ZnKa line is used for analysis, the eect of the ZnKb line (enhancement of Cu)
is not taken into account by the program. In practice, however, the only lines that are
considered for enhancement are the characteristic lines used for the analysis of the other
elements. This can be seen, for example, by calculating the coecients twice for FeSi with
identical conditions; once indicating FeKa and SiKa lines are to be used and once in-
dicating that the FeLa be used (Table 7). The value of the coecients for Fe will change;
this is quite obvious because the magnitude and the sort of the matrix eects on the FeKa
and the FeLa characteristic lines are quite dierent. The value for Si, however, should not
change: in both cases, the same elements are present, and using the same excitation
conditions, there is no reason why the coecients should be dierent as the matrix eects
are the same.
e. The Algorithm of Rousseau
Formulation. Rousseau and Claisse (1974) used a linear relationship to approx-
imate the binary coecients and cross-product coecients:
W
i
R
i
1
X
n
j6i
a
ij1
a
ij2
W
m
W
j
X
n
j6i
X
n
k6i.kj
a
ijk
W
j
W
k
" #
116
The binary inuence coecients are thus approximated by
m
ij
a
B
ij
a
ij1
a
ij2
W
m
117
This model can be used as a stand-alone inuence coecient algorithm, but it has also
been proposed as the starting point for a fundamental parameter algorithm (Rousseau,
1984a). The degree of agreement between the inuence coecient m
ij
and the approxi-
Copyright 2002 Marcel Dekker, Inc.
mation is shown in Figure 13 for the FeNi and the FeCr binaries. The agreement for the
straight line is obviously not as good as with the COLA algorithm, especially in those cases
where the value of the true inuence coecient varies markedly, as is the case for Fe in
FeCr (absorption). Equation (117) has been compared to the three-coecient algorithm of
Lachance by Vrebos and Helsen (1986). They show that the accuracy is somewhat less
than for Lachances method, but for most practical purposes, the Rousseau algorithm
should give acceptable results.
Calculation of the coecients. Rousseau has shown that the fundamental para-
meter equation can be rearranged to
W
i
R
i
1
X
n
j61
a
ij
W
j
" #
118
and he also proposed a method to calculate the a coecients directly from fundamental
parameters, without calculating the intensity rst (Rousseau, 1984a). As a matter of fact,
Rousseau rst calculates the coecients for a given composition and then calculates the
intensity, using Eq. (118). The coecients in Eq. (116) are calculated in a way very similar
to the method described in Sec. V.D.5.d. The compositions involved are given in Table 8.
The specimens referred to in Table 8 are hypothetical specimens. The intensity is cal-
culated from fundamental parameters and requires no actual measurements on real spe-
cimens. For the rst two binaries of Table 8, the inuence coecient is calculated [symbol
a
ij
(0.20,0.80) and a
ij
(0.80,0.20), respectively]. Then the corresponding values are sub-
stituted in Eq. (117):
a
ij
0.20. 0.80 a
ij1
a
ij2
0.80 119a
a
ij
0.80. 0.20 a
ij1
a
ij2
0.20 119b
These equations can be solved for a
ij1
and a
ij2
. Similarly, using compositions 3 and 4 from
Table 8, the corresponding coecients for ik can be calculated. The cross-product
coecients a
ijk
are calculated using Eq. (115).
6. Specimenswith MorethanTwo Compounds
The methods described by Lachance [Eq. (106)] and Rousseau [Eq. (117)] explicitly de-
scribe algorithms to calculate the value of the binary inuence coecient by a rather
simple, hyperbolic or linear, relationship. Combining these binary coecients to describe
the matrix eect for specimens with more than two elements (or compounds) is described
in this subsection. The ternary system FeNiCr is taken here as an example. Figure 14 gives
Table7 Values for the Coefcients for Eq. (109) for Si in FeSi Binaries
SiKa(FeKa) SiKa(FeLa)
a
ij1
5.396 6.284
a
ij2
1.890 0.015
a
ij3
0.409 0.846
Note: In the second column, Fe is measured using the Ka line and in the last
column, the La line is used. Conditions: W tube at 45 kV, in a spectrometer with an
incidence angle of 63
and a 33
take-off angle.
Copyright 2002 Marcel Dekker, Inc.
the relative intensity of FeKa as a function of the weight fraction of Fe in FeNiCr spe-
cimens. There is considerable spread of the intensity of FeKa, even for a constant weight
fraction of Fe. For specimens with a weight fraction of 0.10 Fe, the relative intensity of
FeKa varies between 0.036 and 0.16 (points marked 1 and 2 in Fig. 14). This is due to the
Figure 13 The binary influence coefficient m
FeNi
in FeNi binary systems (j, enhancement, top)
and m
FeCr
in FeCr (j, absorption, bottom) and the approximation by the straight line as suggested
by Rousseau and Claisse. Note the rather large deviations, especially at the high concentration
ranges in FeCr. Conditions: W tube 45 kV, in a spectrometer with an incidence angle of 63
and 33
take-off angle.
Copyright 2002 Marcel Dekker, Inc.
rather dierent eect that Ni and Cr have on Fe:Cr is an absorber for FeKa radiation,
whereas the NiK radiation can enhance FeK radiation through the process of secondary
uorescence (enhancement). For these specimens, the matrix eect M
Fe
can be calculated
from Eq. (75). The total matrix eect on Fe, M
Fe
(FeNiCr), in these specimens, at a xed
Fe concentration of 0.1, for example, varies from 0.63 (for 0.1 Fe in FeNi, point 2) to 2.8
(for 0.1 Fe in FeCr, point 1).
Now, the problem is how to calculate the matrix eect in this case, based on in-
uence coecients. Assume a specimen with the following composition: W
Fe
0.1,
Table 8 Composition of the Specimens Used for the Calculations of the
Coefcients for the Linear Approximation According to Rousseaus Algorithm,
in Weight Fraction
Specimen No. W
i
W
j
W
k
1 0.20 0.80 0.0
2 0.80 0.20 0.0
3 0.20 0.0 0.80
4 0.80 0.0 0.20
5 0.30 0.35 0.35
Figure14 The relative intensity of FeKa as a function of the concentration of Fe, in the presence
of Ni and Cr. For every given weight fraction of Fe, the highest value of the intensity is obtained for
the binary system FeNi (enhancement), whereas the lowest intensity is for the binary FeCr
(absorption). The intermediate values are for ternary specimens, where the concentrations of Ni and
Cr vary in steps of 0.1 weight fraction. At a weight fraction of Fe 0.7, the four data points labeled
a, b, c, and d represent the following specimens (W
Fe
, W
Ni
, W
Cr
): a (0.7, 0.0, 0.3), b(0.7, 0.1,
0.2), c (0.7, 0.2, 0.1), and d (0.7, 0.3, 0.0). Points labeled 1 and 2: see text. Experimental
conditions: W tube at 45 kV, 1-mm Be window, incidence and take-off angles 63
and 33
,
respectively.
Copyright 2002 Marcel Dekker, Inc.
W
Cr
0.3, and W
Ni
0.6 (again, all concentrations are expressed as weight fractions). The
total matrix eect on Fe will be caused by both Ni and Cr and its magnitude will be
between M
Fe
(FeNi; W
Fe
0.1) 0.63 (for Fe in FeNi) and M
Fe
(FeCr; W
Fe
0.1) 2.8
(for Fe in FeCr). It is assumed that the total eect M
Fe
(FeNiCr; W
Fe
0.1) is proportional
to the concentrations of Ni (0.6) and Cr (0.3) in this example. Applying the law of
weighted averages, the total matrix eect is given by
M
Fe
FeNiCr; W
Fe
0.1 M
Fe
FeNi; W
Fe
0.1
W
Ni
W
Ni
W
Cr
M
Fe
FeCr; W
Fe
0.1
W
Cr
W
Ni
W
Cr
120
There is not a strict derivation indicating the validity of Eq. (120) in the general case.
For cases involving absorption only, the derivation is rather straightforward and based
on the additivity law for absorption. For now, let it suce to indicate that the matrix
eect will change gradually when adding an element or when changing the composition
of the specimen slightly; this is described by Eq. (120). By substituting the numerical
values,
M
Fe
FeNiCr; W
Fe
0.1 0.63
0.6
0.6 0.3
2.8
0.3
0.6 0.3
1.35 121
a value of 1.35 is obtained. This is in good agreement with the theoretical value of 1.31.
Equation (120) is based on the availability of binary inuence coecients calculated at
specimen compositions given by W
Fe
0.10; W
Ni
0.90 and W
Fe
0.10; W
Cr
0.90.
Using the more general expressions for matrix eects,
M
Fe
FeNi 1 m
FeNi.bin
W
Ni.bin
122a
and
M
Fe
FeCr 1 m
FeCr.bin
W
Cr.bin
122b
the following is obtained:
M
Fe
FeNiCr 1 m
FeNi.bin
W
Ni.bin
W
Ni
W
Ni
W
Cr
1 m
FeCr.bin
W
Cr.bin
W
Cr
W
Ni
W
Cr
123
where W
Ni,bin
is the concentration of Ni in the binary FeNi (for example, C
Ni,bin
0.90).
Because
W
Ni.bin
W
Cr.bin
1 W
Fe
W
Ni
W
Cr
124
Equation (123) can be rearranged to
M
Fe
FeNiCr 1 m
FeNi.bin
W
Ni
m
FeCr.bin
W
Cr
125
stressing the point again that the inuence coecients are to be calculated for binaries ij
with the composition W
i
. W
j
1 W
i
. This is the reason why Eqs. (106) and (117) use
W
m
instead of W
j
.
However, applying binary coecients to multielement specimens leads to an in-
complete matrix correction because we are trying to describe the eect of Cr and Ni on
Fe in FeNiCr based on the matrix eects in the corresponding binaries only. This eect
Copyright 2002 Marcel Dekker, Inc.
is referred as the crossed eect and has been described by Tertian (1987), who also
proposed a method to correct for this. The proposed method (Tertian, 1987) involves
the use of weighting factors based on the reciprocals of the relative intensities of the
binaries involved. It is a rather cumbersome method, but theoretically valid, and it does
not imply any approximation whatsoever; a discussion is outside the scope of this work.
An easier method is to use the cross-product coecients as used in Eq. (99). The
derivation from Tertian and Vie le Sage (1977) oers some insight in this matter.
Tertian and Vie le Sage (1977) assume that a multielement inuence coecient a
M
ij
can
be approximated as the sum of the binary coecient a
B
ij
and a linear variation with the
other elements:
a
M
ij
a
B
ij
t
ijk
W
k
126
where t
ijk
is a coecient expressing the eect of element k on the inuence coecient a
M
ij
.
Similarly,
a
M
ik
a
B
ik
t
ikj
W
j
127
Substituting Eqs. (126) and (127) in
1 a
M
ij
W
j
a
M
ik
W
k
128
(the superscript M is used to explicitly indicate the use of multi element inuence coe-
cients) yields
1 a
B
ij
W
j
a
B
ik
W
k
a
ijk
W
j
W
k
129
with
a
ijk
t
ijk
t
ikj
130
It is to be realized that crossed eect is introduced by the use of binary coecients; use of
multi element coecients would not lead to crossed eect. Thus, the equation expressing
the matrix eect using binary inuence coecients for specimens with more than two
compounds is
M
i
"
1
X
n
j1
j6i
m
ij.bin
W
j
X
n1
j1
j6i
X
n
kj1
k6i
a
ijk
W
j
W
k
#
131
and is based on cross-product coecients to correct for crossed eect introduced by the
use of binary inuence coecients. The use of the cross-product coecients is not
mandated by the concentration range to be covered (the binary coecients as calculated
by, for example, the algorithm of Lachance are more than adequate) but is a consequence
of the use of binary coecients.
7. Application
In Secs.V.D.4 and V.D.5, several inuence coecient algorithms have been discussed.
Application of the resulting equations for calibration and analysis will be discussed here
and is equally valid for any of the inuence coecient algorithms.
a. Calibration
Step 1. It is assumed that the coecients have been calculated from theory, for
example, using Eq. (84) or (97).
Copyright 2002 Marcel Dekker, Inc.
Step 2. Calculate the matrix correction term [the square brackets in Eq. (81), (85),
(99), (104), and (109)] for all standard specimens and for a given analyte. The coecients
are known (Step 1), and for standard specimens, all weight fractions W
i
and W
j
are
known.
Step 3. Plot the measured intensity of the analyte, multiplied by the corresponding
matrix correction term against analyte weight fraction. Then, determine the best line,
W
i
B
i
K
i
I
i
1 132
by minimizing DW
i
(see Sec. V.A). Note that Eq. (132) is more general than Eq. (50),
which does not correct for matrix eects. This process is repeated for all analytes. Other
methods are also feasible. The most common variant is the one where
W
1
B
i
K
i
I
i
133
is used. This is nearly equivalent to Eq. (132) but with brackets:
W
i
B
i
K
i
I
i
1 134
The term B
i
K
i
I
i
is related directly to the relative intensity R
i
. Corrections for line
overlap should only aect this term.
b. Analysis
For each of the analytes, a set of equations has to be solved for the unknown W
i
, W
j
and
so forth. If the matrix correction term used is the one according to Lachance and Trail
[Eq. (81)] or de Jongh [Eq. (85)], then the set of equations can be solved algebraically (n
linear equations with n unknowns for Lachance and Traill and n1 equations with n1
unknowns for de Jongh). Mostly, however, an iterative method is used. As a rst estimate,
one can simply take the matrix correction term equal to 1. This yields a rst estimate of the
composition W
i
, W
j
, and so on. This rst estimate is used to calculate the matrix cor-
rection terms for all analytes. Subsequently, a new composition estimate can be obtained.
This process is repeated until none of the concentrations changed between subsequent
iterations by more than a preset quantity.
If the matrix correction is done using algorithms which use more than one coecient
{e.g., Claisse and Quintin [Eq. (99)] or Rasberry and Heinrich [Eq. (104)]}, then the
equations are not linear in the unknown concentrations and an algebraic solution is not
possible. An iterative method, such as described earlier can be used.
8. Algorithmswith Empirical Coefficients
Empirical coecients are coecients that are not calculated from theory but from ac-
tually measured specimens using regression analysis (Anderson et al., 1974). They were
the basis of the earliest correction methods, but now they are largely superseded by more
theoretical ones. Such empirically determined coecients tend to mix the matrix correc-
tion with the sensitivity of the spectrometer. On the one hand, the matrix eect is de-
termined by the composition of the sample and physical parameters such as take-o
and incidence angles and tube anode and voltage. These are the same for spectrometers of
similar design. The sensitivity of the spectrometer, on the other hand, depends on the
reectivity of the crystals, the eciency of the detectors, and so on. These parameters are
unique for each spectrometer. Also, if one of the analyte lines is overlapped by another
x-ray line, some of this eect can also aect the value of the inuence coecients. The
Copyright 2002 Marcel Dekker, Inc.
coecients thus determined are instrument-specic and are not transferable to other
instruments.
Stephenson (1971) has noted that the regression equations involved in the determi-
nation of the coecients in such an empirical way become unstable as the degree of
correlation between the independent variables increases. This mandates careful planning
of the experiment, including the composition of the synthetic standards. Klimasara (1994,
1995) has illustrated the use of standard spreadsheet programs for the calculation of the
values of empirical inuence coecients and composition.
a. The Sherman Algorithm
Sherman (1953) was among the rst to propose an algorithm for correction of matrix
eects. For a ternary system, the algorithm can be represented by the following set of
equations:
a
AA
t
A
W
A
a
AB
W
B
a
AC
W
C
0
a
BA
W
A
a
BB
t
B
W
B
a
BC
W
C
0 135
a
CA
W
A
a
CB
W
B
a
CC
t
C
W
C
0
where a
ij
represents the inuence coecient of element j on the analyte i and t
i
is the time
(in s) required to accumulate a preset number of counts. The constants a
ij
are determined
from measurements on specimens with known composition. Determination of the com-
position of an unknown involves the solving of the above set of linear equations
[Eq. (135)]. This set, however, is homogeneous: Its constant terms are all equal to zero. So,
only ratios among the unknown W
i
can be obtained. In order to obtain the weight frac-
tions W
i
, an extra equation is required. Sherman proposed using the sum of all the weight
fractions of all the elements (or components) in the specimen, which ideally, should be
equal to unity. For a ternary specimen,
W
A
W
B
W
C
1 136
Using Eq. (136), one of the equations in the set of Eqs. (135) can be eliminated. The
solution obtained, however, is not unique: For a ternary, any one of the three equations
can be eliminated. This yields three dierent combinations. Furthermore, any of the
three elements can be eliminated in each of the combinations. Hence, a total of 363 9
dierent sets can be derived from Eqs. (135) and (136), and each of these sets will
generate dierent results. In general, the algorithm yields n
2
dierent results for a system
with n elements or compounds. This is clearly undesirable, because it is hard to de-
termine which set will give the most accurate results. Another disadvantage is the fact
that the sum of the elements determined always equals unity, even if the most abundant
element has been neglected. Furthermore, the numerical values of the coecients de-
pend, among other parameters such as geometry and excitation conditions, also on the
number of counts accumulated. Nonquantiable parameters, such as reectivity of the
diracting crystal used in wavelength-dispersive spectrometers, or tube contamination
will also aect the value of the coecients. The coecients determined on a given
spectrometer cannot be used with another instrument; they are not transferable.
The other algorithms discussed use some form of a ratio method: The Lachance and
Traill algorithm, for example, uses relative intensities. The measurements are then done,
relative to a monitor; this reduces, or eliminates, the eect of such nonquantiable
parameters.
Copyright 2002 Marcel Dekker, Inc.
b. The Algorithm of Lucas-Tooth and Price
Lucas-Tooth and Price (1961) developed a correction algorithm, where the matrix eect
was corrected for, using intensity (rather than concentration) of the interfering elements.
The equation can be written as
W
i
B
i
I
i
k
0
X
n
j6i
k
ij
I
j
" #
137
where B
i
is a background term and k
0
and k
ij
are the correction coecients. A total of
n 1 coecients have to be determined, requiring at least n 1 standards. Usually,
however, a much larger number of standards is used. The coecients are then determined
by, for example, a least-squares method. The corrections for the eect of the matrix on the
analyte are done via the intensities of the interfering elements; their concentrations are not
required. The method assumes that the calibration curves of the interfering elements
themselves are all linear; the correction is done using intensities rather than concentra-
tions. The algorithm will, therefore, have a limited range. Its use will be limited to ap-
plications where only one or two elements are to be analyzed (it still involves
measurements of all interfering element intensities) and where a computer of limited
capabilities is used (although calculation of the coecients involves much more compute
capabilities than the subsequent routine analysis of unknowns).
The advantages of the method are as follows:
The method is very fast, because the calculation of composition of the unknowns
requires no iteration.
Analysis of only one element is possible; this requires, however, the determination of
all relevant correction factors.
Very simple algorithm, requiring very little calculation.
c. Algorithms Based on Concentrations
Algorithms similar to Eq. (137) have been proposed, using corrections based on con-
centrations rather than intensities. The values of the coecients were then to be derived
from multiple-regression analysis on a large suite of standards. The main aim was to
obtain correction factors that could be determined on one spectrometer and used, without
alteration, on another instrument. In practice, the coecients still have to be adjusted
because of the intimate and inseparable entanglement of spectrometer-dependent factors
with matrix eects. Furthermore, compared to the algorithms based on intensities, some of
the advantages of the latter are not retained: A calibration for all elements present is now
required, calculation of the composition of unknowns requires iteration, and so forth.
In principle, methods based on theoretically calculated inuence coecients are re-
commended.
VI. CONCLUSION
Among the advantages of XRF analysis are the facts that the method is nondestructive
and allows direct analysis involving little or no specimen preparation. Analysis of major
and minor constituents requires correction for matrix eects of variable (from one spe-
cimen to another) magnitude. If the matrix varies appreciably from one specimen to the
next, then even the intensity of elements present at a trace level can be subject to matrix
eects and a correction is required. Several methods for matrix correction have been
Copyright 2002 Marcel Dekker, Inc.
described. Each of these methods have their own advantages and disadvantages. These, by
themselves, do not generally lead to the selection of best method. The choice of the
method to use is also determined by the particular application.
From the previous sections, it may appear that the mathematical methods are more
powerful than the compensation methods. Yet, if only one or two elements at a trace level
in liquids have to be determined, compensation methods (either standard addition or the
use of an internal standard) can turn out to be better suited than, for examples, rigorous
fundamental parameter calculations. Compensation methods will correct for the eect of
an unknown, but constant, matrix. Also, they do not require the analysis of all con-
stituents in the specimen. The mathematical methods (fundamental parameters as well as
methods based on theoretical inuence coecients), on the other hand can handle cases in
which the matrix eect is more variable from one specimen to another. In this respect, they
appear to be more exible than the compensation methods, but they do require more
knowledge of the complete matrix. All elements contributing signicantly to the matrix
eect must be quantied (either by x-ray measurement or by another technique) even if the
determination of their concentrations is not required by the person who submits the
sample to the analyst. Once a particular algorithm is selected, it is customary to use for all
analytes. However, it must be stressed that this is not a requirement. There is only one
requirement for adequate matrix correction: Each analyte should be corrected adequately,
by whatever method.
If complete analysis (covering all major elements) is required, the analyst has the
choice between the fundamental parameter method and algorithms, based on inuence
coecients. Commonly, fundamental parameter methods are (or were) used in research
environments rather than for routine analysis in industry. This choice is more often made
on considerations such as the availability of the programs and computers than on dif-
ferences in analytical capabilities. Inuence coecient algorithms tend to be used in
combination with more standards compared to fundamental parameter methods, because
their structure and simple mathematical representation facilitates interpretation of the
data (establishing a relationship between concentration and intensity, corrected for matrix
eect). The nal choice, however, has to be made by the analyst.
REFERENCES
Anderson CH, Mander JE, Leitner JW. Adv X-Ray Anal 17:214, 1974.
Australian Standard 2563-1982, Wavelength Dispersive X-ray Fluorescence Spectrometers
Methods of Test for Determination of Precision. North Sydney, NSW: Standards Association
of Australia, 1982.
Bambynek W, Crasemann B, Fink RW, Freund HU, Mark H, Swift CD, Price RE, Venugopala
Rao P. Rev Mod Phys 44:716, 1972.
Bearden JA. Rev Mod Phys 39:78, 1967.
Beattie HJ, Brissey RM. Anal Chem 26:980, 1954.
Bonetto RD, Riveros JA. X-Ray Spectrom 14:2, 1985.
Claisse F, Quintin M. Can Spectrosc 12:129, 1967.
Criss JW. Adv X-Ray Anal 23:93, 1980a.
Criss JW. Adv X-Ray Anal 23:111, 1980b.
Criss JW, Birks LS. Anal Chem 40:1080, 1968.
Criss JW, Birks LS, Gilfrich JV. Anal Chem 50:33, 1978.
de Boer DKG. Spectrochim Acta 44B:1171, 1989.
de Jongh WK. X-Ray Spectrom 2:151, 1973.
Copyright 2002 Marcel Dekker, Inc.
de Jongh WK. X-Ray Spectrom 8:52, 1979.
DeGroot PB. Adv X-Ray Anal 33:53, 1990.
Draper NR, Smith H. Applied Regression Analysis. New York: Wiley, 1966.
Feather CE, Willis JP. X-Ray Spectrom 5:41, 1976.
Garbauskas MF, Goehner RP. Adv X-Ray Anal 26:345, 1983.
Gillam E, Heal HT. Br J Appl Phys 3:353, 1952.
Heinrich KFJ. In: McKinley TD, Heinrich KFJ, Wittry DB, eds. The Electron Microprobe.
New York: Wiley, 1966, p 296.
Holynska B, Markowicz A. X-Ray Spectrom 10:61, 1981.
Hower J. Am Mineral 44:19, 1959.
Hughes H, Hurley P. Analyst 112:1445, 1987.
Hunter CB, Rhodes JR. X-Ray Spectrom 1:107, 1972.
Ingham MN, Vrebos BAR. Adv X-Ray Anal 37:717, 1994.
ISO, Determination of Nickel and Vanadium in Liquid FuelsWavelength-Dispersive X-Ray
Fluorescence Method, ISO 14597. Geneva: ISO, 1995.
Johnson W. International Report BISRA MG,D,Conf Proc,610,67, 1967.
Kaufmann M, Mantler M, Weber F. Adv X-Ray Anal 37:205, 1994.
Klimasara AJ. Adv X-Ray Anal 37:647, 1994.
Klimasara AJ. Workshop on the Use of Spread Sheets in XRF Analysis, 44th Annual Denver X-Ray
Conference, Colorado Springs, CO, 1995.
Lachance GR. X-Ray Spectrom 8:190, 1979.
Lachance GR. International Conference on Industrial Inorganic Elemental Analysis, Metz, France,
1981.
Lachance GR. Adv X-Ray Anal 31:471, 1988.
Lachance GR, Claisse F. Adv X-Ray Anal 23:87, 1980.
Lachance GR, Traill RJ. Can Spectrosc 11:43, 1966.
Leroux J, Thinh TP. Revised Tables of Mass Attenuation Coecients. Quebec: Corporation
Scientique Claisse, 1977.
Li-Xing Z. X-Ray Spectrom 13:52, 1984.
Lubecki A, Holynska B, Wasilewska M. Spectrochim Acta 23B:465, 1968.
Lucas-Tooth HJ, Price BJ. Metallurgia 64:149, 1961.
Mainardi RT, Fernandez JE, Nores M. X-Ray Spectrom 11:70, 1982.
Mantler M. Adv X-Ray Anal 36:27, 1993.
McMaster WH, Delgrande NK, Mallet JH, Hubbel JH. Compilation of X-Ray Cross Sections,
UCRL 50174, Sec II, Rev 1, 1969.
Mu ller RO. Spectrochemical Analysis by X-Ray Fluorescence. New York: Plenum Press, 1972,
chap 9.
Pella PA, Feng LY, Small JA. X-Ray Spectrom 14:125, 1985.
Pella PA, Tao GY, Lachance GR. X-Ray Spectrom 15:251, 1986.
Pollai G, Ebel H. Spectrochim Acta 26B:761, 1971.
Pollai G, Mantler M, Ebel H. Spectrochim Acta 26B:733, 1971.
Rasberry SD, Heinrich KFJ. Anal Chem 46:81, 1974.
Rousseau RM. X-Ray Spectrom 13:121, 1984a.
Rousseau RM. X-Ray Spectrom 13:3, 1984b.
Rousseau RM, Claisse F. X-Ray Spectrom 3:31, 1974.
Sherman J. The Correlation Between Fluorescent X-Ray Intensity and Chemical Composition.
ASTM Special Publication No 157. Philadelphia: ASTM, 1953, p 27.
Sherman J. Spectrochim Acta 7:283, 1955.
Shiraiwa T, Fujino N. Jpn J Appl Phys 5:886, 1966.
Shiraiwa T, Fujino N. Bull Chem Soc Japan 40:2289, 1967.
Shiraiwa T, Fujino N. Adv X-Ray Anal 11:63, 1968.
Shiraiwa T, Fujino N. X-Ray Spectrom 3:64, 1974.
Sparks CJ. Adv X-Ray Anal 19:19, 1976.
Copyright 2002 Marcel Dekker, Inc.
Stephenson DA. Anal Chem 43:310, 1971.
Tao GY, Pella PA, Rousseau RM. NBSGSC, a FORTRAN program for quantitative X-Ray
Fluorescence Analysis. NBS Technical Note 1213. Gaithersburg, MD: National Bureau of
Standards, 1985.
Tertian R. Adv X-Ray Anal 19:85, 1976.
Tertian R. X-Ray Spectrom 16:261, 1987.
Tertian R, Vie le Sage R. X-Ray Spectrom 6:123, 1977.
Vrebos BAR, Helsen JA. X-Ray Spectrom 15:167, 1986.
Vrebos BAR, Pella PA. X-Ray Spectrom 17:3, 1988.
Veigele WJ. In: Robinson JW, ed. Handbook of Spectroscopy. Cleveland, OH: CRC, 1974, p 28.
Venugopala Rao P, Chen MH, Crasemann B. Phys Rev A 5:997, 1972.
Wadleigh KR. X-Ray Spectrom 16:41, 1987.
Wood PR, Urch DS. J Phys F: Metal Phys 8:543, 1978.
SUGGESTIONS FOR FURTHER READING
Bertin EP. Principles and Practice of X-Ray Spectrometric Analysis. 2nd ed. New York: Plenum
Press, 1975.
Jenkins R, Gould RW, Gedcke D. Quantitative X-Ray Spectrometry. 2nd ed. New York: Marcel
Dekker, 1995.
Tertian R, Claisse F. Principles of Quantitative X-Ray Fluorescence Analysis. New York: Wiley,
1982.
Copyright 2002 Marcel Dekker, Inc.
6
Quantication in XRF Analysis
of Intermediate-Thickness Samples
Andrzej A. Markowicz
Vienna, Austria
Rene E. Van Grieken
University of Antwerp, Antwerp, Belgium
I. INTRODUCTION
A number of approaches have been developed for quantitation in x-ray uorescence
(XRF) analysis of intermediate-thickness samples whose mass per unit area m fullls the
relation
m
thin
< m < m
thick
1
where m
thin
and m
thick
are the values of mass per unit area for thin and thick samples [for
a denition, see Chapter 1, Eqs. (93) and (95)]. Intermediate samples can be preferable
to thick specimens because remaining uncertainties about mass-attenuation coecients
have a smaller eect on the analysis results, less material is required, the sensitivity
is more favorable for low-Z elements, and secondary enhancement eects are less
important.
Historically, the oldest correction method applied in quantitative XRF analysis of
intermediate-thickness samples is the emissiontransmission (ET) method in which the
specic x-ray intensities from a sample are measured successively with and without a
target positioned adjacent to the back of the sample in a xed geometry. Recently, the
ET method has thoroughly been evaluated, and both the advantages and limitations of
the technique are well identied. A number of modications of the ET method, de-
veloped in recent years, allowed an essential extension of its applicability range. To
avoid additional measurements that are inevitable in the emissiontransmission method,
some alternative correction procedures based on the use of scattered primary radiation
were also developed. In recent years, however, only a few papers have been published in
this eld.
The underlying principles as well as the ranges of applicability and the limitations of
the correction procedures applied to XRF analysis of both homogeneous and hetero-
geneous intermediate-thickness samples are outlined here.
Copyright 2002 Marcel Dekker, Inc.
II. EMISSION---TRANSMISSIONMETHOD
In the absence of enhancement eects and assuming monochromatic excitation, the mass
per unit area of the element i. m
i
, for homogeneous intermediate-thickness samples, can be
calculated from [see Eq. (90) in Chapter 1]:
m
i
I
i
E
i
B
i
Ab
corr
2
where
B
i
GI
0
E
0
E
i
t
0
i
E
0
.
i
p
i
1 1,j
i
sin
1
3
and Ab
corr
is the absorption correction factor given by
Ab
corr
jE
0
csc
1
jE
i
csc
2
m
1 expfjE
0
csc
1
jE
i
csc
2
mg
4
[The symbols used in Eqs. (2)(4) are explained in Chapter 1, Eqs. (88) and (89).]
The value of the constant B
i
(sometimes called the sensitivity factor) can be de-
termined either experimentally as the slope of the straight calibration line for the ith
element obtained with thin homogeneous samples or semiempirically based on both the
experimentally determined GI
0
E
0
value and the relevant fundamental parameters (Yap
et al., 1987; Markowicz et al., 1992a).
The absorption correction factor Ab
corr
represents the combined attenuation of
the primary and uorescent radiations in the whole specimen and can be determined
individually for each sample by transmission experiments (Leroux and Mahmud, 1996;
Giauque et al., 1973). These are done by measuring the x-ray intensities with and without
the specimen from a thick multielement target located at a position adjacent to the back of
the specimen, as shown in Figure 1.
If I
i
S
. I
i
T
. and I
i
0
are the intensities after background correction from the
sample alone, from the sample plus target, and from the target alone, respectively, then the
combined fraction of the exciting and uorescent radiations transmitted through the total
sample thickness is expressed by
expfjE
0
csc
1
jE
i
csc
2
mg
I
i
T
I
i
S
I
i
0
H 5
After a simple transformation, Eq. (2) can be rewritten as
Figure1 Experimental procedure used in the emissiontransmission method for the correction of
matrix absorption effects.
Copyright 2002 Marcel Dekker, Inc.
m
i
I
i
E
i
B
i
ln H
1 H
6
The emissiontransmission method can only be applied in the quantitative XRF
analysis of homogeneous samples of which the mass per unit area is smaller than the
critical value, m
crit
. dened by
m
crit
ln H
crit
jE
0
csc
1
jE
i
csc
2
7
where H
crit
is the critical value of the transmission factor dened by Eq. (5); in practice,
H
crit
0.1 or 0.05.
To minimize possible absorption correction errors resulting from enhancement of
the specimen radiation by scattered target radiation, targets that yield a high ratio of
scattered to uorescent radiation should not be used.
Giauque et al. (1979) developed a modied version of the emissiontransmission
method. Using data from the attenuation measurements and Eq. (5), the values of
jE
0
csc
1
jE
i
csc
2
m j
T
m are calculated for the energies of characteristic
x-rays of all elements present in a thick multielement target. If these values are plotted
versus the uorescence x-ray energy on a loglog scale (Fig. 2), an approximate value for
Figure 2 Curves of [m(E
0
) csc
1
j(E
i
) csc
2
]m and mm(E
i
) csc
2
values versus fluorescence
x-ray energy for an NBS SRM 1632 coal specimen. (From Giauque et al., 1979. Reprinted with
permission from Analytical Chemistry. Copyright American Chemical Society.)
Copyright 2002 Marcel Dekker, Inc.
jE
0
csc
1
can be obtained by extrapolation of the curve to the energy of the excitation
radiation. In turn, values for mjE
i
csc
2
can be calculated, a curve for these values
drawn, and a new value for mjE
0
csc
1
established. This last step is iterated several
times. Using data from the latter curve, the absorption correction factors for all radiations
of interest can be calculated from Eq. (4). If some elements to be determined are major or
minor constituents, a few separate curves for mjE
i
csc
2
values should be plotted be-
tween the preselected x-ray energies corresponding to the relevant absorption edges. In the
emissiontransmission method of Giauque et al. (1979), the incoherent scattered radiation
corrected for matrix absorption is used as the internal standard to compensate for var-
iations in sample mass, x-ray tube output, and sample geometry.
In practical applications of the emissiontransmission method, the values of the
transmission factor H [Eq. (5)] determined for a few energies of the characteristic x-rays
of some elements present in a multielement target are used to construct a curve
ln lnH
1
lnj
T
m versus ln E. which enables one to calculate the absorption cor-
rection factor Ab
corr
for any energy. In some cases, however, the relationship of
ln lnH
1
versus ln E exhibits some discontinuities which correspond to the absorption
edges of minor (major) elements present in the unknown samples. In such cases the
approach proposed by Giauque et al. (1979), based on the construction of separate
curves for the predened energy regions, can only be used if at least two experimental
points are available for each energy region. When only one or no experimental point is
available, a modied version of the emissiontransmission method can be used (Mar-
kowicz and Haselberger, 1992). To explain the modied procedure, let us assume that a
multielement target consists of Ca, Ti, Fe, Zn, Sr, Zr, and Pb, and the material to be
analyzed contains some minor elements, such as Fe and Ca. From the ET measure-
ments, the values of A A H
1
are easily obtained for higher energies (e.g., for the
characteristics x-rays of Fe, Zn, Sr, Zr, and Pb in this case) (see Fig. 3). First, a straight
line is tted to the points corresponding to Zn, Pb, Sr, and Zr (region I). Second, a
straight line of the same slope and passing through a point for Fe is constructed (region
II). If no experimental point is available for Ca, the discontinuity for the K absorption
edge of Ca has to be taken into account by applying a computation routine. In the rst
step, the product
1
W
Ca
j
above
j
below
is calculated, where W
Ca
is the weight
fraction of Ca obtained by using a direct extrapolation of curve II and j
above
and j
below
are the mass absorption coecients just above and below the K absorption edge of Ca,
respectively. Next, the corrected value of the total mass absorption coecient
j
cor
T
ln A,m
1
csc
2
is calculated for the energy of the CaK
abs
edge, where A is
taken from curve II and
2
is the emerging angle for the characteristic x-rays. Based on
the j
cor
T
value, a straight line passing through point C and having the same slope as that
one in region II is constructed (the coordinates of the point C are x ln CaK
abs
and
y lnln A m
1
csc
2
). In the next step, the values of ln j
T
m and the absorption
correction factor Ab
corr
are calculated for the energy of Ca characteristic x-rays based
on the straight line for region III. Because the true value of the weight fraction for Ca is
unknown, all calculations for the absorption edge of Ca (region III) are iterated until the
following convergence is obtained:
W
Ca
n
W
Ca
n1
W
Ca
n
0.001 8
where W
Ca
n
and W
Ca
n1
are the values of the weight fractions for Ca obtained in the
subsequent iterations.
Copyright 2002 Marcel Dekker, Inc.
A calculation procedure applied for the Ca absorption edge can easily be repeated
for any other absorption discontinuity at a lower energy and, nally, a total absorption
curve ln ln A versus ln E is obtained.
The importance of the additional correction for the discontinuities in the absorption
properties of the samples to be analyzed was demonstrated, among others, for the analysis
of coal samples. Neglecting the absorption edges correction for Fe and Ca at the con-
centration level of around 1% resulted in considerable errors of the analysis, up to 60%
for the determination of Ca (Markowicz and Haselberger, 1992).
Van Dyck et al. (1980) developed a correction method that allows calculations of the
absorption coecients (and absorption correction factors as well) at any energy for in-
termediate-thickness samples, without additional measurements, by using the ratio of the
x-ray signals from a Zr wire positioned in front of the sample and from a Pd foil placed
behind the sample, both in a xed geometry, as shown in Figure 4. The Zr wire provides an
external reference signal ZrKc. which is applied for normalization of all measured
uorescent intensities to reduce considerably the eect of variations in exciting x-ray tube
intensity and of dead-time losses.
The coecients for higher energies are calculated with an iterative program from the
experimentally measured absorption coecient at the PdL energy (2.9 keV), jE
Pd
. In the
rst step, the total attenuation coecient at the PdL energy, caused exclusively by the low-
Z elements (e.g., Z < 17) in the sample that show no characteristic peak above 3.0 keV in
the spectrum, is calculated from the normalized measured intensities, I
Pd
T
and I
Pd
0
and
Figure3 Graphical explanation of the idea for the modified version of the ET method: see text for
details. (From Markowicz and Haselberger, 1992. Reprinted with permission of Pergamon Press Ltd.)
Copyright 2002 Marcel Dekker, Inc.
the dierent characteristic peaks recorded in the spectrum. This attenuation coecient due
to low-Z elements, j
low-Z
E
Pd
, is given by (Van Dyck et al., 1980)
j
low-Z
E
Pd
lnI
Pd
0
,I
Pd
T
,mcsc
2
jE
0
csc
1
,csc
2
P
n
0
j1
W
j
j
j
E
Pd
1
P
n
0
j1
W
j
9
where n
0
is the number of characteristic peak in the XRF spectrum and j
j
E
Pd
is the mass
attenuation coecient for the PdL energy in the element j (McMaster et al., 1969), giving
rise to a characteristic peak in the XRF spectrum.
The weight fraction of the element j in the sample, W
j
, is correlated with the recorded
characteristic peak through the sensitivity factor, routinely obtained by measuring thin
single or multielement standards. As a rst approximation, the characteristic peak in-
tensities are not corrected for absorption. In a second step, based on the j
low-Z
E
Pd
value,
the absorption coecients of the low-Z matrix for other energies, j
low-Z
E, are calculated
quantitatively by assuming nearly parallel properties of the logarithmic absorption curves
ln jE versus ln E.
The total mass attenuation coecient jE for the characteristic x-rays of an element
in the whole sample can now be calculated, taking into account the contributions j
j
E of
the high-Z elements:
jE 1
X
n
0
j1
W
j
!
j
low-Z
E
X
n
0
j1
W
j
j
j
E 10
The overall procedure is summarized and schematically represented in Figure 5.
Better jE values [Eq. (10)] are obtained in the second and following loops by
carrying out appropriate absorption corrections to the characteristic intensities from
which the W
j
values are derived using jE and jE
0
values from the previous loop and
by including in Eq. (9) the jE
0
values, taken as zero in the rst loop. The iteration is
stopped when the dierence in jE between two loops is negligible.
Subroutine ENHANC is applied to evaluate the enhancement eect of the PdL
x-rays caused by all elements in the sample of which the x-ray energy is higher than the
L
1
. L
2
and L
3
absorption edges of Pd. The intensity caused by this enhancement eect can
be considered to result in an apparent increase in I
Pd
0
. The enhancement contributions,
I
enhPd
, are added to the I
Pd
0
in Eq. (9). This total I
Pd
Oeff
value and the detected I
Pd
T
signal allow us to calculate the correct j values via Eqs. (9) and (10).
A comprehensive discussion of the inuence of secondary enhancement of the PdL
x-rays by the samples as well as of the inuence of two other complicating factors, grain
Figure 4 Measurement geometry applied in an automatic absorption correction method. (From
Van Dyck et al., 1980. Reprinted by permission of John Wiley & Sons, Ltd.)
Copyright 2002 Marcel Dekker, Inc.
size eects, and a heterogeneous sample load is presented in the work of Van Dyck et al.
(1980). The grain size and sample heterogeneity eects induce inaccuracies on the ab-
sorption coecient determinations that may well reach 20% for particulate samples, such
as intermediate-thickness deposits of geological materials. Thus, this approach has the
same limitations as all emissiontransmission methods applied to heterogeneous samples.
A. Accuracy and Limitations of the Emission--Transmission Method
The overall (total) absolute uncertainty of the mass per unit area of the ith element, m
i
, can
be calculated fromthe lawof error propagation applied to Eq. (2) (Markowicz et al., 1992b):
m
i
Ab
corr
B
i
2
I
i
E
i
I
i
E
i
B
i
2
Ab
corr
I
i
E
i
Ab
corr
B
2
i
2
B
i
2
" #
1,2
11
where I
i
E
i
is the absolute uncertainty of the characteristic x-ray intensity I
i
E
i
(usually
provided by a computer routine applied for spectrum evaluation), Ab
corr
is the absolute
uncertainty of the absorption correction factor, and B
i
is the absolute uncertainty of the
sensitivity (or calibration) factor B
i
(usually provided by a computer calibration routine,
as a mean percent dierence between the experimental and tted values for the calibration
factors). The total relative uncertainty of the mass per unit area m
i
,m
i
is calculated from
m
i
m
i
I
i
E
i
I
i
E
i
Ab
corr
Ab
corr
2
B
i
B
i
2
" #
1,2
12
Figure 5 The calculation steps. From the experimentally measured total absorption coefficient at
the PdL energy, m(E
Pd
), the calculated contribution from high-Z elements (giving a characteristic
peak in the spectrum) is substracted; through this low-Z matrix contribution at the PdL energy, the
total low-Z absorption curve (dashed line) is calculated; the contribution from high-Z elements (e.g.,
Ca and Fe) is then added, to yield the total absorption curve (thick line). (From Van Dyck et al.,
1980. Reprinted by permission of John Wiley & Sons, Ltd.)
Copyright 2002 Marcel Dekker, Inc.
In order to calculate Ab
corr
, one has to apply the law of error propagation to Eqs. (4)
and (5):
Ab
corr
A 1 ln A
A 1
2
A 12a
with A H
1
[Eq. (5)] and
A
I
i
2
I
i
T
I
i
2
I
i
2
I
i
2
I
i
T
I
i
4
I
i
2
I
i
2
I
i
T
I
i
4
" #
1,2
12b
where I
i
0
and I
i
T
are the absolute uncertainities of the characteristic x-ray
intensities obtained from the target alone and from the sample plus target, respectively.
After a simple transformation, m
i
,m
i
can be given by
m
i
m
i
I
i
E
i
I
i
E
i
A 1 ln A
2
A 1
2
A
2
ln A
2
A
2
B
i
B
i
2
" #
1,2
13
It is obvious that the total relative uncertainty of the weight fraction of the ith element
W
i
,W
i
is equal to m
i
,m
i
.
Equations (13) and (12b) can be used for calculating the total uncertainities of
the emissiontransmission method as well as the contribution from various sources of
the uncertainties I
i
E
i
. Ab
corr
. B
i
. It has been demonstrated that in the de-
termination of trace elements, the dominant contribution to the total uncertainty is
from the uncertainty of the peak-area calculation for the characteristic x-rays of the
element of interest. In this case, the contribution from the uncertainties of the ab-
sorption correction factor and sensitivity can practically be neglected. In the determi-
nation of major elements, the largest contribution to the total uncertainty is from the
uncertainty of the absorption correction factor and,or sensitivity; in this case, the
contribution of the uncertainty of the peak-area calculation can safely be neglected.
Such a detailed analysis of the contribution of various sources of uncertainties can
be useful when there is a need to identify the most critical point of the whole
analytical procedure and some reduction of the total uncertainty of the analysis is
required.
Another factor which inuences the accuracy of the emissiontransmission
method is the heterogeneity of the sample loading. Markowicz and Abdunnabi (1991)
derived general expressions describing the accuracy of the ET method for various
types of sample loading heterogeneities (including incompletely loaded samples) within
a wide range of sample thicknesses and characteristics x-ray energies. The results,
conrmed by experiments, clearly show that inhomogeneities of sample loading may
critically aect the accuracy of the ET method, in particular in the analysis of in-
termediate-thickness samples with strong inhomogeneities and large values for the
jE
0
csc
1
mand jE
i
csc
2
m factors. The most critical inuence of sample loading
heterogeneity on the accuracy of the analytical results is observed for incompletely
loaded samples of intermediate thickness. The method proposed in the work of
Markowicz and Abdunnabi (1991) gives a possibility for evaluating the errors arising
from some types of sample loading heterogeneities as well as for evaluating the range
of applicability of the simplest version of the ET method (Markowicz et al., 1992b) if
a certain value for the total uncertainty W
i
,W
i
is accepted.
Copyright 2002 Marcel Dekker, Inc.
III. ABSORPTIONCORRECTIONMETHODSVIASCATTEREDPRIMARY
RADIATION
The use of scattered primary radiation in XRF analysis provides an alternative to the
common problem of matching standards of similar composition to samples to be analyzed.
The backscatter peaks are sometimes treated as uorescent peaks from internal standards
because they suer matrix absorption similar to that of uorescent peaks and behave si-
milarly with instrumental variations. They also provide the only direct spectral measure of
the total or average matrix of geological, biological, or other materials containing large
quantities of light elements, such as carbon, nitrogen, and oxygen, usually not observed by
their characteristic x-ray peaks.
X-ray uorescence matrix correction methods based on the use of scattered radiation
have mostly been applied in quantitative analysis of innitely thick samples under a wide
variety of experimental conditions, including discrete and continuum primary radiation
sources and detection by both wavelength- and energy-dispersive systems (Andermann
and Kemp, 1958; Kalman and Heller, 1962; Taylor and Andermann, 1971, 1973; Leo-
nardo and Saitta, 1977; Livingstone, 1982; Kikkert, 1983; Markowicz, 1984). The scat-
tered-radiation methods utilize the incoherent (Compton) and,or coherent (Rayleigh)
scatter peaks from line excitation sources or the intense high-energy region from con-
tinuum sources.
A. Absorption Corrections Based on Incoherent Scattered Radiation
For specimens of less than innite thickness, the intensity I
Com
of Compton-scattered
radiation can be expressed by (Meier and Unger, 1976)
I
Com
k
0
I
0
o
Com
E
0
1 expfjE
0
csc
1
jE
Com
csc
2
mg
jE
0
csc
1
jE
Com
csc
2
14
where k
0
is a constant for a given measurement geometry and detection eciency, o
Com
E
0
is the Compton mass-scattering coecient of the sample material for the primary radiation
of energy E
0
cm
2
,g. jE
Com
is the total mass-attenuation coecient of the sample
material for Compton-scattered primary radiation of energy E
Com
cm
2
,g [see Eq. (52) in
Chapter 1]. Equation (14) is valid for monochromatic excitation.
Assuming that the relation of the atomic number to the mass number is constant for
every element to be found in the sample and that jE
0
jE
Com
, the following sim-
plied formula for the intensity of Compton-scattered radiation from a multielement
sample can be obtained:
I
Com
k
0
1
I
0
f1 exp jE
0
csc
1
1 csc
2
,csc
1
m g
jE
0
csc
1
1 csc
2
,csc
1
15
where k
0
1
is a constant for given geometry of measurement and the energy of the incident
radiation. The value of the product k
0
1
I
0
is determined experimentally from a reference
scatterer and is valid as long as the reference scatterer has a matrix that is not too dierent
from that of the specimens.
Kieser and Mulligan (1979) worked out a method based on the use of the incoherent
scatter radiation, which gives accurate mass absorption coecients for a limited average Z
range. The mass absorption coecient jE
0
for specimens of intermediate thickness is
Copyright 2002 Marcel Dekker, Inc.
found from Eq. (15) after a numerical solution. To obtain a value of the mass absorption
coecient jE at any energy E, Kieser and Mulligan (1979) assumed that the slope of the
curve log jE versus log E is constant for all elements (approximately 2.7) over a range
of x-ray energies. The proposed Compton-scattered method for the determination of an
intermediate specimens mass absorption coecient at any energy can be applied as long
as no absorption edge of a major or minor element intervenes. When the values of the jE
are determined, calculation of the absorption correction factor [Eq. (4)] is straightforward
if, of course, the mass of the sample to be analyzed is known.
A modied uorescent Compton correction method for quantitative XRF of in-
termediate specimens was developed by Holynska and Markowicz (1979). The method is
based on the use of the measured x-ray uorescent intensities of all determined elements
and the intensity of Compton-backscattered radiation. The authors derived the following
expression for the determination of the mass per unit area of the element i, m
i
:
m
i
a
i
I
i
1
P
n
0
l1
a
il
I
l
b
i
I
Com
,m
16
where a
i
. a
il
. and b
i
are constant coecients obtained experimentally on the basis of
standard samples and n
0
is the number of the elements to be determined, including the ith
element. As is seen from Eq. (16), the absorption matrix correction is carried out via the
intensities of characteristic x-rays of all elements determined and the intensity I
Com
, re-
ecting, for the most part, the variations in the composition of light matrix. To apply the
absorption matrix correction, Eq. (16), the total mass per unit area of the specimen must
be evaluated, for example, by sample weighing. This uorescentCompton correction
method can be used in the XRF analysis of homogeneous intermediate-thickness samples
in a limited range of mass per unit area (m<10m
thin
).
B. Absorption Corrections Based on Both Coherent and Incoherent
Scattered Radiations
Several absorption correction methods based on both coherent and incoherent scattered
radiations have been developed, mostly in the 1970s and 1980s, and applied in quantitative
XRF analysis of intermediate samples. This group of correction methods is represented
either by relatively simple approaches (Bazan and Bonner, 1976; Markowicz, 1979) or by
very sophisticated fundamental parameter procedures (Nielson, 1977; Van Dyck and Van
Grieken, 1980; Nielson et al., 1982; Nielson and Rogers, 1984) providing superior ana-
lytical exibility.
For specimens of less than innite thickness, the intensity I
coh
of coherent scattered
radiation can be calculated from
I
coh
k
00
I
0
o
coh
E
0
jE
0
csc
1
csc
2
1 exp jE
0
csc
1
csc
2
m f g 17
where k
00
is a constant for a given measurement geometry and detection eciency for the
primary x-rays of energy E
0
, and o
coh
E
0
is coherent mass-scattering coecient of the
sample material for the primary radiation (cm
2
,g) [see Eq. (73) in Chapter 1].
Bazan and Bonner (1976) showed, for the rst time, a linear relation between the
eective absorption coecient (dened as the sum of the sample absorption coecients for
exciting and characteristic x-rays) and the ratio of incoherent to coherent scattering.
Copyright 2002 Marcel Dekker, Inc.
However, the coecients of the calibration line varied somewhat with the matrix, and this
hampered practical applications of this simple approach.
Markowicz (1979) found that, theoretically the sensitivity of the absorption cor-
rection via the incoherent,coherent scattered x-ray intensities ratio is better than that of
the absorption procedure involving each of the scattered radiations individually. For in-
termediate-thickness samples, in a limited range of rather small values of mass per unit
area, the intensities of the Compton-scattered radiation I
Com
and the coherent scattered
radiation I
coh
are dierent functions of the total mass-attenuation coecient of the
incident radiation jE
0
; the intensity I
Com
is a linearly decreasing function and the
intensity I
coh
appears to be a linearly increasing function of the jE
0
. For a limited
range of jE
0
values, the following simple expression can be used to evaluate jE
0
(Markowicz, 1979):
jE
0
C
1
C
2
m C
3
r C
4
mr 18
where
r
I
Com
I
coh
and C
1
C
4
are constants calculated by the least-squares t on the basis of experimental
results for standard samples.
The values of the total mass-attenuation coecient of the uorescent radiation in a
whole sample, jE
i
, is obtained from the simple dependence of the jE
i
,jE
0
ratio on
the values jE
0
, calculated separately for each element to be determined. Finally, sample
weighing provides the value of mass per unit area, m, and the calculation of the absorption
correction factor Ab
corr
via Eq. (4) can be simply performed if, of course, the values of the
eective angles
1
and
2
are evaluated experimentally or theoretically. The applicability
of the proposed matrix correction method (Markowicz, 1979) involving both incoherent
and coherent scattered primary radiations is limited to XRF analysis of intermediate-
thickness samples of mass per unit area smaller than about 10m
thin
.
A backscattered fundamental parameters (BFP) method for quantitative XRF
analysis of intermediate samples of variable composition and thickness was developed by
Nielson (1977). The method utilizes thin-lm multielement calibration of the spectrometer
and mathematical matrix correction in which the samples are modeled as a composite of
heavy elements, which are quantied through their characteristic radiation, and light
elements, estimated through the coherent and incoherent x-ray scatter peaks. Figure 6
schematically illustrates the basis for analyzing the heavy elements (Z13) and the light
elements (H, C, N, O, Na, and others), which must be estimated by the dierence from the
scattered x-ray peaks. The BFP method utilizes coherently and incoherently scattered
x-rays to identify and estimate the quantities of two light elements representative of the
light-element portion of sample matrix. The quantities of the two light elements a and b
are estimated by solving for W
a
and W
b
in the simultaneous equations
gI
coh
X
n
0
j1
W
j
o
cohj
E
0
W
a
o
coha
E
0
W
b
o
cohb
E
0
19
and
hI
Com
X
n
0
j1
W
j
o
Comj
E
0
W
a
o
Coma
E
0
W
b
o
Comb
E
0
20
Copyright 2002 Marcel Dekker, Inc.
where g and h are the geometry-dependent calibration factors determined experimentally
by using any standard of known total composition. Because several light-element pairs
may satisfy Eqs. (19) and (20), the pair is chosen whose incoherent,coherent scattering
cross-sectional ratios lie immediately on either side of the ratio of the observed scatter
attributable to light elements (Nielson, 1977):
o
Coma
E
0
o
coha
E
0
<
hI
Com
P
n
0
j1
W
j
o
Comj
E
0
gI
coh
P
n
0
j1
W
j
o
cohj
E
0
<
o
Comb
E
0
o
cohb
E
0
21
The heavy- and light-element concentrations are used in computing the absorption
correction factor Ab
corr
[Eq. (4)] and the enhancement correction factor [Eq. (91) in
Chapter 1]. Because the concentrations and corrections are interdependent, all calculations
are carried out by iteration [for more details see (Nielson, 1977)].
To improve the sensitivity of the determination of low-Z elements, Sanders et al.
(1983) extended the previously described BFP method. The new method utilizes the co-
herent and incoherent backscatter intensities to compute matrix corrections (Nielson and
Sanders, 1982) from the combined results of two separate energy-dispersive XRF
(EDXRF) data from dierent (TiKc and ZrKc) excitation sources. The Ti-excited spec-
trum allows a more sensitive determination of elements in the AlCa range.
The coherent,incoherent scatter ratio is also applied in an absorption correction
procedure developed by Van Dyck and Van Grieken (1980) for monochromatic x-ray
excitation. In this method, coherent and incoherent scattered radiations are used to cal-
culate, rst, the eective mass of the sample and, second, the absorption coecients for
x-rays of interest and, hence, the absorption correction factors. The eective thickness is
the sample thickness weighted at every point for the excitation-detection eciency, in the
same way as the measured characteristic radiation is weighted.
Figure 6 Light-element contributions to x-ray scattering, from which absorption corrections are
computed. (From Nielson, 1977. Reprinted with permission from Analytical Chemistry. Copyright
American Chemical Society.)
Copyright 2002 Marcel Dekker, Inc.
Assuming that the major elements of the sample do not dier too greatly in atomic
number, the eective thickness m
e
can be calculated from (Van Espen et al., 1979)
m
eff
I
Com
b
0
Ab
corr
E
Com
b
0
I
coh
Ab
corr
E
Com
,a
0
I
Com
Ab
corr
E
coh
b
1
f ga
1
b
1
1
22
where a
0
. a
1
. b
0
. and b
1
are experimental constants obtained by tting the results of
measured standards; for mixtures or compounds, the coherent and incoherent scatter
factors S
coh
and S
Com
, (in fact, the relevant mass-scattering coecients), are given by
S
coh
a
0
X
n
j1
W
j
Z
a
1
j
and
S
Com
b
0
X
n
j1
W
j
Z
b
1
j
23
respectively. Ab
corr
(E
Com
) and Ab
corr
(E
coh
) are the absorption correction factors for the
incoherent and coherent scatter radiation, respectively, as dened in Eq. (4).
A reasonably accurate eective mass is obtained by modeling the sample as a
composite of high-Z elements, calculated from their characteristics peaks using Eq. (2),
and of a light matrix with mass per unit area, m
low-Z
, evaluated from the coherent and
incoherent scatter peaks after subtraction of the high-Z element contribution. The method
for the determination of the eective thickness allows the analysis of samples of hetero-
geneous thickness and irregular shape.
More recently, Arau jo et al. (1990) have developed a very similar procedure for
eective sample mass assessment of intermediate thickness samples via the coherently and
incoherently scattered radiation, as a rst step in the automated matrix eect evaluation,
for the case of ltered polychromatic continuum excitation with a Rh tube. Validation
with geological standards and deposited slurries on Mylar foils and Nuclepore lters gave
very satisfactory results.
In the method proposed by Van Dyck and Van Grieken (1980), calculation of the
mass-attenuation coecient for x-rays of interest is preceded by an evaluation of the
mass-attenuation coecient m (2.956 keV) at the ArKa energy (2.956 keV). This energy is
preferred because it is at the lower end of the energy range that can safely be used in
conventional EDXRF analysis and because, when working under vacuum, it is situated
in a peak-free part of an XRF spectrum. The value of m (2.956 keV) is derived from the
ratio coherent to incoherent scatter intensities R, based on the relationship (see Fig. 7)
of the calculated mass-attenuation coecient at 2.956 keV versus the measured R ratio:
j2.956 keV g
0
g
1
R g
2
R
2
24
where g
0
. g
1
. and g
2
are constant coecients derived by means of a least-squares t based
on the experimental results with standard samples. To improve the accuracy of the
method, the mass absorption coecient of the low-Z matrix at 2.956 keV, m
low-Z
(2.956 keV), must be calculated from the measured ratio of coherent to incoherent scatter
intensities, corrected for the high-Z elements contribution using their characteristic x-ray
intensities. Through this low-Z matrix contribution at 2.956 keV, the total low-Z ab-
sorption curve is calculated (in full analogy to the low-Z matrix contribution at the PdL
Copyright 2002 Marcel Dekker, Inc.
energy in Fig. 5). Finally, the mass absorption coecients for the dierent x-ray energies E
are calculated by adding the low-Z and high-Z absorption contributions [see the total
absorption cure, Eq. (10) and Fig. 5].
The proposed procedure for the automatic determination of the mass-attenuation
coecient, based on the coherent,incoherent scatter ratio, has several obvious merits
compared with the emissiontransmission absorption correction method. First, there is
no supplementary measurement or work needed [apart from experimentally obtaining
the semiempirical dependence of the j (2.956 keV) on the R ratio (Fig. 7) and sensi-
tivity factors B
i
, Eq. (2), on the basis of thin standard samples], because the additional
information on a sample composition is present in the spectrum itself. Second, because
the energies of coherently and incoherently scattered primary radiation, from which the
information is extracted, are higher than that of the incident radiation used in trans-
mission experiments, secondary eects (i.e., grain size eects, inhomogeneous thickness
of the sample, and irregular sample surface) are less important. A third positive point
Figure 7 Calculated mass absorption coefficient at 2.956 keV versus measured coherent,incoher-
ent scattered intensity ratios for pure elements and compounds (circles) and a graphitesulfur
mixture (crosses). (From Van Dyck and Van Grieken, 1980. Reprinted with permission from
Analytical Chemistry. Copyright American Chemical Society.)
Copyright 2002 Marcel Dekker, Inc.
of the described procedure (Van Dyck and Van Grieken, 1980) is the independence,
within certain limits, of the analytical results on the mass of the specimen. Also, it is
worth emphasizing the capability of determining reasonably accurate mass absorption
coecients when the mean atomic number of the sample varies drastically.
Most of the existing correction procedures in XRF analysis of intermediate-thickness
samples ignore the enhancement eect. It appears, however, that for some special cases,
the enhancement eect should be taken into account. Van Dyck et al. (1986) derived
theoretical formulas for secondary uorescent x-ray intensities in medium-thickness
samples based on the Sherman equations. Their computer routine for enhancement cor-
rections was incorporated into an overall program for evaluation of x-ray spectra and
calculation of x-ray absorption correction factors from scatter peaks.
The methods, based on coherent and incoherent scatter radiation, developed by
Van Dyck and Van Grieken (1980) for overall matrix eect corrections in the case of
monochromatic excitation and by Arau jo et al. (1990) for eective sample mass
determination in case of polychromatic radiation have later been expanded by He and
Van Espen (1991) into a general and versatile procedure for quantitative EDXRF with
polychromatic excitation. The comprehensive method uses the scattered radiation to es-
timate the composition and mass of the low-Z element matrix (with optimal use of the
fundamental parameter approach) and also uses the characteristic x-ray peaks to estimate
the concentrations of the higher-Z elements. An iterative process is then executed. Using
the initially estimated sample mass and composition, the weight fraction of each element is
calculated until the calculated composition converges. The absorption and enhancement
corrections are calculated from the latest iteration. Better sample mass, low-Z matrix
composition, and analyte concentrations are computed after this. This overall procedure is
repeated again until subsequent overall iterations do not yield signicant dierences. The
method is, in principle, applicable to samples of any thickness and composition and to
polychromatic excitation. The procedure has been implemented as part of the popular
software package AXIL-QXAS (Van Espen et al., 1997), which includes spectrum ac-
quisition and spectrum analysis and runs on a PC. Excellent quantitative results have been
obtained with it.
Wegrzynek et al. (1993) have recently developed a direct correction procedure for the
enhancement eect in intermediate thickness samples, which is based on the results of the
emissiontransmission measurements and does not require any iterative calculations. The
enhancement term ENH
i
is calculated for the intermediate-thickness samples from
(Wegrzynek et al., 1993)
ENH
i
1
2j
i
E
0
sin
1
jE
i
sin
2
!
m
& '
1
1
25a
SUM1
X
j.q
t
i
E
q
K
j
q
E
0
W
j
sin
1
jE
0
ln
jE
0
,jE
q
sin
1
1
&
sin
2
jE
i
ln
jE
i
,jE
q
sin
2
1
!'
25b
Copyright 2002 Marcel Dekker, Inc.
SUM2
X
j.q
t
i
E
q
K
j
q
E
0
W
j
sin
1
jE
0
ln
jE
0
,jE
q
sin
1
1
&
sin
2
jE
i
ln
jE
i
,jE
q
sin
2
1
!'
25c
SUM3
X
j.q
(
t
i
E
q
K
j
q
E
0
W
j
sin
1
jE
0
sin
2
jE
i
!
exp
jE
0
m
sin
1
! & (
exp
jE
i
m
sin
2
!'
Ei jE
q
m
exp
jE
0
sin
1
jE
i
sin
2
!
m
& '
sin
1
jE
0
Ei jE
q
jE
0
sin1
!
m
& '
sin
2
jE
i
Ei jE
q
jE
i
sin
2
!
m
& '
sin
1
jE
0
Ei jE
q
jE
0
sin1
!
m
& '
sin
2
jE
i
Ei jE
q
jE
i
sin
2
!
m
& ')'
25d
Eix
Z
x
1
expt dt
t
ln jxj
X
1
n1
x
n
n n!
Exponential integral
0.577215664 . . . Euler constant
where
t
i
E
q
is the photoelectric mass absorption coecient for the ith element at the
energy of the qth characteristic line of the jth enhancing element (E
q
),
jE
q
is the total mass absorption coecient for the sample at the energy E
q
,
K
j
q
E
0
t
j
E
0
1
1
J
j
q
!
.
j
p
j
q
J
j
q
and p
j
q
are the jump ratio for the relevant absorption edge and probability of the
emission of the qth characteristic x-ray of the jth element, respectively.
The calculation of the enhancement eect correction is included in a complete pro-
cedure based on the emissiontransmission measurements (Holynska et al., 1994). In the
rst step, the calibration of the XRFsystemis performed [to nd the values of GI
0
E
0
E
i
],
and sin
1
and sin
2
are determined from the x-ray measurements for single-element
standard samples of known mass per unit area. In the next step, the concentrations of the
elements not enhanced by any other element present in the sample are calculated (with the
absorption eect correction based on the emissiontransmission measurements). In the nal
step, the correction for the enhancement eect is made; the values of the fundamental
parameters are taken from the published tables, whereas the values of jE
0
. jE
i
, and
jE
q
are derived from the results of the emissiontransmission measurements.
The enhancement factor for the intermediate thickness samples, ENH
i
, can also be
assessed by using another simple formula (Markowicz et al., 1992b; Austrian Matrix
Correction Routine, 1990):
ENH
i
X
j
ENH
i
thick
W
j
1 expfjE
j
mg 26
Copyright 2002 Marcel Dekker, Inc.
with (ENH
i
)
thick
is the enhancement factor for the thick sample of the same composition as
the intermediate thickness sample to be analyzed:
ENH
i
thick
1
2
1
1
J
j
.
j
p
j
j
j
E
0
j
i
E
0
j
i
E
j
ln 1
jE
0
, sin
1
jE
j
sin
1
jE
0
ln 1
jE
i
, sin
2
jE
j
sin
2
jE
i
1
!
26a
where p
j
is the transition probability for a given spectral line in an x-ray series of the
enhancing jth element. Also in this case, the emissiontransmission measurements can be
used to support the enhancement eect correction. The enhancement factor is calculated
either for each spectral line of the enhancing element separately or for a weighted average
energy of the whole x-ray series. The dierences in the results obtained by using the two
approaches are practically negligible (Markowicz et al., 1992b). Equation (26), although
very simple, can also be applied in order to (1) dene the region of sample thickness where
the enhancement eect is negligible (or smaller than a certain predened level) and (2) to
determine the minimum sample thickness for which the bulk enhancement correction can
already be used.
IV. QUANTITATIONFORINTERMEDIATE-THICKNESS GRANULAR SPECIMENS
The fundamentals of the quantitative XRF analysis of granular specimens involve typical
problems of radiation physics (i.e., the interaction of photons with a specimen of nite
size).
A. Particle Size Effects in XRFAnalysis of Thin and Intermediate-Thickness
Specimens
The analysis of granular materials by XRF, when traditional methods of fusing or
grinding cannot be used, requires careful consideration of the so-called particle size eects.
These eects exist in XRF analysis of any granular materials, irrespective of the mass per
unit area of the specimen, and may constitute a major source of error in quantitative
analysis. The size of particles aects not only the intensity of characteristics x-rays but the
intensity of both backscattered and transmitted x-rays and low-energy g-radiation as well
(Van Dyck et al., 1980; Berry et al., 1969).
Dierent models have been proposed to account for the inuence of particle size on
the characteristic x-ray intensity leaving the sample (Berry et al., 1969; Claisse and
Samson, 1962; Lubecki et al., 1968; Hunter and Rhodes, 1972; Rhodes and Hunter, 1972;
Hawthorne and Gardner, 1978; Krasnopolskaya and Volkov, 1986). Many of these
models involve relatively complex calculations, particularly when the particle size eects
for thick granular samples must be evaluated. This section is limited to the particle size
eects in the quantitative XRF analysis of thin and intermediate-thickness (monolayer)
samples.
For granular intermediate-thickness samples, the mass per unit area of the element,
i. m
i
, is given by (Rhodes and Hunter, 1972)
Copyright 2002 Marcel Dekker, Inc.
m
i
I
i
E
i
B
i
F
i
27
where B
i
is the sensitivity factor as in Eq. (1) and F
i
is the heterogeneity factor which for a
certain discrete particle size is dened by
F
i
1 expj
f
E
0
csc
1
j
f
E
i
csc
2
a
r
f
E
0
csc
1
j
f
E
i
csc
2
a
r
28
where j
f
E
0
and j
f
E
i
are the linear attenuation coecients (cm
1
) for primary and
uorescent radiation in uorescent particles, respectively, and a
r
is the radiometric particle
diameter.
The radiometric diameter, introduced by Claisse and Samson (1962), represents the
mean geometrical path of x-rays through one particle. These and many other authors (Berry
et al., 1969; Lubecki et al., 1968; Hunter and Rhodes, 1972; Rhodes and Hunter, 1972) have
taken a
r
as simply equal to the volume of the grain divided by the particle area presented to
the radiation, averaged over all possible orientations of the grain. Hence, for spherical
particles, the radiometric diameter is equal to 0.67a (with a geometric diameter). One can
easily visualize, however, that such an approach assumes equal weighting of the con-
tribution from all possible radiative paths to the average. In view of the dierent absorption
eects themselves, this is untrue. Markowicz et al. (1980) introduced, instead of the com-
monly used radiometric diameter approach, the concept of an eective absorption-weighted
radiometric diameter for uorescent radiation, depending on both the geometry and ab-
sorption eects, and provided a comparison of these two approaches for single spherical
particles for two excitation-detection geometries ( and /2). This alternative approach
allows quantitative evaluation of the discrepancies resulting from the concept a
r
particle
volume,average area. It was concluded that for the geometry, both approaches give
practically the same results (maximum relative dierences amount only to 5%), but for /2
geometry, the radiometric diameter approach can safely be applied only for very small
particles and,or at relatively high energies of primary radiation (Markowicz et al., 1980).
For samples with a certain particle size distribution described by a function fa
r
, the
heterogeneity factor F
i
can be calculated according to the formula (Rhodes and Hunter,
1972)
F
i
Z
a
r max
a
r min
fa
r
1 expa
r
j
f
E
0
csc
1
j
f
E
i
csc
2
da
r
j
f
E
0
csc
1
j
f
E
i
csc
2
a
r
29
where a
r min
and a
r max
are the smallest and the largest radiometric particle diameters,
respectively, and
fa
r
dV
f
V
f
t
da
r
30
where dV
f
is the volume of uorescent particles having a size between a
r
and a
r
da
r
and
V
f
t
is the total volume of the uorescent particles.
The theoretical predictions given by Eqs. (28) and (29) have been compared with the
experimental results for the heterogeneity factors obtained for samples with some discrete
particle sizes and various particle size distributions (Holynska and Markowicz, 1981). For
granular samples of copper sulde, a satisfactory agreement has been obtained; for samples
of iron oxide, some discrepancies due to agglomeration of the particles have been observed.
Copyright 2002 Marcel Dekker, Inc.
An in-depth study of the inuence of sample thickness, excitation energy, and
excitation-detection geometry on the particle size eects in XRF analysis of intermediate-
thickness samples was carried out and the results are presented in the work of Van Dyck
et al. (1985).
B. Correction Methods for the Particle Size Effect in XRFAnalysis
of Intermediate-Thickness Specimens
As already mentioned, the heterogeneity factor F
i
describing the magnitude of the particle
size eect in XRF analysis of thin and monolayer samples can simply be calculated if the
particle size or the function of the particle size distribution is known. This occurs in XRF
analysis of air particulates (or aerosols), for example, when special sampling techniques,
involving a cascade impactor, are applied. At dierent stages of the cascade impactor, the
particles of denite sizes are collected (Katz, 1977) and calculation of the heterogeneity
factor F
i
may be straightforward if, of course, the kind of uorescent particles is known.
A simple empirical particle size correction factor 1 ba
2
was proposed by Criss
(1976), in which a is the particle diameter b is a coecient that depends on particle
composition and experimental conditions. The author has provided a table of the values
for b for the determination of 48 dierent elements in 200 compounds using either a Cr- or
W-target x-ray tube.
Another correction for the particle size eect based on the model of Berry et al.
(1969), also requiring evaluation of the particle size in a sample, is due to Nielson (1977).
All the particle size corrections mentioned are of limited applicability because eva-
luation of a particle size or particle size distribution function must be done before XRF
analysis. Moreover, in many cases it may be necessary to consider how the indirectly
determined particle sizes relate to true sizes. However, even when there is some uncertainty
in the sizes and compositions of the particles, which are input parameters in the particle
size correction, it is better to make an appropriate correction than no correction at all.
1. Empirical Particle Size Correction Method Using Dual Measurements
To overcome the problems encountered when calculated particle size correction factors are
applied in the quantitative XRF analysis of granular samples, a particle size correction
method based on dual measurements of characteristic x-rays excited by x- or g-radiation of
two dierent energies was developed (Holynska and Markowicz, 1982). The method utilizes
the dierence in the particle size eect for two excitation energies and oers the possibility
of experimental detection and correction of this eect. In general, these two excitation
energies should be chosen so that the eect of particle size is small for one of them and large
for the other. Thus, the measured intensities of the characteristic x-rays of the element to be
determined are dierent functions of the particle size. The ratio of these two intensities
is sensitive to the particle size and it can be used for obtaining the particle size correction
factor K
i
1,F
i
. First, a calibration curve giving the relationship between the correction
factor K
i2
and the ratio I
i1
,I
i2
rel
must be plotted. The ratio I
i1
,I
i2
rel
is given by
I
i1
I
i2
rel
I
i1
,I
i2
I
i1
,I
i2
hom
31
where I
i1
,I
i2
is the ratio of the intensities of the characteristic x-rays of the ith element
excited in a granular sample by primary x-rays of two dierent energies (indexes 1 and 2,
respectively) and I
i1
,I
i2
hom
is the same for a thin homogeneous sample. Taking into
Copyright 2002 Marcel Dekker, Inc.
account that in the thin-sample technique, interelement eects may be neglected, single-
element standard samples can be used to obtain the I
i1
,I
i2
hom
ratio. The calibration curve
mentioned earlier can be obtained either theoretically or experimentally with the use of
calibration samples of known discrete particle size fractions. It has been shown (Holynska
and Markowicz, 1982) that there is also a possibility of applying such a calibration curve
for granular samples with various particle size distribution functions. For a p geometry,
the calibration curves for the determination of the correction factor K
i2
are described by
the equation (Markowicz, 1983)
K
i2
I
i1
I
i2
rel
t
0
a
r
1 e
t
0
a
r
32
where
t
0
j
f
E
01
j
f
E
i
f
E
02
j
f
E
i
33
j
f
E
02
j
f
E
i
34
j
f
E
01
and j
f
E
02
are the linear absorption coecients for primary radiation of two
dierent energies (indexes 1 and 2, respectively) in uorescent particles (cm
1
). The par-
ticle size correction factor K
i2
is given by
K
i2
a
r
1 e
a
r
35
The proposed method of particle size correction was veried experimentally for the
determination of copper, applying
238
Pu and
241
Am radioisotopes as sources of primary
radiation, and a satisfactory agreement between theoretical predictions and experimental
results was reported (Holynska and Markowicz, 1982).
2. Applicabilityof the Particle Size Correction Method
Accurate determination of the correction factor K
i2
is mainly aected by uctuations
resulting from counting statistics in all measured intensities of the characteristic radiation.
The absolute error K
i2
in determining the correction factor K
i2
can be calculated from
the formula
K
i2
t
0
a
r
1 e
t
0
a
r
S
c
36
where S
c
is standard deviation resulting from counting statistics for the ratio I
i1
,I
i2
rel
.
The relative error in determining the particle size correction factor K
i2
is given by
K
i2
K
i2
S
c
I
i1
,I
i2
rel
S
r
37
where S
r
is relative standard deviation resulting from counting statistics for the ratio
I
i1
,I
i2
rel
.
Equations (32) and (37) enable us to estimate the applicability of the particle size
correction method. This can be done with the aid of Figure 8. From two families of curves
in Figure 8, one can estimate the maximum value of the particle size correction factor
K
i2
max
that can be determined with the particle size correction method for given values of
S
r
and t
0
. Figure 9 presents the theoretical relationship of the maximum value of the
correction factor K
i2
max
with the parameter t
0
for dierent values of the relative standard
Copyright 2002 Marcel Dekker, Inc.
deviation S
r
; Figure 10 shows the theoretical relationship of K
i2
max
with S
r
for dierent
values of t
0
. The families of curves shown in Figures 9 and 10 allow us to determine the
application limits of the particle size correction method [i.e., to determine K
i2
max
in
various congurations]. In practice, however, it is more interesting to know the maximum
value of the radiometric particle diameter for which the particle size eect can still be
corrected. This can be determined from Eq. (32).
3. Selectionof the OptimumMeasurement Conditions
When the ith element is present in a given chemical compound and the maximum values of
the radiometric diameter of the particle in a sample are known, it is possible to dene the
optimum value t
0
opt
of the parameter t
0
for a given energy of exciting radiation E
02
.
Figure 8 Theoretical relationship of the particle size correction factor K
i2
with the ratio of the
intensities I
i1
,I
i2
rel
for different values of the parameter t
0
(family curves) and with the error K
i2
for different values of the relative standard deviation S
r
(family of straight lines). (From Markowicz,
1983. Reprinted by permission of John Wiley & Sons, Ltd.)
Copyright 2002 Marcel Dekker, Inc.
The value of t
0
opt
can be determined from Figure 9 for a given value of the relative standard
deviation S
r
. Thus, one can obtain the following inequality, which should be fullled by
the parameter t
0
:
t
0
t
0
opt
38
This means that the value of the energy of the exciting radiation E
01
of the correction
source should fulll the inequality
E
01
! E
01
opt
39
where E
01
opt
is the energy of the exciting radiation E
01
for which t
0
t
0
opt
. Taking into
account that the eciency of photoexcitation depends on the energy of the primary
Figure 9 Theoretical relationship of the maximum value of the particle size correction factor
K
i2
max
with the parameter t
0
for different values of the relative standard deviation S
r
. t
0
opt
is the
optimum value parameter t
0
(see text for details). (From Markowicz, 1983. Reprinted by permission
of John Wiley & Sons, Ltd.)
Copyright 2002 Marcel Dekker, Inc.
radiation, the practical conclusion can be drawn that the values of both the parameter t
0
and the energy of the exciting radiation E
01
should be as close as possible to the values of
t
0
opt
and E
01
opt
, respectively.
On the other hand, for a given pair of sources of primary radiation (i.e., for a given
value of the parameter t
0
), it is possible to estimate, using Figure 10, the maximum ac-
ceptable value S
r
0
of the relative standard deviation of the ratio I
i1
,I
i2
rel
. In con-
sequence, the appropriate measurement times and,or activities of the sources of primary
radiation can be selected.
Although the correction method may look complicated, it is currently the only
real correction procedure dealing with the particle size eect in XRF analysis of granular
intermediate-thickness specimens. The idea of applying dierent excitation x-ray energies to
estimate particle size corrections was also exploited by Nielson and Rogers (1986).
Figure 10 Theoretical relationship of the maximum value of the particle size correction factor
K
i2
max
with the relative standard deviation S
r
for different values of the parameter t
0
. S
r
0
is the
maximum acceptable value of the relative standard deviation S
r
(see text for details). (From
Markowicz, 1983. Reprinted by permission of John Wiley & Sons, Ltd.)
Copyright 2002 Marcel Dekker, Inc.
REFERENCES
Andermann G, Kemp JW. Anal Chem 30:1306, 1958.
Arau jo MF, Van Espen P, Van Grieken R. X-ray Specrom 19:29, 1990.
Austrian Matrix Correction Routine, Description of the Program, Technical University of Graz,
Austria, 1990.
Bazan F, Bonner NA. Adv X-ray Anal 19:381, 1976.
Berry PF, Furuta T, Rhodes JR. Adv X-ray Anal 12:612, 1969.
Claisse F, Samson C. Adv X-ray Anal 5:335, 1962.
Criss JW. Anal Chem 48:179, 1976.
Giauque RD, Garrett RB, Goda LY. Anal Chem 51:511, 1979.
Giauque RD, Goulding FS, Jaklevic JM, Pehl RH. Anal Chem 45:671, 1973.
Hawthorne AR, Gardner RP. X-ray Spectrom 7:198, 1978.
He F, Van Espen PJ. Anal Chem 63:2237, 1991.
Holynska B, Markowicz A. X-ray Spectrom 8:2, 1979.
Holynska B, Markowicz A. X-ray Spectrom 10:61, 1981.
Holynska B, Markowicz A. X-ray Spectrom 11:117, 1982.
Holynska B, Ptasinski J, Wegrzynek D. Appl Radiat Isot 45:409, 1994.
Hunter CB, Rhodes JR. X-ray Spectrom 1:107, 1972.
Kalman ZH, Heller L. Anal Chem 34:946, 1962.
Katz M (ed), Methods of Air Sampling and Analysis. Washington, DC: American Public Health
Association, 1977, p 592.
Kieser R, Mulligan TJ. X-ray Spectrom 8:164, 1979.
Kikkert J. Adv X-ray Anal 26:401, 1983.
Krasnopolskaya NN, Volkov VF. X-ray Spectrom 15:3, 1986.
Leonardo L, Saitta M. X-ray Spectrom 6:181, 1977.
Leroux J, Mahmud M. Anal Chem 38:76, 1966.
Livingstone LG. X-ray Spectrom 11:89, 1982.
Lubecki A, Holynska B, Wasilewska M. Spectrochim Acta B23:465, 1968.
Markowicz A. X-ray Spectrom 8:14, 1979.
Markowicz A. X-ray Spectrom 12:134, 1983.
Markowicz A. X-ray Spectrom 13:166, 1984.
Markowicz A, Abdunnabi AA. X-ray Spectrom 20:97, 1991.
Markowicz A, Haselberger N. Appl Radiat Isot 43:777, 1992.
Markowicz A, Haselberger N, Mulenga P. X-ray Spectrom 21:271, 1992a.
Markowicz A, Van Dyck P, Van Grieken R. X-ray Spectrom 9:52, 1980.
Markowicz A, Haselberger N, El Hassam HS, Sewando MSA. J Radioanal Nucl Chem 158:409,
1992b.
McMaster WH, Delgrande M, Mallett JH, Hubbell JM. University of California. Lawrence
Radiation Laboratory Report, UCPL-50174, 1969.
Meier H, Unger E. J Radioanal Chem 32:413, 1976.
Nielson KK. Anal Chem 49:641, 1977.
Nielson KK, Rogers VC. Adv X-ray Anal 27:449, 1984.
Nielson KK, Rogers VC. Adv X-ray Anal 29:587, 1986.
Nielson KK, Sanders RW. The SAP3 Computer Program for Quantitative Multielement Analysis by
Energy-Dispersive X-ray Fluorescence, US DOE Report PNL-4173, 1982.
Nielson KK, Sanders RW, Evans JC. Anal Chem 54:1782, 1982.
Rhodes JR, Hunter CB. X-ray Spectrom 1:113, 1972.
Sanders RW, Olsen KB, Weiner WC, Nielson KK. Anal Chem 55:1911, 1983.
Taylor DL, Andermann G. Anal Chem 43:712, 1971.
Taylor DL, Andermann G. Appl Spectrosc 27:352, 1973.
Van Dyck PM, Van Grieken RE. Anal Chem 52:1859, 1980.
Van Dyck P, Markowicz A, Van Grieken R. X-ray Spectrom 9:70, 1980.
Copyright 2002 Marcel Dekker, Inc.
Van Dyck P, Markowicz A, Van Grieken R. X-ray Spectrom 14:183, 1985.
Van Dyck PM, To ro k SB, Van Grieken RE. Anal Chem 58:1761, 1986.
Van Espen P, Janssens K, Nobels J. J Chemom Intell Lab Syst 1:109, 1997.
Van Espen P, Vant dack L, Adams F, Van Grieken R. Anal Chem 51:961, 1979.
Wegrzynek D, Holynska B, Pilarski T. X-ray Spectrom 22:80, 1993.
Yap CT, Kump P, Tang SM, Wijesinghe L. Appl Spectrosc 41:80, 1987.
Copyright 2002 Marcel Dekker, Inc.
7
Radioisotope-Excited X-ray Analysis
Stanislaw Piorek*
Niton Corporation, Billerica, Massachusetts
I. INTRODUCTION
Radioisotope x-ray uorescence (XRF) and x-ray preferential absorption (XRA) techni-
ques are used extensively for the analysis of materials, covering such diverse applications
as analysis of alloys, coal, environmental samples, paper, waste materials, and metalli-
ferous mineral ores and products (Rhodes, 1971; Rhodes and Rautala, 1983; Watt, 1978
and 1983; Watt and Stener, 1985; Piorek 1997). Many of these analyses are undertaken in
the harsh environment of industrial plants and in the eld. Some are continuous on-line
analyses of material being processed in industry, where instantaneous analysis information
is required for the control of rapidly changing processes.
Radioisotope x-ray analysis systems are often tailored to a specic but limited range
of applications. They are similar and often considerably less expensive than analysis
systems based on x-ray tubes, but these attributes are often gained at the expense of
exibility of use for a wide range of applications.
Operators making analyses in the eld or in industrial plants are usually less skilled
than those working in the laboratory with x-ray tube systems. Manufacturers of radio-
isotope x-ray analysis systems compensate for this by producing simple semiautomated or
fully automated systems whose output, calibrated for the specic application, is given
directly in terms of concentrations of elements required or in terms of a simple pass=fail-
type decision.
Radioisotope x-ray techniques are preferred to x-ray tube techniques when simpli-
city, ruggedness, reliability, and cost of equipment are important, when minimum size,
weight, and power consumption are necessary, when a very constant and predictable x-ray
output is required, when the use of high-energy x-rays is advantageous, and when short
x-ray path lengths are required to minimize the absorption of low-energy x-rays in air.
Also of signicant analytical importance is the fact that the radioisotope excitation is
usually monoenergetic (monochromatic) as opposed to polyenergetic (polychromatic)
excitation characteristic for x-ray tubes.
*Previous aliation: Metorex Inc., Princeton, New Jersey.
Copyright 2002 Marcel Dekker, Inc.
X-ray uorescence techniques based on the x-ray tubeBragg crystal spectrometer
are considerably more sensitive than those based on radioisotope sources. This high
sensitivity is due to the excellent x-ray resolving power of the crystal spectrometer, which is
superior to that of the gas-lled or solid-state detector typically used with radioisotopes.
Radioisotopes cannot be used with crystal spectrometers because of the low geometrical
eciency of this spectrometer coupled with the fact that the x-ray photon output of
radioisotope sources is relatively very low, about six orders of magnitude less than that of
x-ray tubes used with crystal spectrometers.
Presently, the bulk of use of XRF analysis is in portable and on-line equipment.
Portable systems outnumber on-line installation, and the gap increases from year to year.
A signicant number of radioisotope benchtop XRF systems have been installed for
quality control applications, as o-line auxiliary instruments. These, however, are very
quickly yielding the eld to XRF systems based on low-power x-ray tubes.
For some applications, x-ray preferential absorption (XRA) and x-ray scattering
(XRS) techniques are preferred to XRF techniques, particularly when coarse particulate
material is to be analyzed. Radioisotopes are the only practical source of x-rays for these
applications, because to penetrate deep into the material, high energy, usually above
100 keV, x-rays are required. The most important applications of XRA and XRS tech-
niques are the on-line analysis of particulate material on conveyors.
Some of the terminology used in this chapter is now briey dened. The element
whose concentration in the sample is to be determined is the analyte and the other ele-
ments of the sample are the matrix elements, or simply the matrix.
Sensitivity of an analytical method for a given analyte is dened as net change (in-
crease) of the measured signal of the analyte per unit concentration change (increase) of
that analyte. This term is notoriously, but wrongly, used in place of minimum detectable
level. A minimum detectable level (DL) for a given analyte is understood as that amount of
analyte in a sample that produces spectral signal equal to or greater than three standard
deviations of a signal obtained on a sample with no analyte present. The DL improves
when sensitivity of analysis improves.
The common link between all techniques and applications discussed here is the de-
pendence of the analysis primarily on the absorption of x- and g-rays. Compton and
coherent scattering are the other important interactions taking place in the sample. The
terms x-ray and g-ray can often be used interchangeably. The term x-ray is always used
when discussing uorescent x-rays. Radioisotope sources emit either g-rays directly from
the nucleus or uorescent x-rays emitted following the ejection of an atomic electron.
g-Rays emitted by radioisotopes usually have energies greater than 50 keV. X-ray uor-
escence analysis depends on both x-ray and g-ray excitation, but most XRA and XRS
analyses are based on the use of g-rays. The term high-energy g-ray is used when the
g-ray interaction in the sample is essentially entirely due to Compton scattering (typically
above 300 keV) which is being used to determine either the bulk density or mass per unit
area of the sample. The term low-energy g-ray is used when photoelectric interactions
are important to the analysis.
This chapter reviews radioisotope-excited x-ray uorescence, preferential absorp-
tion, and scattering techniques. The characteristics of radioisotope sources and x-ray
detectors are described, and then the x-ray analytical techniques are presented. The choice
of radioisotope technique for a specic application is discussed along with major factors
aecting the overall accuracy of analysis. This is followed by a summary of applications of
these techniques, with a more detailed account given of some of the most representative
applications, particularly those of considerable industrial importance.
Copyright 2002 Marcel Dekker, Inc.
II. BASICEQUATIONS
The basic equations for x-ray analysis are given in Chapter 1. Some additional equations
used for XRF, XRA, and XRS analyses are presented here. The typical geometries (Watt
and Stener, 1985; Jenkins et al., 1981 b) of the radioisotope source, sample, and detector
used are shown in Figures 1 and 2.
Figure 1 The three geometries for radioisotope-excited x-ray fluorescence analysis: (a) annular
source, (b) central source, and (c) side source. (From Jenkins et al., 1981b.)
Copyright 2002 Marcel Dekker, Inc.
A. Absorption of X-rays
The intensity, I, of a narrow beam of monoenergetic x- or g-rays transmitted through a
sample (already shown in Chapter 1) is given by
I I
0
e
mrt
1
where
m
X
n
i1
W
i
m
i
2
Figure 2 Arrangement of radioisotope g-ray source, collimators, and scintillation detector used in
x-ray preferential absorption analysis. (From Watt and Steffner, 1985.)
Copyright 2002 Marcel Dekker, Inc.
and
X
n
i1
W
i
1 3
I
0
is the intensity of x-rays detected without the sample. m and t are the mass absorption
coecient and path length of x-rays in the sample, respectively, r is the bulk density of the
sample, and m
i
and W
i
are the mass absorption coecient and weight fraction of the ith
element in the sample, respectively.
Equation (1) also holds for broad beams of x-rays when the cross section for pho-
toelectric absorption is much greater than that for Compton and coherent scattering [i.e.,
particularly for low-energy x-rays and high-atomic-number (Z) elements]. This assump-
tion can be veried by reference to Appendices VIVIII of Chapter 1.
B. Fluorescent X-ray Intensity
When a monoenergetic beam of x-rays excites the K shell x-rays of the analyte i in an
innite-thickness sample and both the incident and emitted x-rays are normal to the
sample surface, the detected intensity I
i
of the Ka x-rays of analyte i is given approximately
by Eqs. (94) and (90) in Chapter 1, namely
I
i
GeE
i
a
i
E
0
I
0
E
0
mE
0
mE
i
4
where
G Geometrical constant
eE
i
Intrinsic eciency of the detector to the x-rays of the analyte i
a
i
E
0
W
i
t
0
i
E
0
o
i
r
i
1 1=j
i
I
0
E
0
The source emission (photons=s)
mE
0
; mE
i
Mass absorption coecients for the exciting radiation with energy
E
0
and the characteristic radiation with energy E
i
, respectively, in
the sample (cm
2
/g)
t
0
i
E
0
Total photoelectric mass absorption coecient for the ith element
at energy E
0
(cm
2
/g)
o
i
The K shell uorescent yield for the analyte i
r
i
Relative transition probability for Ka lines of analyte i
j
i
Jump ratio
Enhancement eects (discussed in Chapter 5, Sec. II.B.2) have been assumed to be
negligible.
The intensities of L and M shell uorescent x-rays can be calculated from equations
similar to Eq. (4). For radioisotopes emitting x-rays of more than one energy, I
i
can be
separately calculated for each emitted energy and the total uorescent x-ray intensity
determined by summing the products of I
i
and the probability of emission of the given
x-ray energy from the radioisotope.
C. Scattered X-ray Intensities
X-rays are scattered from the sample and its surroundings to the detector by a mechanism
of coherent and Compton (or incoherent)-scattering interactions. There is no loss of
Copyright 2002 Marcel Dekker, Inc.
energy of incident radiation during coherent scattering. The energy E of the Compton-
scattered x-ray is given by [Eq. (52) of Chapter 1]
E
E
0
1 g1 cos y
5
where E
0
is the energy (in the units of keV) of the incident x- or g-ray, g E
0
=511, and y
is the scattering angle. The scattering angle y, measured from the direction of the incident
x-ray, in most radioisotope XRF systems (Fig. 1) is in the range 90
150
. The loss in
energy of the incident x-ray due to Compton scattering at 90
, 120
, and 150
is shown in
Figure 3 and can be seen to be relatively very small at energies below 20 keV. The de-
tected intensity I
s
of x-rays scattered from an innitely thick sample to the detector is
given by
I
s
GI
0
E
0
T
s
e
s
P
m
si
yW
i
P
m
i
m
si
W
i
6
where G; I
0
E
0
, and W
i
are the same as in Eq. (4), T
s
is the transmission of the scattered
x-rays through the lter and the detector window; e
s
is the eciency of the detector for the
scattered x-rays; m
si
y is the dierential scattering cross section for the x-rays scattered by
the ith element toward the detector, and m
si
is the mass absorption coecient of the
scattered x-rays for the ith element of the sample.
Equation (6) holds for both coherent and Compton scattering when the appropriate
scattering cross section [Eqs. (73) and (65) of Chapter 1] is used. It assumes that the
photoelectric absorption cross section of the x-rays in the sample is much greater than
that for scattering and that incident and emergent x-rays are normal to the sample
surface.
The dierential and total Compton-scattering cross sections per atom are propor-
tional to Z=A, where A is the atomic weight of the atom, and, except for hydrogen, are
almost independent of the atomic number of the atom. Hence, the scattered intensity is
approximately inversely proportional to the sum of the mass absorption coecients in the
sample of the incident and emerging x-rays.
Figure 3 Loss in energy of x-rays in Compton scattering at angles 90
, 120
, and 150
.
Copyright 2002 Marcel Dekker, Inc.
The cross section for coherent scattering is highest for small scattering angles, low-
energy x-rays, and high-atomic-number atoms. For angles greater than 90
, the cross
section is low and varies by only a factor of about 2.
D. X-ray Fluorescence Analysis
The concentration of the analyte in the sample is determined from measurement of the
intensity of its uorescent x-rays, often combined with measurements of the intensities of
the uorescent x-rays of matrix elements and=or the Compton-scattered x-rays. Often the
denominator of Eq. (4) is proportional to that of Eq. (6); that is,
mE
0
mE
i
/
X
m
i
m
si
W
i
7
Hence, Eqs. (4) and (6) can be combined to give
W
i
k
I
i
I
Com
8
where k is a constant. Note that the intensity of scattered radiation, I
s
, is replaced here by
I
Com
, because Eq. (6) holds for both coherent and Compton scattering when the appro-
priate scattering cross section is used. If a major matrix constituent, a, has an absorption
edge energy between the energy of the characteristic x-rays of the analyte and the energy of
the incoherently (Compton) scattered primary radiation, then
W
i
kI
i
1
I
Com
k
1
m
a
W
a
9
where k
1
is a constant and the subscript a refers to the major matrix constituent. Equa-
tions (8) and (9) can be checked for accuracy for any specic application by substituting
mass absorption coecients (see Appendices VIIIX of Chapter 1) and elemental con-
centrations into Eqs. (4) and (6).
Relation (8) has a great analytical signicance; in many applications, it linearizes the
calibration curve for a given analyte, thus making analysis more accurate and robust.
Analysis of Eqs. (4) and (3) reveals that the intensity of the characteristic x-rays of a
given analyte is a function not only of this analyte concentration but also of the con-
centrations of all the other elements in the sample. This poses diculty in solving this
equation directly for W
i
. However, by adopting, for example, a simple approximation
proposed by Lucas-Tooth and Price (L-TP) (1962), this problem can be solved. The L-TP
approximation states that because the x-ray intensities of elements are functions of their
respective concentrations, one can substitute their measured x-ray intensities for con-
centrations of matrix elements. In this way, Eq. (4), via Eq. (3), can be solved for W
i
,
which is now expressed, for each analyte, by all measured x-ray intensities of the elements
in the sample:
W
i
b
1
I
i
k
0
X
jn
j6n
k
ij
I
j
!
10
The I s are the x-ray intensities of the elements measured, k
ij
are correction coecients
determined by multivariable linear least-squares tting, and W
i
is concentration of the
analyte i. There are as many of this equations as there are analytes to be measured.
The L-TP model allows for the calibration of an x-ray analyzer with the suite of
calibration samples without the necessity of knowing assays for any element but the
Copyright 2002 Marcel Dekker, Inc.
desired analyte. Due to its simplicity and ruggedness, this approach is often used in the
calibration of benchtop and portable x-ray analyzers; the analyzer measures x-ray in-
tensities of the important interfering elements in calibration samples and then develops a
calibration equation for the analyte(s) by multivariable, linear least-squares tting.
E. X-ray Preferential Absorption Analysis
X-ray preferential absorption analysis is based on the measurement of the intensities of
two or more monoenergetic x-rays transmitted through the sample (Fig. 2). Sensitivity of
analysis depends on the selective absorption of the transmitted x-rays by the analyte
compared with absorption by the sample matrix. The greater the dierence (or contrast)
between the mass absorption coecients of the analyte and matrix for the transmitted
x-rays, the better the sensitivity of analysis for that analyte. From Eqs. (1)(3), the
concentration of the analyte is given by
W
i
lnI
0
=I=rt m
M
m
i
m
M
11
where m
i
is the mass absorption coecient of the x-rays in the analyte and m
M
is the ab-
sorption coecient of the matrix, given by
m
M
X
n
j6i
m
j
W
j
12
and the subscript j refers to the matrix elements and SW
j
1 W
i
. The concentration of
the analyte can thus be determined if the product of the bulk density and thickness of the
sample is known and the mass absorption coecients of the matrix elements are ap-
proximately equal or the composition of the matrix does not vary.
In practice, XRA analysis usually involves measurements of transmission of narrow
beams of x-ray, at two x-ray energies, through the sample (Watt and Stener, 1985). This
is called dual-energy (x- or g-ray) transmission (DUET) analysis. The beams are usually
coincident and thus not dierently aected by heterogeneity of the sample in the beam
path (Fig. 2).
From Eq. (11), the concentration W
i
of the desired analyte is
W
i
m
0
M
Rm
00
M
m
0
i
m
0
M
Rm
00
i
m
00
M
13
where
R
lnI
0
=I
0
lnI
0
=I
00
14
and
R
m
0
M
m
0
i
m
0
M
W
i
m
00
M
m
00
i
m
00
M
W
i
m
0
m
00
15
The prime and double prime refer to the rst and second x-ray energies, respectively. The
concentration is thus determined independently of the density and thickness of the sample
through which the coincident x-ray beams pass.
Copyright 2002 Marcel Dekker, Inc.
The sensitivity of analysis is high when m
0
i
)m
0
M
and when rt is large. The analysis is
accurate when the ratio m
0
M
=m
00
M
is constant, independently of variations in composition of
the sample matrix. This ratio is approximately constant when the x-ray energies of the
transmitted x-ray are just above and below the K shell absorption edge energy of the
analyte and at higher energies when, at each x-ray energy, the mass absorption coecients
of all matrix elements are about the same. In the latter case, the energy of the higher-
energy x-ray is usually chosen so that the mass absorption coecients of the analyte and
matrix elements are the same, the transmission measurement thus determining the mass
per unit area of sample in the x-ray beam.
The uncertainty in determination of R [Eq. (14)] caused by counting statistics is
dR
R
rt
dI=I
0
m
0
!
2
dI=I
00
m
00
!
2
s
16
where
dI
I
0
!
2
dI
0
0
I
0
0
!
2
dI
0
I
0
!
2
17
and
dI
I
00
!
2
dI
00
0
I
00
0
!
2
dI
00
I
00
!
2
18
and where dI=I is the relative counting statistical uncertainty and m is the mass absorption
coecient of the x-ray in the sample. The corresponding uncertainty in determining the
concentration of the analyte can be found by substituting RdR for R in Eq. (13).
The uncertainty in the determination of the concentration of the analyte, caused by
an increase in the concentration (dC) of one matrix element k replacing another matrix
element l, can be calculated by increasing the mass absorption coecient of the sample
matrix by
m
M
new m
M
old m
k
m
l
dC 19
and substituting the new mass absorption coecient into Eq. (13).
These equations accurately predict all aspects of XRA analysis (Watt and Stener,
1985) except when the sample is so highly heterogeneous that within the beam of x-rays,
there are signicant dierences in absorption of the x-rays.
F. X-ray Scattering Analysis
Of the two x-ray scattering methods of analysis, one relies on comparison of the detected
intensities of the Compton-scattered and coherent scattered x-ray (Schatzler, 1979),
whereas the other one is based on determination of the intensity of the Compton-scattered
x-rays (Fookes et al., 1975). The former method is essentially a measure of the ratio of the
dierential scattering cross sections of the two components, which is proportional to
between Z and Z
2
(Siegbahn, 1965). The latter methods depends on the absorption of
x-rays in the sample, which, in the photoelectric region, is proportional to between Z
4
=A
and Z
5
=A (Siegbahn, 1965). Thus, the method is very similar to XRA analysis and is
considerably more sensitive than the Compton-scattered to coherent scattered ratio x-ray
method. Both methods are accurate only when the changes in detected x-ray intensities
Copyright 2002 Marcel Dekker, Inc.
caused by changes in the concentration of the analyte are much greater than those caused
by changes in the concentration of the matrix elements.
The sensitivity of both techniques and uncertainties due to variations in con-
centrations of matrix constituents can be predicted using Eq. (6), where the photoelectric
absorption cross section in the sample is much greater than the scattering cross section.
Dual-energy scattering techniques (Outokumpu Mintec, 1986), analogous to dual-
energy preferential absorption techniques, are used to minimize the eects of sample
heterogeneity. The x-ray scattering techniques are used in applications in which only one
side of the sample is accessible and thickness of the sample is too great to allow sucient
penetration of x-rays. Compared with DUET analysis, the main disadvantage is that
narrow beams of x-rays cannot often be used because of the lower geometrical eciency of
the source, sample, and detector. Hence, multiple scattered x-rays are detected with a
consequent loss in accuracy of analysis.
III. RADIOISOTOPE X-RAYSOURCES ANDDETECTORS
The characteristics of radioisotope x-ray sources and detectors are described here. A full
understanding of the dierent characteristics of scintillation, proportional, and solid-state
detectors is essential because of the need to tailor their use to specic applications and to
environmental conditions in the eld and in industrial plants.
A. Radioisotope Sources
There are only a fewradioisotope sources that are usedfrequently for x-ray analysis; these are
listed with their most important characteristics in Table 1. Also included are two radio-
isotopes that emit high-energy x-rays and they are used most frequently with the x-ray
sources to correct for changes in sample mass per unit area, thickness, or bulk density.
The activity of radioisotopes is specied in terms of the rate of disintegration of the
radioactive atoms [i.e., decays per second, or becquerel (Bq)]. One becquerel (Bq), an SI
unit, is dened as one disintegration per second. The unit of becquerel replaces the non-SI
unit, the curie (Ci), which equals 3.7610
10
becquerel. Unfortunately, the unit of becquerel
is not very practical. For example, a typical, useful activity of 100 mCi has to be expressed
in gigabecquerels (GBq). A practical conversion relation between the two units is
100 mCi 3:7 GBq
The typical number of x- or g-rays emitted per disintegration by the given radioisotope is
listed in Table 1 so that the essential parameter of the radioisotope source, the number of
x-rays or g-rays emitted per second, can be calculated. The emission rate of radioisotope
decreases with time according to the law of natural decay, the number of radioisotope
atoms decaying from N
0
to N after an elapsed time t being given by
N N
0
e
0:693t=T
1=2
20
where T
1=2
is the so-called half-life of the radioisotope. The source decays to half of its
original emission rate during the time equal to its half-life. The radioisotope source is
usually replaced after one to two half-lives.
The physical size of radioisotope x-ray sources is small. Figure 4 shows the en-
capsulations of typical cylindrical and annular sources of
109
Cd (Amersham, 1986).
Cylindrical sources used in portable analyzers are usually 8-mm-diameter by 5-mm-thick
Copyright 2002 Marcel Dekker, Inc.
capsules and are often referred to as pellets. When economics justies it, a special source
may be designed for a particular type of the analyzer. A so-called lollipop source, de-
signed specically for light-element analysis probes made by Metorex Int. is such an ex-
ample (Amersham IEC.600 series). It is made in a form of a at copper ring, 1 mm thick,
15 mm in diameter, with a 8-mm opening. One side of this at copper annulus is elec-
troplated with a
55
Fe isotope, over which a few-micrometer-thick Ni ashing is applied.
Such source geometry allows for a very close coupling between sample and a proportional
detector window. Consequently, the quantitative analysis of light elements, down to A1, is
possible without the nuisance of helium purge or vacuum.
There are international codes for the safe use of radioisotopes, and a simple in-
troduction to radiation protection has been published (Martin and Harbison, 1986). Each
organization using radioactive substances is required to hold a license, issued in most
countries by a government health department or atomic energy authority. The Interna-
tional Commission on Radiological Protection (ICRP, 1985) recommends that, for
members of the public, it would be prudent to limit exposures to radiation on the basis of a
lifetime average annual dose of 1 millisievert (mSv). Table 1 lists the approximate, typical
dose rates at 1 m from each radioisotope source, assuming no absorption of the emitted
radiation within the source or by air. The x-ray dose is inversely proportional to the square
of the distance from the source. X-ray doses received during the operation of x- and g-ray
instrumentation and gauges are trivial compared with the maximum permitted doses be-
cause of the low x-ray output of radioisotope sources, careful design of operating tech-
niques, and x-ray shielding.
Table1 Properties of Radioisotope Sources Used for XRF, XRA, and XRS Analysis and
Determination of Bulk Density r, Mass per Unit Area (rt), and Thickness t in X-ray Analysis
Radio-
isotope
Half-life
(years)
X- or g-ray
energy (keV)
Photons per
disintegration
Dose at 1 m
from 1 GBq
(27 mCi) (mSv=h)
Analytical
technique
55
Fe 2.7 MnK x-rays (5.9, 6.5) 0.28
a
XRF
238
Pu 88 UL x-rays (1330) 0.13
a
XRF
244
Cm 17.8 PuL x-rays (1421) 0.08
a
XRF
109
Cd 1.3 AgK x-rays (22, 25) 1.02
a
XRF
g-rays at 88 0.04
125
I 0.16 g-rays at 35 0.07 2.7 XRF
TeK x-rays (2732) 1.38
241
Am 433 g-rays at 59.5 0.36 3.6 XRF, XRA, XRS
153
Gd 0.66 EuK x-rays (4148) 1.10 27 XRA
g-rays at 97 0.30
g-rays at 103 0.20
57
Co 0.74 g-rays at 122 0.86 24 XRF, XRA, XRS
g-rays at 136 0.11
133
Ba 10.8 g-rays at 81 0.34 65 XRA, XRS,
g-rays at 276 0.07 rt
g-rays at 303 0.18
g-rays at 356 0.62
0.09
137
Cs 30.2 g-rays at 662 0.85 83.7 rt
a
It is difcult to assign a radiological protection meaning to the dose of low-energy x-rays.
Copyright 2002 Marcel Dekker, Inc.
Figure 4 Encapsulation of disk and annular
109
Cd g-ray sources. Dimensions are in millimeters.
(From Amersham, 1986.)
Copyright 2002 Marcel Dekker, Inc.
The International Organization for Standardization (ISO, Geneva) has produced a
system for classifying sealed radioisotope sources based on safety requirements for typical
uses (Amersham, 1986). Prototype sealed radioisotope sources undergo temperature, ex-
ternal pressure, impact, vibration, and puncture tests (Table 2), which increase in severity
as the class designation increases from 1 to 6. The ISO classies the test requirements for
specic types of application of the sealed sources. The classication for low-energy g-ray
gauges and XRF analysis instruments used in industry is C33222 [i.e., from Table 2, the
rst classication 3 is temperature ( 740
C and 180
C (20 min.),
80
C (1 h)
740
C (20 min.)
180
C (1 h)
740
C(20 min.),
400
C (1 h) and
thermal shock
from 400
C to 20
C
740
C (20 min.),
600
C (1 h),
thermal shock from
600
C to 20
C
740
C (20 min.),
800
C (1 h)
thermal shock from
800
C to 20
C
External
pressure
b
No test 25 kPa absolute to
atmospheric pressure
25 kPa absolute to
2 MPa absolute
25 kPa absolute
to 7 MPa absolute
25 kPa absolute
to 70 MPa absolute
25 kPa absolute to
170 MPa absolute
Impact
c
No test 50 g from 1 m 200 g from 1 m 2 kg from 1 m 5 kg from 1 m 20 kg from 1 m
Vibrations No test 30 min, 25500 Hz at
5gn peak amplitude
30 min, 2550 Hz at
5gn peak amplitude;
5090 Hz at 0.635 mm
amplitude peak-to-peak;
90500 Hz at 10gn
90 min, 2580 Hz
at 1.5 mm amplitude
peak-to-peak;
802000 Hz at 20gn
Puncture
d
No test 1 g from 1 m 10 g from 1 m 50 g from 1 m 300 g from 1 m 1 kg from 1 m
a
Details of the testing procedures are given in ISO.2919 and BS.5288. A further class X can be used when a special test procedure has been adopted.
b
External pressure 100 kPa 1 atm (approximate).
c
The source, positioned on a steel anvil, is struck by a steel hammer of the required weight; the hammer has a at striking surface, 25 mm in diameter, with the edges
rounded.
d
The source, positioned on a hardened steel anvil, is struck by a hardened pin, 6 mm long and 3 mm diameter, with a hemispherical end, xed to a hammer of the required
weight.
Note: The shaded cells of the table when read from top to bottom will correspond to class C33222, which is the minimum requirement for low-energy g-ray and x-ray sources
used in XRF instrumentation and g-ray gauges.
Copyright 2002 Marcel Dekker, Inc.
relative, and its is therefore more convenient and accurate to express their FWHMs in
units of energy rather than percent. Energy resolution, expressed as the full width at
half(peak)-maximum (FWHM) and shown for each type of detector as continuous line in
Figure 6, was calculated from equations given by Jenkins et al. (1981d). It is important to
remember that practical energy resolution for any detector is always worse than that
quoted by 10% to sometimes 20% relative. This is because the spectra measured from real
samples usually contain large amounts of backscattered radiation, and, additionally, the
intensities of the analytes may be very high.
As mentioned earlier, the energy resolution of solid-state detectors is superior to that
for proportional and scintillation detectors (see Fig. 26 in Chapter 2). Figure 7 shows the
calculated energy spectrum for the detection of 8-keV x-rays in each detector and also the
energies of the Ka x-rays in the 69-keV energy range. Figure 8 shows the dierence in
energy of Ka x-rays between adjacent atomic number elements.
Table 3 compares the dierence in Ka x-ray energies with the energy resolution for
aluminum, iron, and tin based on data given in Figures 6 and 8. Solid-state detectors are the
only detectors that can fully resolve Ka x-rays of adjacent Z elements. The factors aecting
their resolution are discussed in detail in Chapter 3 (Sec. III.D). Proportional detectors have
an energy resolution less than twice the energy dierence in Ka x-rays of adjacent Zelements.
Hence, their energy-resolving power is useful even if there are adjacent Z elements in the
sample. Scintillation detectors (see Chapter 2, Sec. III.F.2) have suchlimited resolving power
that other techniques must be used to discriminate between adjacent Z elements, such as
balanced lters. This achieved, however, at the expense of some loss in sensitivity of analysis.
Figure 9 illustrates the resolving capabilities of detectors by showing calculated energy
spectra for identical concentrations of Fe and Ni in the same sample.
Figure 5 Definition of energy resolution, FWHM, of an x-ray detector, which is measured at the
half height of MnKa peak. Note the shaded area, which indicates the width of background window.
Copyright 2002 Marcel Dekker, Inc.
2. Energy Resolutionof Detector and MinimumDetectable Level
The energy resolution of the detector not only determines its ability to resolve the x-rays of
adjacent elements but also decides about the minimum quantity of the element that can be
detected by the detector, a so-called minimum detection limit, or the detection limit (DL).
The better the resolution (i.e., the narrower the x-ray peak), the better (i.e., smaller) the
minimum detectable quantity of element. The detection limit is frequently dened by
DL
3
S
I
bgd
t
r
21
where S is the sensitivity (in counts=s per unit concentration of analyte), I
bgd
is the
background intensity measured in the analytes integration window, and t is the mea-
surement time.
Figure 6 X-ray energy resolution (FWHM) of scintillation, proportional, and solid-state
detectors. The continuous lines are calculated (data from Jenkins et al., 1981d), the silicon detector
results being based on a small detector (10 mm
2
63 mm): The (diamonds) resolutions are for various
solid-state detectors (EG&G Ortec, 1986); () typical and (6) best for the high-resolution
proportional detectors (Metorex, 1986); and (open circles) typical and (solid circles) best resolutions
for specific NaI scintillation detectors (Harshaw).
Copyright 2002 Marcel Dekker, Inc.
The background intensity, I
bgd
, is measured in the same window as the intensity
of a given x-ray peak. The width of the window is usually set equal to the peak
FWHM. If a sample containing a certain percentage of the analyte is measured with a
detector of energy resolution E
1
(where E
1
FWHM), it will generate in its FWHM-
wide window a certain intensity, I
1
. If the detector resolution now degrades to, say, E
2
(i.e., the analytes x-ray peak is now wider), the intensity I
2
will still be equal to I
1
(as
long as the measurement window is FWHM-wide). This is because the energy de-
posited by the photon in the detector is represented by the area of the whole peak.
Therefore, if the peak is wider, it has to be smaller, and if it is narrower (better re-
solution), it has to be tallerfor the area of the peak to remain constant (see Fig. 5).
However, the background intensity will be larger in the case of a worse energy re-
solution, E
2
, than in the case of E
1
, because the integration window for the back-
ground intensity is wider in the case E
2
than in the case E
1
. Then, it follows from
Eq. (19) that the DL for the E
2
case will be larger (worse) than in the case of E
1
energy resolution. Therefore, for the best DLs, it is of paramount importance to use
the detector with the best energy resolution.
3. Detector Efficiency for X-raysand Detector-Sensitive Area
Figure 10 shows the calculated eciencies of scintillation and solid-state detectors over
the energy range 1150 keV (see also Chapter 3). At low energies, the decrease in ef-
ciency is due to the absorption of x-rays in the beryllium window at the front of the
detector.
Figure 7 The calculated energy spectra for the detection of 8 keV x-rays in scintillation,
proportional, and solid-state detectors. (From Watt, 1983.)
Copyright 2002 Marcel Dekker, Inc.
The eciency of the detector for registering the x-rays at high x-ray energies is
determined by the probability that the x-ray interacts with atoms in the sensitive volume of
the detector. The most ecient detectors are those with a high atomic number and a high
mass per unit area. For x-ray energies above about 40 keV, solid-state detectors made of
germanium are preferred to those made of silicon. Also, of the two detectors made of the
same material, the thicker one will have a better eciency for high-energy x-rays. This is
why even the smallest Si(Li) detector (usually at least 3 mm thick) is superior to the
siliconp-i-n diode detector, which is usually not thicker than 0.5 mm. The eciency of
Table 3 Difference in Energy of the Ka X-rays of Adjacent Atomic Number Elements,
and the Energy Resolution of Three Types of Detector
Atomic No.
of element
Energy of
Ka x-rays (eV)
Dierence in
Ka energies (eV)
Energy resolution of detector, FWHM (in eV)
Solid
state
Gas-lled
proportional Scintillation
13 (Al) 1,490 253 117 425 3,000
26 (Fe) 6,400 527 160 660 6,200
50 (Sn) 25,300 1087 275 1,750 12,200
Figure 8 The energy difference between the Ka x-rays of adjacent atomic number elements.
Copyright 2002 Marcel Dekker, Inc.
Figure 9 Comparison of energy resolution of Si(Li) (FWHM160 eV) and gas proportional
detector (FWHM660 eV). The solid line is spectrum of the 1-to-1 ratio of iron and nickel collected
with the Si(Li) detector. The dotted spectrum is generated with a gas proportional detector for the
same sample.
Figure 10 Calculated efficiencies of NaI scintillation detector and silicon and germanium solid-
state detectors used in XRF analysis.
Copyright 2002 Marcel Dekker, Inc.
proportional detectors depends on the type and pressure of the lling gas and the diameter
of the detector (Fig. 11).
The sensitive area for scintillation detectors is usually from 1000 to 2000 mm
2
; for
proportional detectors, it ranges from about 500 to 1000 mm
2
; for silicon solid-state de-
tectors, it is 10100 mm
2
; and for the commercially available silicon p-i-n diode detectors,
Figure11 X-ray detection efficiencies of Metorex International (formerly Outokumpu Electronics)
proportional detectors with different types of gas fillings. The number code below each graph is gas
pressure, bar (1), window thickness (in mm) (2), gas mixture (3), low background (4), high efficiency
(5) long lifetime (6), high count rate (7), and high resolution (8). The ratings shown are excellent
(xxx), good (xx), fair (x).
Copyright 2002 Marcel Dekker, Inc.
it is currently 313 mm
2
. Hence, in general, count rates are highest for the poor-
er-resolution detectors. The sensitive area of the detector determines the type of mea-
surement geometry that can be used with each detector type. Central source geometry
(Fig. 1b) is normally used with scintillation and proportional detectors, because the large
sensitive area compensates for the shadowing eect of the centrally located source. An
annular source geometry (Fig. 1a) is characteristic for Si(Li) solid-state detectors and
especially for their stationary installations. A side source geometry (Fig. 1c) is the only
choice for the small-area detectors, specically the most recent semiconductor detectors
such as HgI
2
, silicon p-i-n-diode, CdTe, CdZnTe, and so forth. The eciency of solid-state
detectors is discussed in some depth in Chapter 3 (Sec. III.F).
4. Ratio of Full-Energy Peak toTotal Spectrum
The ratio of the area of the full-energy peak to the total area of the x-ray spectrum is
critical to the sensitivity of XRF analysis. The spectrum outside the full-energy peak is
caused by many factors.
The rst and most important is the occurrence of the so-called escape peak
(Jenkins et al., 1981c), resulting from incomplete photoelectric absorption of the incident
x-ray in the detector material followed by escape of some of its uorescent x-rays from
the detector. For example, if an x-ray photon of iron energy 6.4 keV enters the active
volume of the detector, its energy is being absorbed by the detector material also by the
mechanism of excitation of characteristic x-rays of the detector material, such as Si, Ar,
and so forth. If, in turn, the excited x-ray photon of, say, silicon, of energy 1.74 keV, is
not absorbed in the detector but wanders outside of it, then the total energy deposited in
the detector by the original iron photon will be smaller by the energy of silicon x-ray
photon which escaped from the system. The energy left in the detector equals the
dierence in energy of the x-ray entering the detector and that of the escaping uor-
escent x-ray. This will give rise to the small peak always located left to the original x-ray
peak, at a distance equal to the energy of characteristic x-rays of the detector material.
Thus, for example, for silicon-based detectors, the escape peak may be observed
1.74 keV left to any original photopeak. The escape peak is greatest for the higher-Z
detectors. The ratio of x-rays in the escape and full-energy peaks is highest for pro-
portional detectors with gas llings of Xe (Ka x-ray of 29.7 keV), Kr (12.6 keV) and Nal
(iodine Ka of 28.5 keV) scintillation detectors. However, even silicon (1.74 keV) has
about 12% of the detected counts in the escape peak when excited by 2 keV or higher-
energy x-rays. The magnitude of the escape peak is also strongly dependent on the size,
shape, and geometry of the detector itself. The smaller the detector, the more likely the
escape of the characteristic x-ray of detector material is, because it is more likely for the
photon to be generated at the wall of the detector. Also, escape peaks are more intense
for more intense original photopeaks.
Other factors that lead to incomplete absorption of the energy of the x-ray in the
detector are Compton scattering of the incident x-ray in the detector, with the scattered
x-ray or Compton electron escaping from the sensitive volume; alternatively, the incident
x-ray may be photoelectrically absorbed in the detector, but the photoelectron escapes
from the sensitive volume of detector before losing all its energy. These phenomena, which
are dierent manifestations of the escape mechanism, result in the contribution to the
spectrum at energies dierent from the energy of the original x-ray photon. The full-energy
peak to total spectrum is highest for the high-Z detector materials and for low-energy
x-rays. It is lowest for the low-Z gases used in some proportional detectors.
Copyright 2002 Marcel Dekker, Inc.
Apart from the phenomena taking place within the detector itself, the total spectrum
is also a function of the instrument design. For example, typically most of the radiation
reaching the detector is not characteristic x-ray radiation from elements in the sample, but
primary radiation from the source scattered on sample and its environment. It is, there-
fore, very important that the design of the instrument be focused on minimizing and
optimizing of measurement geometry in order to reduce to absolute minimum any para-
sitic radiation reaching the detector.
5. Comments onthe Characteristics of Proportional Detectors
The characteristics of proportional detectors vary considerably with type of lling gas and
its pressure and are much more variable than the characteristics of scintillation and solid-
state detectors. The best energy resolution is obtained by using Penning mixtures as gas
llings of these detectors (Jarvinen and Sipila, 1984b). Although the improvement in en-
ergy resolution is relatively small (Fig. 6), it is critically important for applications in the
atomic number range 2630 (iron to zinc). The low average ionization energy of Penning
mixtures also leads to other important advantages (Jarvinen and Sipila, 1984b): the vol-
tage required is lower, hence the gas pressure can be higher. This leads to a higher e-
ciency of detection, fewer wall eects, and smaller escape peaks and, consequently, to a
higher ratio of peak to total spectrum. The life of the detector is also increased to more
than 10
13
counts because of the use of only noble gases.
The characteristics of proportional detectors supplied by Metorex International
(formerly Outokumpu Oy) are summarized in Figure 11. The recommended gas llings for
proportional detectors depend on the specic analysis application. The eciency of de-
tection of low-energy x-rays is limited by the absorption of the x-rays in the beryllium
window. Proportional detectors with lower gas pressures are used in the detection of low-
energy x-rays because the thinner windows do not withstand high pressures. For those
low-energy x-rays, the detectors lled with neon gas and tted with Be windows as thin as
13 mm are commercially available.
6. Developmentsin Solid-State Detectors
The solid-state detector (SSD) is the best type of x-ray detector for XRF analysis, but,
until recently, its potential has not been fully realized, particularly in industrial and eld
use, because of the need for liquid-nitrogen cooling. The only successful exceptions have
been pioneering and notable Kevex designs, known as Analyst
TM
Model 6700a factory
oor modeland X-SITE
TM
a very rst portable, solid-state detector alloy analyzer
(Spiegel and Horowitz, 1981; Kevex Corp.). Both of these models featured a traditional
Si(Li), liquid-nitrogen-cooled detector and either annular or capsule-type radioisotope
source(s). Perhaps the best testimonial to the success of these designs of the early eighties is
the fact that owners of quite a few still working systems spare no eort and ingenuity to
keep them alive, despite a total abandonment by the original manufacturer.
There has been much promising research into mercuric iodide, cadmium telluride
and gallium arsenide SSDs (Cuzin, 1987) which can operate at or near ambient tem-
peratures. The eld-eect transistor (FET) of the low-noise charge preamplier associated
with these detectors must be cooled to at least 720
C. Over
the last decade, the Peltier-cooled mercuric iodide detector has been used in the com-
mercially available, portable x-ray analyzers (Berry and Voots, 1989; Piorek, 1997).
Copyright 2002 Marcel Dekker, Inc.
However, production yields for these detectors are still problematic, and the same applies
to other semiconductor detector materials such as CdZnTe. It is reasonable to say that
should the demand for those detectors be as high as for silicon in the electronics industry,
we would most likely had mastered the technology of these promising semiconductor
materials a long time ago. Developments have led to the production of silicon-lithium-
drifted detectors that can be operated at temperatures much higher than that of liquid
nitrogen (195
C). Madden et al. (1986) used these silicon detectors cooled in a Peltier
cryostat. The front-end assembly, mounted in the cryostat, contains a silicon detector and
a FET and is mounted on a four-stage Peltier cooling cell. With the assembly under high
vacuum, a temperature of 774
s
2
instr
s
2
stat
s
2
het
s
2
cal
s
2
samp:plac:
s
2
matrx
s
2
part:size
s
2
??
q
22
where
s
instr
Instrumental component such as thermal drift
s
stat
Uncertainty due to statistics of counting
s
het
Uncertainty reecting sample heterogeneity
s
cal
Reects uncertainties contributed by reference analysis and calibra-
tion t
s
samp:plac:
Uncertainty due to sample placement
s
matrx
Uncertainty associated with matrix eects
s
part:size
Uncertainty caused by varying particle size of analyzed material
s
??
Any other potential, unidentied uncertainty
This formula assumes that all uncertainties included are random and normally distributed
and that systematic uncertainty (bias) is negligible. Under these assumptions, the s
tot
can
be regarded as a good measure of the accuracy of analysis.
The rst two uncertainty components, s
instr
and s
stat
, are often treated together
as main constituents of instrument precision (repeatability), whereas all the other
uncertainties identied in Eq. (22) can be regarded as contributing to user or application
related uncertainty. Thus, Eq. (22) can be written as follows:
s
tot
s
2
instr:precis:
s
2
applic:
q
23
where
s
instr:precis:
s
2
instr
s
2
stat
q
24
and
s
applic:
s
2
het
s
2
cal
s
2
samp:plac:
s
2
matrx
s
2
part:size
s
2
??
q
25
Some of the uncertainties can be identied and easily controlled or minimized by the
analyst, such as uncertainty due to statistics of counting, or calibration uncertainty. For
example, the statistical uncertainty of counting, s
stat
, can be easily reduced by extending
Copyright 2002 Marcel Dekker, Inc.
the measurement time, whereas the uncertainty associated with calibration, s
cal
, can be
reduced by using more accurately analyzed calibration samples and by selection of a better
calibration equation.
Sample placement uncertainties, s
samp:plac:
, can be reduced by always following the
same procedure for preparing and loading the sample into the measuring chamber of the
analyzer, orpartiallyby spinning the sample during measurement.
Uncertainties due to matrix variability, s
matrx
, and particle size, s
part:size
, can be sig-
nicantly reduced by proper sample preparation techniques such as matrix dilution
(Jenkins et al., 1981a) and by grinding, pelletizing, or fusing the sample with spectrometric
ux, respectively.
Although all these uncertainty minimization techniques are valid and readily ap-
plicable in laboratory environments, they are an unattainable luxury in eld and industrial
applications where grinding, fusing or any type of sample preparation is simply out of the
question.
This is why the accuracy of analysis in eld and industrial environments tend to be
governed by the application-related uncertainties rather than by the usually much smaller
contribution from the instruments precision. It is also for this reason that the measure-
ment time may often be selected from the point of view of process throughput rather than
overall uncertainty of analysis; usually, good instrumental precision allows the analyst to
shorten the measurement time without sacricing the former. Nevertheless, the mea-
surement time is always a compromise between the throughput of the process and ana-
lytical requirements.
In the coating industry, the time for the determination of the coating mass of tin
and zinc on steel must be less than a few seconds if the result is to be useful; hence
continuous analysis directly on the main coating line is essential. In mineral con-
centrators, rapid changes in the grade of ore entering the plant and the time taken for
the ore to pass through the plant (about 15 min) make it essential for the process
slurries to be analyzed within 5 min. This can be achieved by continuous analysis either
directly in stream or on slurries in sampling loops (bylines). In scrap yards, alloys must
be sorted into dierent types in short periods: otherwise the sorting operation is not
economical. The lower the price per pound of scrap, the faster the sorting has to be. In
this case, it is essential to have analysis equipment that is both portable and capable of
producing rapid results.
It is important to realize that XRF analysis does require some nite time period
within which a meaningful measurement can take place. The consequence of this fact is
that XRF analysis results in process control applications are, most of the time, the average
rather than point estimates of the process parameters such as coating thickness, sulfur
concentration of diesel fuel, and so forth.
B. Uncertainties of Sampling and Sample Presentation
Uncertainties originated by sampling processes are probably one of the most signicant
contributors to the overall analytical uncertaintythe point we attempted to make in the
previous subsection. This is especially true for eld and industrial applications of XRF
methods, where sample preparation and conditioning is nonexistent, and which, as we
stated in the Introduction to this chapter, are mostly radioisotope based. That is also why
we devote so much space in this section to this issue.
All analyses involve sampling, some more than others. In the continuous on-line
measurement of material on conveyors or in process slurry streams, the analysis is
Copyright 2002 Marcel Dekker, Inc.
averaged over large quantities of material, but still has the potential for sampling un-
certainties because not all of the material on the belt or in the slurry stream is viewed. If
material is continuously sampled from the main stream and fed through a sampling loop
past the x-ray analyzer, the uncertainties in sampling the main stream and in viewing only
part of the sample loop must both be considered.
A mathematical approach to sampling of a particulate media has been detailed by
Gy (1982) and Pitard (1993). The most readily accessible information on sampling is given
in the various international standards for the sampling of process material from moving
conveyors and from slurry streams (ISO, Geneva). These describe preferred sampling
practices and estimate the accuracies of sampling, including the eects of dierent top sizes
and the number and mass of sample increments taken. These standards are regularly
amended.
Sampling uncertainties are of critical importance to the exploration for and assess-
ment of deposits and the mining of both metalliferous ores and coal. After all, it is possible
to analyze only a very small proportion of the ore. There are established practices for
choosing where to sample (e.g., on a regular spaced grid), the number of samples to be
taken, and the weight of each sample to be taken. This complex eld of sampling geo-
statistics is thoroughly covered in textbooks (David, 1971; Ingamells and Pitard, 1986;
Isaaks and Srivastava, 1989).
1. Uncertainty Dueto Sample Heterogeneity
Samping uncertainties, specically in analysis of particulate type of material, are mainly
caused by heterogeneity of the measured medium. Should the analyte be uniformly
distributed throughout the mass=volume of the measured material, any sample taken
for analysis would contain the same concentration of the analyte. However, even this
ideal condition would be compromised by the morphology of the sample. If the sample
is liquid, then the dissolved analyte can, indeed, be homogeneously dispersed in a
liquid matrix. However, if the analyzed medium is ore or soil, or any other powder, the
chances of nonuniform distribution of the analyte in matrix increase dramatically. Not
only might the grains of the analyte not mix well with grains of the matrix, but any
handling of the sample will promote tendencies for segregation of one type of grain
from the other. A very important contributing factor is the size of the sample taken for
XRF analysis. The smaller sample is likely to emphasize its heterogeneity as compared
to a larger sample. The main consequence of heterogeneity of the material analyzed is
the fact that the sample of this material taken for analysis may not be representative of
the material.
If the uncertainty cannot be minimized, the next best way of dealing with it is to
estimate its magnitude. It is rather easy to evaluate the contribution of material hetero-
geneity factor to the total uncertainty. The only requirement is the availability of a suf-
cient quantity of analyzed material. Two series of measurements are required. In the rst,
a sample of material to be analyzed is measured in x-ray analyzer for at least seven times,
without being disturbed in any way. This series of measurements will yield a standard
deviation, which will correspond to s
tot
, as per Eq. (20), except for the s
het
, that being
equal to zero because only one sample was measured. Next, at least seven dierent samples
of the same material are measured under the same conditions as before (i.e., the same
measuring cup, counting time, etc.). This second series of results will yield a standard
deviation, which will correspond to s
tot
, exactly as per Eq. (20), now including the het-
erogeneity factor, s
het
. By subtracting the squared standard deviation of the rst series
Copyright 2002 Marcel Dekker, Inc.
from that of the second and taking the square root of the dierence, a numerical value of
s
het
is readily obtained. For example, in soil analysis for lead, the two standard deviations
obtained were 25 and 65 mg=kg, respectively. Therefore, after subtraction, the s
het
was
60 mg=kg, a signicant contribution to the total uncertainty.
It is very important to realize the fact that in the majority of analyses of granular or
particulate material, its heterogeneity determines the overall accuracy of analysis. This is
true not only for XRF, which by its nature is a surface measurement method, but also for
other instrumental techniques such as atomic absorption or inductively coupled plasma,
as long as an aliquot analyzed is not representative of the material from which it was
taken. It is therefore of paramount importance that good homogenization is applied to
analyzed material and that the amount of sample taken is sucient to be representative of
it, especially if verication or validation of XRF results with other analytical techniques is
required.
2. Sample Heterogeneityand Detection Limits
A detection limit (DL) for a given analyte is understood to be that amount of analyte in a
sample that produces a spectral signal equal to or greater than three standard deviations of
a signal obtained on a sample with no analyte present.
The authors own data and values cited in literature indicate that a typical DL for lead
in soil is about 50 kg=kg when a
109
Cd isotope and semiconductor detector are used for
analysis. The
109
Cd isotope x-rays reach to a depth of about 1.3 cm into a silica-based soil.
However, if there is leadinthis sample, the excitedleadLseries x-rays canreachthe surface of
soil only if they are generated at a depth not greater than 0.20.3 cm, as measured from the
soil surface. A 50-mg=kg DL for lead means, in practice, about 650 mg of total lead as po-
tentially seen by the analyzer in a soil cylinder of 13 g mass (cylinder base area of 5 cm
2
by
1.3 cm high by 2 g=cm
3
of assumed specic density of silica soil). However, the lead x-rays
reach the detector from a depth of only 0.3 cm and, therefore, the eective mass of total lead
seen is only 150 mg. This mass of lead translates into a 0.030-cm (or 12 mils)-diameter
sphere of pure lead. Even in a form of an oxide, the size of the lead oxide grain would not be
much more. If this small, by all means, lead grain happens to be within the rst 3 mm of soil,
the XRFanalyzer will be able tomeasure it. However, if the leadgrainends upat deeper layer
of soil, it will never be noticed. Obviously, depending on its characteristic x-rays energy and
excitation source, any other element will be similarly aected by its location in the sample.
This perhaps extreme but simple example illustrates how illusionary and misleading the
concept of detection limit may be in case of granular material.
C. Choice of Radioisotope X-rayTechnique
The choice of the most appropriate radioisotope x-ray technique for a specic application
depends on the requirements of accuracy of sampling and x-ray analysis and the time
available for the analysis. The simplest radioisotope x-ray technique that satises these
requirements is usually chosen.
In the laboratory, where many dierent types of analyses can be performed, XRF
analysis with a solid-state detector is the most exible method. For industrial and eld
applications, there is often a more restricted range of analyses paired with a greater need
for simple and reliable equipment, the cost of which may also be an important con-
sideration. X-ray uorescence analysis based on scintillation or proportional detectors is
often, but not always, the best approach. These considerations lead to the use of a much
wider range of radioisotope x-ray techniques in industry than in the laboratory. The choice
Copyright 2002 Marcel Dekker, Inc.
of technique is more complicated. Should an ore sample be ground before analysis and L
shell XRF be used for a high-Z element, such as uranium, or can K shell XRF techniques
determine the concentration with sucient accuracy despite the heterogeneity of the
sample? Should the analysis technique be chosen to make a direct measurement on-line, in
a sampling loop of the process stream, or on a sample taken to a laboratory? Should XRF
techniques even be used in an on-line application, given the inaccuracies introduced by the
heterogeneity of the material to be analyzed? The practical alternative may be to use XRA
or XRS techniques or a nuclear technique, based on more penetrating radiation, such as
high-energy g-rays or neutrons. The applications discussed in the next section indicate the
preferred solutions to some important analysis applications, particularly in industry.
VI. APPLICATIONS
Table 4 lists some important applications of radioisotope x-ray techniques based on XRF
equipment, usually referred to as laboratory or portable elemental analyzers. The analysis
techniques involve the use of scintillation, proportional, and semiconductor detectors.
These analyzers are in widespread use in many application areas and in many industries
(Rhodes, 1971; Rhodes and Rautala, 1983). The total number used worldwide probably is
about 10,000. Some major suppliers of radioisotope based instruments are listed in Table 5.
A photograph of a portable model of an x-ray analyzer for alloys, build around a silicon
p-i-n diode detector is shown in Figure 23 and a laboratory model based on gas-lled
proportional detectors is shown in Figure 24. Many solid-state detector systems are also in
routine use in laboratories.
Table 6 lists types of on-line analysis systems in routine use. These include systems
for the on-line analysis of mineral slurries, owing powders, coal, coal slurries, paper,
determination of sulfur in oil and petroleum products, and analysis of coatings. Most of
the analysis systems tend to be based on scintillation and proportional detectors, but some
of the more recently developed systems also use solid-state detectors. Table 6 is also an
attempt to list some commercial suppliers of the on-line equipment. Example of the ap-
plication of these techniques are now discussed in details, with emphasis given to appli-
cations of industrial importance. Some examples illustrate the interaction of sampling and
sample presentation with the selection of the appropriate radioisotope technique.
A. Identification of Alloys
Historically, the rst applications of portable XRF analyzers were in mining and pro-
specting. It was only with the advent of the on-board memory and microprocessors that the
portable XRF analyzers found wider acceptance and use for analytically more demanding
alloy identication and analysis. Since then, several thousand of these analyzers have been
sold, making alloy sorting and analysis a agship application for portable XRF analy-
zers. These analyzers can assay the alloy and=or identify it by its grade or common name.
Rapid sorting of alloys is required in many areas of the metals industry, such as
smelting, fabrication, inventory and incoming material control, and the sorting of scrap
(Piorek and Rhodes, 1986; Berry, 1981; Piorek, 1989). Some common alloy groups include
nickel alloys, copper alloys, stainless and high-temperature steels, and carbon and chro-
miummolybdenum steels. Although 4050 elements are involved in the alloying process,
in any given alloy there are only 1020, and of these, only about 10 are required for the
identication of specic alloy.
Copyright 2002 Marcel Dekker, Inc.
The main requirements of analytical equipment for alloy identication are port-
ability, speed and reliability of identication, and an ability to be used by unskilled
operators. Balanced-lter techniques have two main disadvantages. Concentrations of at
least several elements must be determined; hence, separate measurements must be made
with several sets of balanced lters. The sensitivity of analysis is insucient for the lower
concentrations of some specic elements in the alloys.
Piorek and Rhodes (1986) showed that by using XRF analysis based on a 111-MBq
(3 mCi)
109
Cd source and a high-resolution proportional detector, many alloys can be
identied in one measurement. Measurement is rst made to identify the alloy by group.
The spectrum of the unknown alloy is then compared with the key features of spectra of
known alloys in the group, which are stored in a memory chip in the equipment. This so-
called grade-identication mode of analysis is particularly useful because it does not
require the operator to be procient in the specics of alloy composition. In the grade-
identication mode, the analyzer uses a pattern-recognition algorithm to compare the
x-ray intensities of the measured sample with those of the stored references of alloy
standards. This approach is very fast, taking about 35 s to identify a single alloy. In
Table 4 Some Typical Applications of Radioisotope Based Laboratory and Portable
Elemental Analyzers
Application Typical examples
Alloy sorting and
identification
Low-alloy steels; stainless steels; nickel alloys; high-temperature alloys;
titanium, aluminum alloys; specialty alloys; metal scrap
Mining and mineral Copper, lead, zinc, tin, arsenic, molybdenum, nickel, iron, chromium,
bismuth, and uranium in commercial-grade ores, concentrates,
and tailings; titanium and iron in silica sand; silicon, potassium,
titanium, and iron in clays; phosphate rock
Pulp and paper Thickness of silicone coatings on paper and polymer membranes; calcium,
titanium, filler in paper
Environmental Soil screening for metals (Cr, Cu, Ni, Pb, Zn, As, Cd, Hg, Sb); hazardous
materials (e.g., lead, arsenic, chromium, or cadmium in waste sludge);
trace elements in wastewater discharge; metals in air particulates on
filters; chlorine (halogens) in waste oil; sulfur in diesel fuel
Fibers, films and
coatings
Copper, zinc, tin, gold, silver and chromium plating thicknesses; metals
in plating solutions; silver in photographic film; manganese coating
thickness on magnetic tape; titanium on glass; ruthenium on electrodes
Chemicals and
process control
Lead, titanium, and zinc in paint; sulfur, iron, alumina, silica, and calcium
in cement; vanadium in catalysts; palladium and gold coatings on silica
spheres uses as catalysts; zinc, chromium, nickel in plating baths
Plastics Calcium, lead, tin, and chlorine in PVC; zinc and bromine in polystyrene;
chlorine in urethane rubbers; bromine and chlorine in butyl rubbers;
silicon in polythene; TiO
2
in nylon; bromine in Styrofoam
Agricultural Fertilizers (calcium, phosphates, potassium); copper, chromium, and arsenic
in wood preservatives and treated wood; bromine in almonds; ironzinc
ratio in meat for grading; minerals in cattle feed; titanium in fillers
Cosmetics Titanium, iron, lead in powders
Pharmaceutical Metals in vitamin pills; zinc in insulin
Petroleum products Lead, calcium, sulfur, vanadium, and chlorine in gasoline or oil;
sulfur in petroleum coke; sulfur and ash in coal; lubricating oils additives
Copyright 2002 Marcel Dekker, Inc.
Table 5 Manufacturers and Suppliers of Radioisotope-Based Laboratory and Portable
Elemental Analyzers
Manufacturer
a
Analyzer type Detector type
Amdel Laboratory Scintillation
ASOMA
b
Laboratory Gas proportional
BRGM Laboratory Solid-state Si(Li)
Metorex Int.
b
Portable, laboratory High-resolution gas
proportional; solid state:
Si(Li) or silicon p-i-n diode
Niton Corp. Portable Solid state: silicon
p-i-n diode, CdZnTe
Outokumpu Electronics, Oy Portable, laboratory High-resolution
gas proportional;
solid state: Si(Li)
Oxford Instruments
b
Laboratory Gas proportional
Radiation Monitoring Devices Portable Solid state: CdTe
TN Technologies Portable Solid state: mercuric
iodide, HgI
2
a
See the list of manufacturers and their addresses in the Appendix of this chapter.
b
These manufacturers reacted to the dynamics of the market and industry by switching from isotope-based
instrumentation to x-ray-tube-excited devices.
Figure 23 The Metorex Int. portable alloy analyzer based on high-resolution silicon p-i-n diode
detector, MetalMaster
TM
2000. (Courtesy Metorex Int.)
Copyright 2002 Marcel Dekker, Inc.
comparison, assaying the alloy may take up to 4 min in order to maintain low measure-
ment uncertainty. Table 7 shows the identication results for dierent alloys by group,
obtained with the analyzer pictured in Figure 23. The probability of correct identication
is satisfactory for all alloys except the carbon steels. The results in Table 7 were generated
by measuring each alloy sample for 5 s for at least 10 times and comparing its spectral
features against the library of about 145 dierent alloys. The number of correct identi-
cations ratioed to the total number of trials for a given alloy group was then recorded in
Table 7 as a measure of probability of correct identication. Some of the identication
failures are for alloys very close in composition, for which the main alloying elements
dier in concentrations by less than 1%. The most dicult identication is for carbon
steels in which the concentrations of alloying elements are very low in the presence of
almost 100% iron, and the dierence in concentrations of the same elements between two
grades approaches the sensitivity of the XRF technique. Therefore, the XRF is not a
method of choice for identication of carbon steels, for which better techniques, such as
optical emission spectrometry, exist (Piorek et al., 1997).
By using a silicon p-i-n diode detector in place of a gas proportional one, it has been
possible to extend positive identication to such specic alloys as B1900 and B1900Hf,
which dier from each other by less than about 1% Hf, in the presence of several percent
of W and Ta and a balance of nickel.
Overall, the portable alloy analyzers oer a much simpler approach to identication
of alloys, with a reliability of identication as good as that for more complex techniques.
Despite the signicant dierences in design, the portable alloy analyzers exhibit si-
milar performance in the assay mode; that is, they oer measurement precisions ranging
from 0.01% to 0.5% absolute depending on the element, its concentration range, and the
Figure 24 A laboratory, benchtop, elemental analyzer based on a high-resolution proportional
counter, X-MET
TM
820. (Courtesy of Outokumpu Electronics.)
Copyright 2002 Marcel Dekker, Inc.
matrix. Several references treat this subject in greater detail (Fookes et al., 1975; Jenkins
et al., 1981c, 1981d; Piorek and Rhodes, 1986; Piorek 1989; Spiegel and Horowitz, 1981;
Berry, 1981). Table 8 shows typical performance data for a contemporary portable XRF
alloy analyzer, such as illustrated in Figure 23.
Table 6 On-line Analysis Systems Based on Radioisotope X-ray Sources
a
Application Analytical technique, detector
b
Manufacturer
c
Metal content of
mineral slurries
XRF/XRA, s and ss, in-stream
XRF, p, sample line
Amdel
Asoma
XRF, ss, sample line Outokumpu Mintec, Oy
XRF, ss, in-stream Texas Nuclear
XRF, ss, sample line Ramsey
Metal content of clay
and mineral powders
XRF, p, sample line
XRF, ss, sample line
Asoma
Outokumpu Mintec, Oy
Iron and chromium in
ore on converyors
Dual-energy XRS, s, on-line Outokumpu Mintec, Oy
Ore sorting Dual-energy XRS, s, on-line Outokumpu Mintec, Oy
Ash in coal on conveyor Dual-energy XRS, s, on-line MCI; Harrison Cooper; SAI
XRS, s, on-line EMAG
XRS, s, sample line Humboldt-Wedag
XRS, p, sample-line Sortex
Solids weight fraction and
ash in coal in slurries
XRF, neutron, and g
transmission, s, on-line
Amdel
Tin content of galvanizing
solutions
XRS, p Rigaku
Calcium in cement raw mix XRF, s Rigaku; Outokumpy Mintec,
Oy
Sulfur in oil, diesel fuel,
gasoline
XRF, ion chamber,
sample line
Yokogawa
XRF, p, sample-line Mitsubishi; Metorex Int.
Lead in gasoline XRF, p, sample-line Metorex Int.
Metals in plating bath
solutions (Ni, Cu, Cr, Ta, etc.)
XRF, p, sample-line Asoma; Metorex Int.
Cement analysis for
Ca, Si, Mg, Al, S, Fe
XRF, p, sample-line Metorex Int.
Corrosion products (Cr, Fe) in
steam generator feedwater
of nuclear power plants
XRF, p, sample-line Detora Analytical
Ash content and/or mineral
filler material in paper
XRA, p, b-rays
transmission, on-line
Sentrol; Yokogawa;
Paul Lippke
XRF, p, on-line Sentrol
Coating mass of: Zn, Sn/Cr,
Sn/Ni, Zn/Fe, Sn/Pb on steel
and other substrates
XRF, p, on-line Data Measurement; FAG;
Gammametrics
a
X-ray tube techniques can be used in some of these applications.
b
Detectors: s scintillation, p gas proportional, ss solid state [usually Si(Li)].
c
See the list of manufactures and their addresses in the Appendix of this chapter.
Copyright 2002 Marcel Dekker, Inc.
B. Determination of Uraniumand Gold in Ore
During the exploration of mining of metalliferous ores, large numbers of ore samples must
be analyzed to compensate for the inherent variability of expression of the ore. The ore
may be analyzed in the laboratory or, for higher-Z elements, at the mine face or in situ in
boreholes.
K shell XRF techniques are often preferred for the analysis for uranium; because of
the penetration of the uranium K x-rays in the ore, little or no crushing of samples is
required and, additionally, the uranium concentration is averaged over much larger
samples. Uranium can be determined down to 20 mg=g or about 0.7 oz.=ton (1s) in 30 s
using a 222-MBq (6 mCi)
57
Co and a 28-mm
2
65-mm germanium detector (EG&G Or-
tec, 1986). This technique was used routinely to survey samples for uranium and other
Table 7 Performance of a Portable X-ray Analyzer in Alloy Identication
Alloy group Measured elements
Identication results
(% feasible)
Nickel and cobalt alloys Ti, Cr, Fe, Co, Ni, Cu, Nb, Mo, W, Hf, Ta 100
Copper alloys Mn, Fe, Ni, Cu, Zn, Pb, Sn 90100; 98100
Stainless steels and
high-temperature alloys
Ti, Cr, Mn, Fe, Co, Ni, Cu, Nb, Mo 90100; 100
Cr/Mo steels Cr, Fe, Ni, Mo 90100; 100
Low-alloy steels Cr, Mn, Fe, Ni, Cu, Nb, Mo 6580; 90100
Titanium alloys Ti, V, Mn, Cr, Zr, Mo, Sn 95100;
Aluminum alloys Mn, Fe, Cr, Cu, Zn 90100; 95100
Note: If two results are given, the rst refers to a gas-lled proportional detector, and the second to a solid state,
silicon p-i-n diode detector.
Source: From Piorex et al., 1997.
Table 8 Standard Performance Data for a Typical, Contemporary, Commercially
Available Portable XRF Alloy Analyzer
Alloy group Ti Cr Mn Fe Co Ni Cu Zn Nb/Mo Sn Pb
Low-alloy steels 0.01 0.04 0.1 0.25 0.25 0.10 0.05 0.10 0.006 0.15 0.15
0.02 0.1 0.2 0.50 0.25 0.15 0.01
Stainless, hi-temp steels 0.015 0.20 0.10 0.20 0.20 0.20 0.06 0.20 0.01 0.30 0.05
0.03 0.30 0.20 0.30 0.30 0.10 0.03 0.30
Ni/Co alloys 0.10 0.20 0.10 0.12 0.10 0.20 0.05 0.30 0.02 0.30 0.15
0.50 0.30 0.50 0.50 0.50 0.30 0.08
Cu alloys, brass/bronze 0.02 0.10 0.02 0.02 0.05 0.05 0.15 0.07 0.01 0.008 0.20
0.06 0.06 0.08 0.40 0.30 0.20 0.30
Aluminum alloys 0.02 0.05 0.10 0.05 0.05 0.04 0.05 0.06 0.003 0.005 0.01
0.20 0.10 0.005 0.20 0.02
Titanium alloys 0.40 0.10 0.10 0.06 0.05 0.05 0.02 0.02 0.008 0.005 0.01
Notes: The values listed are typical precision ranges in percent absolute for total assay time per sample not longer
than 60 s. The differences are due to the analyzer model and/or the radioisotope(s) used.
Precision data listed are obtained with a 185-MBq (5 mCi)
109
Cd and 1.48-GBq (40 mCi)
55
Fe sources.
Source: From Piorek et al., 1997.
Copyright 2002 Marcel Dekker, Inc.
high-Z elements (Z ! 40) in the laboratory by the Australian company Geopeko in the
extensive exploration programs that found the large Ranger uranium deposit at Jabiru in
the Northern Territory, Australia (G. Sherrington, personal communication, 1987). It is
also used for borehole logging, with detection limits of 0.04 wt% (1s) for uranium,
tungsten, and lead and 6 mg=g for gold in 120 s counting time using a 518-MBq (14 mCi)
57
Co source and a small silicon detector in a 32-mm-diameter borehole probe (Scitech
Corp. undated).
An analysis system has been developed for the in situ determination of gold in ore at
the mine face (Hill and Garber, 1978). The hand-held probe consists of a 4.4-GBq
(120 mCi)
109
Cd (88-keV g-rays) source, a 200-mm
2
67-mm-thick germanium detector,
and a small Dewar ask containing liquid nitrogen, which must be replenished after 6 h of
operation. The sensitivity to gold has been optimized by careful choice of the incident
g-ray energy and by measuring gold Kb x-rays whose energy is greater than the energies of
most of the Compton-scattered g-rays. The precision for a 30-s scan time is 20 mg=g (1s)
and 2 mg=g for 100630-s scans. The technique is suitable for use in high-grade gold mines
but has insucient sensitivity to be applied widely in gold mining, in which 0.3 mg=g or
0.01 oz.=ton (1s) is normally required.
C. On-Line Determination of Coating Mass
Tin on steel (tinplate), zinc on steel (galvanized iron), zinc and aluminum on steel (zin-
calum), iron oxide on plastic (magnetic tape), and silicon coating on paper are only a few
examples of coated products manufactured in large quantities. The coatings are applied at
high speed. Accurate control of the coating mass per unit area is essential to economize the
operation.
Two XRF techniques can be used to determine the mass of the coating. Fluorescent
x-rays of the coating element can be excited and their intensity measured. The intensity of
the coating element x-rays increases with coating thickness. Alternatively, uorescent
x-rays of element in the base material can be excited. Their intensity decreases with the
increase in coating mass because of the absorption of the incident and excited x-rays in the
coating. Both radioisotopes and x-ray tubes are used as the source of x-rays, with
radioisotope sources preferred, except for those applications requiring very fast response,
such as 0.1 s.
Radioisotope XRF techniques for the on-line determination of coating mass are
based on the use of proportional detectors. These are preferred to scintillation detectors
because of their better energy resolution and because they can be used at the relatively high
temperatures that occur above the hot tin and galvanized iron coating processes. In
commercially available systems (Table 6), the analysis head unit continuously scans across
the width of the strip so that coating mass can be controlled across the whole strip. These
commercial systems are used worldwide in most high-throughput coating operations. They
can also be used to determine the separate coating masses of multiple coatings.
Coated products are usually sold with a specied minimum coating mass. The ac-
curate coating mass determination has led to coatings being controlled much closer to the
minimum specication. The variations in coating mass obtained on the zinc galvanizing
line of John Lysaght Pty. Ltd., Port Kembia, Australia, corresponding to no gauge, gauge
with manual control, and gauge with automatic control are shown in Figure 25. In 1977,
improved control of zinc coating mass led to savings of A$300,000 per year per line; si-
milarly, at the nearby Australian Iron and Steel Pty. Ltd., savings of A$1 million per year
(Watt, 1978) were made for tinplate.
Copyright 2002 Marcel Dekker, Inc.
Another typical coating mass application for XRF instruments is measurement of
zinc phosphate coating on steel (Johnson et al., 1989), where phosphorus x-ray intensity is
used rather than zinc to increase the sensitivity of analysis. A precision of 2 mg=cm
2
was
achieved in a 30 s with gas-lled proportional detector and a 1.5 GBq (40 mCi)
55
Fe source,
for up to 450 mg=cm
2
coating mass.
D. On-StreamAnalysis of Metalliferous Mineral Slurries
Most metalliferous minerals are concentrated from their ores by froth otation. The grade
of ore fed to the concentrator can vary rapidly; hence, to control the otation process, the
concentrations of valuable minerals in the plant process slurries should be determined
continuously. The concentrations, in the slurry solids, of such base metal minerals as
nickel, copper, zinc, and lead are usually in the range of 0.315 wt% for feed streams, tens
of weight % in concentrate streams, and 0.030.3 wt% in tailings (residue) streams. The
Figure 25 The frequency distribution for product zinc coating mass before and after installation
of an on-line zinc coating mass gauge at John Lysaght Pty. Ltd. Product minimum specification is
300 g/m
2
. (From Watt, 1978).
Copyright 2002 Marcel Dekker, Inc.
solids weight fraction is in the range of 1550 wt%. The time for analysis (less than 5 min)
is too short for laboratory analysis of samples taken from the process streams.
A radioisotope x-ray based on the use of scintillation detectors was introduced
in 1973 (Watt, 1985). Cesium-137 g-ray transmission is used to determine the bulk density
of the slurry and, hence, the solid weight fraction. X-ray uorescence techniques based
on direct excitation and detectorradiator assemblies (Fig. 18) are used to determine the
concentration of all but some high-Z elements, such as lead, which are determined by
153
Gd g-ray (about 100 keV) transmission. In each case, the x- or g-ray measurements are
combined with the solids determination to obtain the concentration of elements in the
slurry solids. These techniques are suciently sensitive for all but a few tailing streams
containing very low concentrations of valuable mineral. This limitation was overcome
later by the introduction of a solid-state detector probe.
The radioisotope x-ray system (Fig. 26) is based on probes, each containing a
radioisotope source and a detector, which are immersed directly into the plant process
slurries (Watt, 1985). Electrical signals from the probes are fed to an analyzer unit and its
output to a central computer. Thus, not only is there no need for sampling from the plant
process streams, but all streams are analyzed continuously rather than sequentially.
This radioisotope x-ray system is very dierent from x-ray tube and crystal spec-
trometer systems for on-stream analysis. These systems were developed in the 1960s to
scan slurries sequentially in up to 14 simple sampling loops (Leppala et al., 1971). Analysis
of the slurries involved sampling from the main process stream, running slurries through
long sections of sampling loops to the central analyzer, and subsampling from the sam-
pling loops before presentation to the analyzer. This is a complex and expensive system.
X-ray tube analysis systems developed in the 1980s sequentially route the slurries from up
to ve process streams through a common ow cell viewed by an XRF analyzing unit
(Saarhilo, 1983). The analyzer is mounted in the plant near the process streams and thus
overcomes much of the mechanical complexity of the 14-stream system just discussed.
Both the x-ray tube crystal spectrometer system and the radioisotope XRF solid-state
detector system are routinely used with this new sample loop system. Both systems can
also be used for the continuous analysis of ne powders. The radioisotope system is
capable of determining elements of atomic number as low as 14 (silicon), because with the
short x-ray path length, the absorption of the low-energy x-rays in air is minimized.
The development of dierent systems for the on-stream analysis of mineral slurries
illustrates the strong inuence of sampling and sample presentation on the type of XRF
analysis system used. Various radioisotope (Table 6) and x-ray tube systems are now in
use, with about several hundred plant systems being installed in concentrators throughout
the world. The radioisotope systems, both in stream (Watt, 1985; Berry et al., 1983) and
using short sample loop, and the ve-stream x-ray tube system, are preferred for new
installations because they cost less and are mechanically less complex. Improvements in
plant control based on this analysis information have led to an increase in the recovery of
valuable minerals (Fig. 27), decreased reagent addition, and reduced need for assay and
sampling sta (Watt, 1985). Total savings per concentrator vary in the range from
US$100,000 to several million dollars a year.
E. On-Line Determination of the Ash Content of Coal
Continuous on-line determination of the ash content of coal on conveyors is required for
the control of coal mining, blending, sorting, and preparation operations. The coal is
carried by the conveyors, usually at the rate of 100600 tons=h, the speed of the conveyor
Copyright 2002 Marcel Dekker, Inc.
Figure 26 A system for the on-stream analysis of mineral slurries. The microcomputer outputs solids weight fraction and concentrations of
valuable minerals in the solids of each stream. (Courtesy of AMDL.)
Copyright 2002 Marcel Dekker, Inc.
is about 3 m=s, and the coal particle top size is in the range 10150 mm, depending on the
application. The process requirement for analysis time varies from 1 or 2 s for the fast
sorting of coal to 10 min for the steady-state control used in coal preparation plants.
Three x-ray or low-energy g-ray techniques have been developed: backscatter of
x-rays (H. Fraenkel, private communication, 1987), backscatter of low-energy g-rays
(Fauth et al., 1986; Cierpicz, 1986), and dual-energy g-ray transmission (Fookes et al.,
1983; Gravitis et al., 1987). Each relies on the photoelectric eect, which depends on
atomic number, and on the fact that ash (mainly SiO
2
and Al
2
O
3
with smaller con-
centrations of Fe
2
O
3
) has an eective atomic number, Z, greater than that of the coal
matter (carbon, hydrogen, nitrogen and oxygen).
The x-ray technique (H. Fraenkel, private communication, 1987) depends on the
scatter of approximately 17-keV x-rays from
238
Pu (or
244
Cm) source in the coal and, at
the same time, excitation of iron K x-rays in the coal to correct for the high absorption per
unit mass by Fe
2
O
3
compared with Al
2
O
3
and SiO
2
. Because the low-energy x-rays pe-
netrate only thin layers of coal, the coal is sampled from the conveyor, subsampled, and
ground to 5-mm top size particles, partially dried, and then presented in a moving stream
of controlled geometry to the radioisotope x-ray analysis system. This system compensates
for the eect of variations in Fe
2
O
3
in the ash, a signicant source of uncertainty in some
applications. However, it involves complex sampling, sample handling and processing,
and blockages occur when the coal is very wet.
The low-energy g-ray technique, using an
241
Am (60 keV) source, depends on mea-
surement of the intensity of g-ray scattered from thick layers (<15 cm) of coal. It was rst
used on a high-throughput sample line (Fauth et al., 1986). Although coal must be
sampled, there is no need for the coal to be subsampled and crushed because of the high
Figure 27 Improvements in recovery of mineral concentrators after installation of on-stream
analysis systems. The symbols refer to tin (diamonds), lead (square), zinc (circles) and copper
(asterisks).
Copyright 2002 Marcel Dekker, Inc.
penetration of g-rays in the coal. The technique has been adapted for use on-line (Cierpicz,
1986), the analysis head unit riding on a raft that is spring loaded so that it is always
touching the top of the coal on the conveyor. Its use is restricted to conveyor speeds of less
than 2 m=s and to minimum thickness of 15 cm of coal on the conveyor compared to the
normal practice of 520 cm.
The dual-energy g-ray transmission technique (Fookes et al., 1983; Gravitis et al.,
1987) measures coal directly on the conveyor belt (Fig. 28), eliminating the need for
sampling the coal. The ash content is determined independently of vertical segregation
of coal on the belt, and if segregation across the belt occurs, the narrow beam of g-rays can
be made to scan across the belt to obtain a representative sample. The coal mass per unit
area in the g-ray beam must be <3 or 4 g=cm
2
to achieve sucient sensitivity of analysis.
Variations of iron in the ash limits the accuracy of ash determination in some applications.
The choice of a suitable x- or g-ray analysis technique is highly inuenced by the
complexities in sampling of the coal on the conveyor and the subsequent subsampling and
grinding. Radioisotope techniques that measure directly on-line are preferred to those
involving sampling, and if sampling is necessary, preference is given to those that minimize
sampling and sample presentation.
Dual-energy g-ray transmission is now the preferred technique for the on-line
determination of ash content of coal, except for applications in which unacceptable un-
certainties in ash are caused by variations in iron in the ash. In this case, a high-energy
g-ray technique, which is based on the pair production interaction and is much less sen-
sitive to variations in iron, is preferred. Its main disadvantages are cost and that it must
operate on a sample line (Sowerby, 1986).
F. On-Line Analysis of Paper
Continuous on-line analysis of paper is required for control of the production process.
Paper consists of cellulose, water, and mineral llers. The characteristic and the quality of
various types of paper are, to a great extent, dependent on the quality, quantity, and
distribution of ller materials. These ller materials occupy the spaces between bers and
improve the printing properties of the paper. Filler materials include CaCO
3
, kaolin, talc,
and TiO
2
, and the concentrations of each may vary.
The analysis may be achieved by a combining XRF, XRA, and b-ray transmission
techniques (Kelha et al., 1983). The mass absorption coecients of x-rays in the 110-keV
region are shown in Figure 29, with abrupt changes in the K shell absorption-edge energies
of calcium (4.04 keV) and titanium (4.96 keV). X-ray preferential absorption measure-
ments are made at x-ray energies of 5.9 (
55
Fe), 4.51 (TiKa x-rays), and 3.69 keV (CaKa
x-rays), the latter two energies being contained by exciting a secondary target material
with
55
Fe x-rays (Fig. 30). The total mass per unit area of the paper is determined by b-ray
transmission. The distribution of the ller material through the paper is determined by
making XRF measurements, using
55
Fe x-rays to excite calcium K x-rays, to give the
dierence in CaCO
3
concentrations near the surface on each side of the paper. These
techniques can be used to determine the concentrations of cellulose, kaolin or talc, CaCO
3
and TiO
2
.
G. Determination of Sulfur and Chlorine in Oil
Sulfur in oil is a source of air pollution. Strict environmental controls are applied to limit
sulfur release into the atmosphere. Oil in sample lines from main pipelines is monitored
Copyright 2002 Marcel Dekker, Inc.
routinely by one of two radioisotope x-ray techniques. One technique combines XRF
(
55
Fe and a proportional detector) and b-ray transmission. The other combines XRA,
using 22-keV silver K x-rays that are absorbed equally per unit weight by both carbon and
hydrogen, with a nonnuclear technique to determine the density of the oil. In Japan alone,
more than 100 of the former gauges and several hundred of the latter are used (Table 6).
More recently, this application received an application boost from regulatory agencies
such as U.S. Environmental Protection Agency. Since October of 1993, no on-highway
diesel fuel the United States can contain more than 500 mg=kg of sulfur. This regulation
forced sulfur monitoring for compliance at reneries, distributorships, and local suppliers.
A proportional detector with neon gas lling and a thin, 13-mm Be window, coupled with a
1.48-GBq (40 mCi)
55
Fe isotope allows sulfur analysis in diesel fuel down to 50 mg=kg,
with a precision of 8 mg=kg (1s), in 300 s measurement time (Piorek and Piorek, 1993).
Figure 28 System for on-line determination of the ash content of coal based on dual-energy g-ray
transmission techniques.
Copyright 2002 Marcel Dekker, Inc.
A factor of 5 better results may be obtained by using low-power x-ray tube in the place of
the isotope. The issue of sulfur measurement in petroleum products for compliance with
the EPA regulations has been addressed by American Society for Testing and Measure-
ment (ASTM), with two ASTM Standard Test Methods, D2622 and D4294. Similar
regulations for sulfur in petroleum products for on-highway use are being implemented in
Europe and Asia.
Spent engine oil is burned as fuel in industrial boilers and home heating furnaces.
Because chlorinated solvents have routinely been mixed with much of this oil, there is a
danger that hazardous levels of these compounds or their derivatives will be released into
the atmosphere. The U.S. Environmental Protection Agency has banned the sale of used
oil for fuel if the total halogen level (interpreted as total chlorine level) exceeds 1000 mg=g.
If the total halogen level is more than 4000 mg=g, such oil has to be considered as
Figure 29 Mass absorption coefficients of the most frequently used filler materials, water and
cellulose, relating to the manufacture of paper. (From Kelha et al., 1983.)
Copyright 2002 Marcel Dekker, Inc.
Figure 30 Measuring principle used in the on-line analysis of paper: (a) XRF, (b) XRA, and (c) b-ray absorption. (From Kelha et al.,
1983.)
Copyright 2002 Marcel Dekker, Inc.
hazardous waste and cannot be reused. However, if the total halogen level is greater than
1000 but less than 4000 mg=g, it may be diluted with less contaminated oil to reduce the
halogen content below 1000 mg=g, after which it can be burned as fuel. Piorek and Rhodes
(1987) and Gaskill et al. (1987) showed that a portable XRF analyzer, with a high-
resolution proportional detector and a 0.74-GBq (20 mCi)
55
Fe source, can be used to
analyze for chlorine in oil. The used oil also contains phosphorus, sulfur, potassium, and
calcium, and the intensities of the K x-rays from all these elements were incorporated into
the calibration so that the measured chlorine K x-ray intensity could be properly corrected
for spectral overlap. The analysis time was 30 s per sample, and the overall accuracy
achieved was quite adequate.
H. Analysis of Environmental Samples
Perhaps the second most important group of applications well served by XRF is in the
analysis of a variety of environmental samples. This is especially true of portable XRF
analyzers, which, in a very natural way, drifted frommineral prospecting to soil screening for
metals; the two applications virtually dier only by name. Successes of XRF in analysis of
thin lms and coatings have been parlayed into measurement of air particulates collected on
lter media. Even much more analytically demanding analysis of water can also be per-
formed with XRF using ion-exchange membranes. All these applications will be discussed.
1. Air Particulatesin Ambient Air
The concentrations of particulates suspended in ambient air are determined to identify
sources of air pollution. The air particulates are collected by drawing large volumes of air
through a lter paper, which is then submitted to multielements analysis. Rhodes et al.
(1972) and Florkowski and Piorek (1974) showed that energy-dispersive XRF analysis
based on radioisotope sources and a silicon solid-state detector is a simple and cost-
eective method for determining the elemental concentrations on the lter paper. The
advantage of the solid-state detector is that simultaneous multielement analysis is achieved
with excellent sensitivity and short analysis time.
Analyses were made using 4.44 GBq (120 mCi)
55
Fe, 14.8 GBq (400 mCi)
238
Pu, and
444 MBq (12 mCi)
109
Cd, a 80-mm
2
64-mm-thick silicon detector (energy resolution of
180 eV at 5.9 keV), and counting for 10 min per lter paper with
238
Pu and
109
Cd and 5 min
with
55
Fe. The detection limit (3s of the background under the relevant peak) for the 19
elements measured varied in the range from 0.03 to 0.24 mg=cm
2
. This is more than adequate
for levels required in environmental pollution analysis. Rhodes and Rautala (1983) have
since shown that the same radioisotopes, with activities of 111370 MBq (310 mCi), can be
used with a high-resolution proportional detector to determine the same elements with de-
tection limits (3s) between 0.4 and 12 mg=cm
2
in 4 min. The main competitive techniques are
energy-dispersive techniques using x-ray tubes and solid-state detectors (Chapter 3) and
particle accelerators in proton-induced x-ray emission (PIXE) techniques (Chapter 12). The
energy-dispersive XRF have been used for the last 20 years by various governmental and
international institutions for routine monitoring of ambient air pollution.
2. Air Particulates Monitoringin Stack Emissions and Workplace Air
The dierence between ambient air monitoring and stack emission or workplace air is
only in the levels of concentrations to be measured and frequency of sampling. As in
analysis of ambient air by ltering the air onto a thin membrane substrate, a signicant
Copyright 2002 Marcel Dekker, Inc.
preconcentration of the sample is realized, resulting in improved sensitivity of the XRF
analysis. Although the concentrations of metals in stack emissions and in workplace air
are much higher than in ambient air, they can only be sampled and analyzed for short
periods of time. Nevertheless, despite these restrictions, the portable x-ray analyzer can
still provide sucient sensitivity of measurement. The typical application is analysis of
metals in deposits obtained with personal samplers for breathing zone air used by
workers occupationally exposed to polluted air. Samples of air particulates are col-
lected for 8 h by pumping air through a 37-mm-diameter membrane, at a rate of about
2 L=min. These deposits can then by analyzed by the same portable XRF analyzer which
was used for soil screening during remediation eort. Within the last 3 years, the issue of
metals monitoring in stack emissions from industrial incinerators has been the subject of
regulatory action in the United States and Europe. Regulations in the United States are
pending which will limit emissions of toxic metals from incinerators stacks and will
require continuous on-line monitoring for compliance. Of the several attempts at Con-
tinuous Emission Monitoring (CEM) Systems, XRF proved to be a viable and relatively
inexpensive alternative. The proof of principle experiment using XRF was performed by
Piorek et al. in 1995 (Piorek et al., 1995) during the EPA-sponsored eld test in the
EPAs own hazardous waste incinerator in Little Rock, Arkansas. The test showed that
particulate deposits obtained by ltering stack exhaust for 10 min on lter membrane
can be readily analyzed by the portable XRF system with radioisotopes as excitation
sources (Makov et al., 1968). Table 9 lists the capabilities of the portable analyzer for
metals ltered from air on a membrane substrate (Piorek, 1997). As can be seen, the
detection limits for air particulates collected on the lter are on the order of a micro-
gram per cubic meter of air. Further improvement of the performance is possible using
x-ray tubes as excitation sources and integrating the analyzer into an continuous on-line
monitor.
Table 9 Performance of a Si(Li)-Detector-Based Field Portable X-ray Analyzer in Monitoring
Air Particulates in Workplace Air and Industrial Incinerator Exhaust
Element
Detection limit
(mg/cm
2
)
Detection limit
a
(mg/cm
3
)
PEL
b
in air
(mg/cm
3
)
Detection limit
CEM
c
(mg/cm
3
)
Cr 0.43 3.0 1000 1.5
Mn 0.45 3.2 50 ND
Fe 0.45 3.2 10
4
(as Fe
2
O
3
) ND
Ni 0.29 2.0 1000 ND
Cu 0.29 2.0 1001000 ND
Zn 0.27 2.0 (15) 10
3
ND
As 0.16 1.1 10 1.5
Se 0.16 1.1 200 ND
Pb 0.24 1.7 50 3.1
Cd 0.07 0.5 1003000 3.1
Sb 0.06 0.5 500 1.5
Sn 0.05 0.4 2000 ND
a
For one cubic meter of air (gas) pumped through a Millipore SMWP membrane of 9 cm
2
area.
b
The OSHA Permissible Exposure Limit in work place air, PEL (as per 29 CFR 1910.1000, 1 Jan. 1977).
c
Hazardous waste combustors, revised standards, proposed rule, 40 CFR Part 60 et al., April 19, 1996. NDnot
determined; CEMcontinuous emission monitor. Source: From Piorek, 1997.
Copyright 2002 Marcel Dekker, Inc.
The same technique for analyzing thin-layer samples can be used for measurement of
metallic ions in water, by cycling the water through an ion-exchange membrane. Piorek
and Pasmore (1993) reported the detection limits for transition metals in water at less than
0.1 mg=L if 250 mL of water was cycled seven times through a 37-mm-diameter ion-
exchange membrane, followed by a 200-s analysis with a 0.37-GBq (10 mCi)
109
Cd source.
3. Soil Screeningfor Inorganic Contaminants
On-site analysis of metals in soils, sediments, and mine wastes is required as part of re-
habilitation studies at hazardous waste sites. The distribution of the hazardous metals is
heterogeneous owing to the exigencies of dumping and possible leaching by rainwater.
Many samples must be taken over the surface area and at depths to dene the zones of
metal accumulation and metal depletion. Chappell et al. (1986) and Piorek and Rhodes
(1987) demonstrated that the portable XRF analyzers based on a high-resolution pro-
portional detector and a 3.7-GBq (100 mCi)
244
Cm source can be used to determine
concentrations of arsenic, copper, zinc, and lead in soils in the range of approximately
from 10 to 10,000 mg=g, with an accuracy that is adequate for this type of investigation.
Rapid, on-site analyses can be incorporated immediately into a eld investigation pro-
gram, making it possible to change the density of sampling at any spot, depending on the
results of previous analyses. Since then, portable XRF analyzers have been improved to
include high-resolution semiconductor detector probes and extensively used to measure
in situ metals in soil, such as lead, zinc, copper, arsenic, nickel, chromium, and cadmium.
In the majority of reported cases, the correlation between the XRF and the conrmatory
method results [such as Atomic Absorption Spectrometry (AAS) or Inductively-Coupled
Plasma Atomic Emission Spectrometry (ICP-AES)] was very good (correlation coecient
greater than or equal to 0.9) (Piorek, 1994). Cole et al. (1993) report on an extensive lead
study conducted during the summer of 1991 in Leadville, Colorado, where over 3700
samples of soil were analyzed for lead using portable XRF analyzers. The accuracy of the
lead determination by the XRF in that study was estimated at about 8% relative. Another
very important observation from that study is that the accuracy of the determination of
spatial distribution of lead in the soil does not depend so much on the accuracy of the
individual result as on the total (large) number of samples. In other words, raising the
analytical quality of the individual data has very little eect on the total uncertainty;
however, increasing the number of samples will considerably improve the accuracy of the
determination of spatial distribution of lead. This conclusion has been conrmed by other
researchers as well (Raab et al., 1991).
The portable XRF analyzer has been successfully used during remediation eorts at
the cleanup of soil contaminated with chromium (Waligora, 1997). More than 3000
samples were measured in the eld, providing an immediate feedback to an excavating
crew which was working literally on the heels of the eld analytical team. Figure 31
illustrates typical performance parameters of the contemporary portable XRF analyzer
with a Si(Li) detector probe for metals in soil matrix. As it can be seen, the typical de-
tection limits are in the range 50100 mg=kg. The success of portable XRF in soil
screening led to coining of a new term of eld-portable x-ray uorescence (FPXRF),
probably to distinguish this application from the others.
The FPXRF analyzer oers an attractive, in situ alternative to the traditional la-
boratory analysis of contaminated soil. The operator equipped with the portable XRF
analyzer can perform a large number of multielement measurements in situ by placing the
probe directly on the soil surface, thus allowing for the decisions to be made in real time.
Moreover, the measurements can also be performed at any stage of the remediation
Copyright 2002 Marcel Dekker, Inc.
Figure 31 Typical detection limits (DLs) for metals in soil obtained with field portable X-ray analyzer equipped with a
Si(Li) detector probe. Conditions: Si(Li) detector, 30 mm
2
area; measurement time 200 s; sample matrix: silica sand. K series
excitation with
109
Cd and
241
Am isotopes. Open rectangles correspond to 1s uncertainty of counting; solid rectangles mark
the DL. The DLs for heavy metals determined via their L series x-rays, such as Pb, Hg, Bi, are numerically similar to those
from Zn to Br. (From Piorek, 1994.)
Copyright 2002 Marcel Dekker, Inc.
process to either verify that the cleanup is accomplished or that the work has to continue.
The results obtained with the XRF analyzer will reveal relative variations of contami-
nant(s), their spatial distribution, and pinpoint locations with elevated contamination
levels (so-called hot spots), thus making sampling plans realistic and ecient. This is
particularly important because any contaminated site is likely to have an extremely het-
erogeneous distribution of hazardous chemicals in the soil. For that reason, the sampling
strategies will mandate that many densely distributed samples are taken, in order to ensure
that hot spots are not missed. This type of analysis, which is referred to as screening,
is allowed by the EPA to have uncertainties as large as 50% relative (EPA, 1987). Most
recently, the ocial recognition of the FPXRF as a valid soil characterization method has
materialized in a form of a new EPA Method 6200, which has been published in June of
1997 (EPA, 1997).
4. Analysis of Liquid HazardousWaste
The analysis of liquid hazardous waste (LHW) is required prior to its treatment in order to
properly set up the treatment process and, also, after the treatment to determine the nal
disposition of the treated waste. Liquid waste is a mixture of solids suspended in a liquid
medium. Solid particles vary considerably in particle size and composition, whereas the
liquid phase is often a mixture of immiscible liquids. The sludge remaining after the
treatment of liquid waste is either transported to a waste dump or, preferably, utilized for
soil amendment in agriculture and forestry because it is often rich in nutrients. However,
depending on the origin of the waste, it may also contain a whole range of heavy metals,
which even at moderately elevated concentrations, are very toxic to plants and animals. In
order to avoid the accumulation of toxic metals in soil, the concentration of these metals in
the original waste, as well as in postprocess sludge, has to be known.
Traditionally, the liquid waste samples are analyzed by a ame or graphite furnace
AAS or ICP-AE. Those methods, while oering low detection limits, are slow and labor
intensive. Additionally, they call for rather involved sample preparation (digestion) and
are relatively expensive. Energy-dispersive x-ray uorescence spectrometry was considered
as an alternative analytical method for liquid waste due to its known features of speed,
multielement character, wide dynamic range, and little to no sample preparation re-
quirement. However, the diversity of form and composition makes LHW an extremely
dicult medium to analyze with XRF as a result of the severe matrix eects present in a
sample of LHW. The severity of matrix eects in liquid hazardous waste precludes the use
of many known calculational models correcting for absorption=enhancement phenomena
because mathematical models are valid only for physically and chemically homogeneous
materials. To overcome these diculties, a novel empirical approach was proposed for
stabilization of the sample matrix by dilution (ASTM, 1997; Piorek et al., 1997), For the
purpose of XRF analysis of liquid hazardous waste, a calcined alumina powder (Al
2
O
3
)
was used as a diluent. This powder is characteristic for its highly adsorptive properties
toward liquids. Thus, after adding a small amount of liquid to an aliquot of Al
2
O
3
, the
latter instantly absorbs the liquid. After brief mixing and shaking, the originally liquid
sample is converted to a homogeneous powder. The mass ratio of the diluent to sample has
to be maintained constant from sample to sample and should not be smaller than 3:1,
otherwise the sample=diluent mixture will behave more like a slurry than a powder. This
kind of sample matrix stabilization is eective for liquids as well as liquids with suspended
solids. By diluting and mixing, both the liquid as well as the suspended solids are now
suspended in alumina powder, which does not allow waste solids to settle. Another
important function of alumina is that it dominates the matrix of the sample and, therefore,
Copyright 2002 Marcel Dekker, Inc.
the composition dierences between the original samples of waste are reduced to easily
manageable levels. For calibration, a set of samples of alumina powder is prepared by
spiking with a known amount of standard solutions. The details of the technique are
described in the ASTM Standard Method D6052-97 (ASTM, 1997). Using a portable
Si(Li)-detector-based probe with a 0.74-GBq (20 mCi)
109
Cd source and a 1.11-GBq
(30 mCi)
241
Am source, the 200-s measurement time detection limits obtained with this
method for elements ranging from Cr to As were between 35 and 10 mg=kg, respectively.
For heavy elements such as Hg, Pb, Cd, and Sb, detection limits were 12, 8, 4, and
4 mg=kg, respectively.
5. Analysis of LeadinWall Paint
A prolonged exposure to low levels of lead may lead to serious health problems in young
children, including impairment of the central nervous system, behavioral disorders, and
mental retardation. Paint, dust, soil and, to a lesser extent, food, water, and air are all
potential sources of exposure to lead. After banning the use of leaded gasoline in the U.S.,
lead-based paint is recognized as the principal source of lead in dust and soil. From here it
can easily enter a childs system as the result of hand-to-mouth activity, typical for young
children. These concerns were addressed with federal regulations in the early seventies.
The essence of those regulations is that any applied lead-based paint exhibiting more than
1.0 mg of lead per square centimeter must be removed, as well as any applied to surface
paint that contains more than 0.5% lead by dry weight. In recent years the lead-based
paint danger became the focus of attention in European countries. France, for example,
introduced guidelines for lead levels in paint and followed with enforcement regulations
(Law No. 99-483 of June 9, 1999). It is expected that other countries will follow Frances
example in coming years.
The preferred method for testing applied paint for lead is FPXRF (McKnight et al.,
1989; Piorek et al., 1995). This is because XRF is nondestructive and fast and oers overall
good performance. Additionally, lead x-rays, either K or L series, have high penetrating
power (especially the Ks), and are not interfered with by the other elements which make up
the paint. As the result, the XRF method is quite selective for lead in leaded paint applied
on any substrate. A number of accounts on this subject has been published over the years,
and a number of portable XRF devices was designed for just lead in wall paint mea-
surements. The majority of the instruments used lead K shell x-rays, excited either with
57
Co or
109
Cd (the 88-keV g-rays of this isotope), but some have utilized L shell excita-
tion of lead. The latter approach has been reported to be able to measure lead with a
proportional detector and a 2.1-GBq (60 mCi)
244
Cm source with a precision of
0.05 mg Pb=cm
2
. Probably the most successful overall lead in paint analyzer has been
Model LPA-1 from Radiation Monitoring Devices. The self-contained instrument in a
shape of the lightweight pistol uses a
57
Co isotope and CdTe detector to determine the lead
concentration in applied paint at a 2s condence level in 5 s measurement time. The
analyzer automatically adjusts the measurement time to reach a preset level of total
measurement uncertainty. The instrument is an example of well thought out ergonomical
design, responding well to the needs of the market. The analyzer can be preset for a
specic, local lead abatement action level and it can generate a hard-copy analytical
report. Another interesting entry in the lead in paint market is the Model XL from the
Niton Corporation (1999). Although a marvel of miniaturization, its shape seems less
conforming to the requirements expected of a hand-held device operated continuously for
several hours. This is also the only commercially available analyzer which uses L shell lead
x-rays for analysis and the ratio of the lead La to Lb lines for determination of the depth at
Copyright 2002 Marcel Dekker, Inc.
which leaded paint is present under overcoats of other paint (Shefsky et al., 1997); in those
instances in which leaded paint may be buried under overcoats of paint, the instrument
uses K shell lead lines for measurement, as there are very little aected by top layers of
paint.
I. Analysis of Corrosion Products
An interesting on-line application has recently been reported by Connolly and Millet
(1994) and Connolly and Harvey (1995) for corrosion product monitoring in nuclear
power plant feedwater. By coupling the XRF analyzer probe to a specially designed
sample ow chamber, the authors realized continuous measurement of incremental ac-
cumulation of corrosion products such as iron. The feedwater is circulated through the
sample ow chamber, where it passes through the membrane lter. The solids present in
the water are collected on the lter surface and then measured with an XRF probe. The
measurement is incremental, so that the lter does not have to be removed for analysis
after each sampling cycle. A gas-lled proportional detector and 3.7-GBq (100 mCi)
244
Cm
source allow one to routinely achieve the detection limits for iron of 1 ppb, at a 300-s
sampling period followed by 300 s of measurement time. Further work is under way to
improve these performance and to extend the applicability of this method to other ele-
ments and liquid media.
VII. FUTURE OF RADIOISOTOPE-EXCITEDXRFANALYSIS
As we have stated in the Introduction, radioisotope x-ray analysis is mainly useful in
industrial and eld applications. This trend will continue, although the proportion of on-
line installations for process control to number of eld-portable x-ray analyzers will
constantly decrease. First, it will be the result of a more general contemporary trend in the
instrument industry of bringing the analytical method to the site. Second, the number of
applications that can benet from portable XRF analysis is much greater and more di-
versied than the number of installations for process control. Third, radioisotope-based
on-line installations will gradually be displaced by systems based on x-ray tubes. Last but
not least, on-line installations will face erce competition from new analytical technologies
based on lasers, which, in many cases, oer instantaneous results at a high repetition rate.
As a result, in the very near future, radioisotope x-ray analysis may become synonymous
with eld-portable XRF and, therefore, the future of radioisotope X-ray analysis will most
likely depend on the developments in eld-portable XRF.
The next designs of the portable XRF analyzers will feature new, high-resolution-
semiconductor, room-temperature detectors. Most likely these will be improved detectors
based on silicon diodes (p-i-n detector) or composite semiconductors such as ZnCdTe,
CdTe, GaAs, and so forth. A recently introduced x-ray detector (Xash Detector) built on
a silicon chip as a solid-state drift chamber oers count-rate capabilities well into
50,000 cps at an energy resolution still better than 200 eV (Ro ntec 1997). The portable
XRF analyzer of the future will incorporate a GPS (Global Positioning System) module,
whichif neededwill instantly locate the measurement site on the map andif properly
equippedwill transmit the measurement results via a wireless link to a central data
collection station. An instrument with a GPS chip installed will also be very easy to track
down if misplaced or stolen, a very desirable feature for a radioisotope-containing device.
The FPXRF will travel underground, as an analytical module built into the push-in
probe or in a cone penetrometer. The rst trials were very encouraging (Elam et al., 1997).
Copyright 2002 Marcel Dekker, Inc.
Technological progress will also bring us truly small x-ray sources, which will gradually
replace radioisotopes in applications requiring less than 35-keV-energy x-rays (Moxtek
Inc., private information, 1998). We should also expect an increased number of dedicated,
smart XRF analyzers customized to specic applications, such as the already mentioned
measurement of lead in wall paint. New portable instruments will be smaller, lighter, and
smarter than their predecessors. Perhaps the avor of the future direction of develop-
ment of portable, radioisotope excited XRF is best exemplied by Model XL by Niton
Corp. (Fig. 32), a one-piece, miniature XRF analyzer with built-in ADC=MCA, and
weighting only about 1.6 kg (Niton Corp., 1999).
VIII. CONCLUSIONS
X-ray uorescence spectroscopy is perhaps the rst spectroscopic technique to successfully
enter the eld and plant environments. Radioisotope x- and g-ray techniques have been
widely used for analysis in the laboratory, in industrial plants, and in the eld. Because of
their simplicity, they are preferred to x-ray tube analysis techniques in many applications,
particularly for more routine analyses involving a limited range of sample compositions.
The techniques are less sensitive, however, than those based on x-ray tubecrystal spec-
trometer systems and less exible to widely dierent analysis applications. Radioisotope
x- and g-ray techniques have had great impact on industrial and eld applications. Their use
Figure 32 Niton II, XRF Alloy Analyzer based on a silicon p-i-n diode. (Courtesy Niton Corp.)
Copyright 2002 Marcel Dekker, Inc.
for the continuous on-line analysis in the coating, mineral, coal, paper, and petroleum in-
dustries has led to better control of the industrial processes. This, in turn, has led to products
that more closely meet specications, with consequently large savings in production costs.
Portable XRF analyzers have been around for about 30 years, and during that
period, their design has been rened and their performance greatly improved. A con-
temporary, high-resolution, portable XRF analyzer brings to the eld site not only an
excellent performance often matching that of the laboratory instrument but also un-
surpassed savings in time and labor, contradicting the popular conviction about the in-
herent inferiority of portable instrumentation. The use of portable instruments in the eld
(e.g., for the sorting of alloys and for detection of hazardous waste materials) has led to a
much more rapid analysis of materials and, hence, to the much wider use of x-ray tech-
niques in eld analysis. Field-portable x-ray uorescence is an example of a well-balanced
compromise among portability, ruggedness, reliability, and analytical performance. There
are not that many analytical techniques that can rival FPXRF in simplicity, speed of
operation, and relaxed requirements of sample preparation. Whenever a fast and frequent
elemental analysis for inorganics is required, there is a place for the FPXRF analyzer.
This applies in particular to scenarios in which a quick verication type of analysis is
needed.
APPENDIX: LIST OF COMPANIES THAT MANUFACTURE RADIOISOTOPE-BASED
X-RAYANALYZERS ANDSYSTEMS
AMDL (Australian Mineral Development Laboratories), P.O. Box 114, Eastwood,
SA 5063, Australia
ASOMA Instruments, 11675 Jollyville Road, Austin, TX 78759, USA
BRGM, BP 6009-45060 Orleans, Cedex 02, France
Data Measurement Corporation, P.O. Box 490, Gaithersburg, MD 20877, USA
Detora Analytical, P.O. Box 2747, Alliance, OH 44601-0747, USA
Electrical Engineering and Automation (EMAG), Katowice, Poland
FAG Kugelscher Georg Schafer KG Auf Aktien, Tennenloher Strasse 41,
Erlangen, Germany
Harrison Copper Systems, Inc., AMF Box 22014, Salt Lake City, UT 84122, USA
Harshaw Radiation Detectors, Harshaw=Filtrol Partnership, 6801 Cochran Road,
Solon, OH 44139, USA
Humboldt Wedag, P.O. Box 2729, 4630 Bochum, Germany
Kevex Corporation, P.O. Box 4050, Foster City, CA 94404, USA
Metorex International Oy, P.O. Box 85, SF-02631 Espoo, Finland
Mineral Control Instrumentation Pty. Ltd., P.O. Box 64, Unley, SA 5061, Australia
Mitsubishi Corporation, 6-3, Marunouchi 2-chone, Chiyoda-Ku, Tokyo 100, Japan
Niton Corporation, 900 Middlesex Turnpike, Bldg. 8, Billerica, MA 01821, USA
Outokumpu Oy, PO Box 85, SF-02201, Espoo, Finland
Oxford Instruments, 20 Nueld Way, Abingdon, Oxon OXI4 1TX, UK
Paul Lippke GmbH & Co. KG, Postfach 1760, 5450 Neuwied 1, Germany
Radiation Monitoring Devices, 44 Hunt Street, Watertown, MA 02172, USA
Ramsey Ltd., 385 Enford Rd., Richmond Hill, Ontario, L4C 3G2, Canada
Rigaku Denki, 14-8 Akaoji, Takatsuki-shi, Osaka, Japan
Science Applications Inc., 1257 Tasman Drive, Sunnyvale, CA 94089, USA
Sentrol Systems Ltd., 4401 Steeles Avenue West, North York, Ontario, Canada,
M3N2S4
Copyright 2002 Marcel Dekker, Inc.
Sortex Ltd., Pudding Mill Lane, London E15 2PJ, UK
TN Technologies (Texas Nuclear Corporation), 2555 North Interstate Hwy 35,
Round Rock, TX 78689, USA
Tracor X-ray, Inc., 345 East Middleeld Rd., Mountain View, CA 94043, USA
Yokogawa Hokushin Electric Corporation, 9-32 Nakacho 2-chome, Musashino-shi,
Tokyo, 180, Japan
ACKNOWLEDGMENTS
The author thanks the many scientists, organizations, and companies who generously
supplied information used in this chapter and to the following for permission to reprint
gures used in this chapter: Pergamon Press (Figs. 2, 14, 18, 21, 22, and 28); EG&G Ortec
(Fig. 1); The Radiochemical Centre, Amersham (Fig. 4); the International Atomic Energy
Agency (Fig. 7); The Analyst (Fig. 20); Outokumpu Oy (now Metorex Int.) (Figs. 11 and
23); IEEE (Fig. 16); Automatica (Fig. 25); Australian Mineral Development Laboratories
(Fig. 26); The Australasian Institute of Mining and Metallurgy (Fig. 27); and Acta
Polytechnica Scandinavia (Figs. 29 and 30).
REFERENCES
Amersham. Industrial Gauging and Analytical Instrumentation Sources, Radiochemical Centre,
UK, Catalogue: 1986, pp 15, 5960.
Amptek Inc. Product Summary Catalogue. Bedford MA: Amptek Inc., 1997.
ASTM. Method D6052-97, Standard Test Method for the Preparation and Elemental Analysis of
Liquid Hazardous Waste by Energy Dispersive X-ray Fluorescence. Philadelphia: ASTM, 1997.
Berry PF. Developments in Design and Application of Field-Portable XRF Instruments for On-Site
Alloy Identication and Analysis. 40th ASNT National Fall Conference, Atlanta, 1981.
Berry PF, Garber W, Blake K. Proceedings of 11th Annual Mining and Metallurgy Industries
Symposium, Tucson, AZ, 1983, pp 3142.
Berry PF, Voots GR. In Boogaard, J, Van Dijk GM, eds. Non-Destructive Testing (Proceedings of
the 12th World Conference on Non-Destructive Testing, Amsterdam, The Netherlands, April
2328, 1989). Amsterdam: Elsevier Science Publishers, 1989, p 737.
Chappell RW, Davis AO, Olsen RL. Proc. National Conf. On Management of Uncontrolled
Hazardous Waste Sites, Washington, DC, 1986, p 115.
Cierpicz S. In Gamma, X-ray and Neutron Techniques for the Coal Industry. Vienna: IAEA, 1986,
p 149.
Cole WH, Kuharic CA, Singh AK, Gonzales D, Melon AB. An X-Ray Fluorescence Survey of Lead
Contaminated Residential Soils in Leadville, Colorado, Case Study. EPA Report
EPA=600=R-93=073, Washington, DC: Oce of Research and Development, 1993.
Connolly D, Millett P. Ultrapure Water J 2:61, 1994.
Connolly D, Harvey S. Ultrapure Water J 11:28, 1995.
Cuzin M. Nucl Instrum Methods A253:407, 1987.
David M. Geostatistical Ore Reserve Estimation. Amsterdam: Elsevier, 1977.
EG&G Ortec. Catalogue: Nuclear Instruments and Systems. EG&G Ortec, 1986, pp 316, 320.
EG&G Ortec. Portable Assay Instruments for Detection and=or Measurement of Ore Values,
EG&G Ortec, undated.
Elam WT, Adams J, Hudson KR, McDonald B, Eng D, Robitaille G, Aggarwal I. In Proceedings of
the Fifth International Symposium on Field Analytical Methods for Hazardous Wastes and
Toxic Chemicals, Pittsburgh: Air and Waste Management Association, 1997, p 681.
Copyright 2002 Marcel Dekker, Inc.
Ellis WK, Fookes RA, Gravitis VL, Watt JS. Int J Appl Radiat Isot 20:691, 1969.
EPA. US EPA Data Quality Objectives for Remedial Response ActivitiesDevelopment Process,
EPA=540=G-7=003. Washington, DC: EPA, 1987.
EPA, Method 6200, Field Portable X-Ray Fluorescence Spectrometry for the Determination of
Elemental Concentrations in Soil and Sediment. Washington, DC: EPA, 1997.
Fauth G, Leininger D, Ludke H. In Gamma, X-ray and Neutron Techniques for the Coal Industry.
Vienna: IAEA, 1986, p 165.
Florkowski T, Piorek S. Nukleonika 19 (10):891, 1974.
Fookes RA, Gravitis VL, Watt JS. Anal Chem 47:589, 1975.
Fookes RA, et al. Int J Appl Radiat Isot 34:63, 1983.
Gaskill A, Estes ED, Hardison DL. Evaluation of Techniques for Determining Chlorine in Used
Oils, Vol. 1. Research Triangle Institute Project Number 472U-3255-05, 1987.
Gravitis VL, Greig RA, Watt JS. Australas Instit Min Metall Proc 249:1, 1974.
Gravitis VL, Watt JS, Muldoon LJ, Cochrane EM. Nucl Geophys 1:111, 1987.
Gy PM. In Developments in Geomathematics, Vol. 4. New York: Elsevier, 1982.
Hietala M, Viitanen J. In Rood CO et al., eds. Advances in X-ray Analysis, Vol. 21. New York:
Plenum Press, 1978, p 193.
Hill RF, Garber W. IEEE Trans Nucl Sci NS-25:790, 1978.
Homann P. Fresenius Z Anal Chem 323:801, 1986.
ICRP. Quantitative Bases for Developing a Unied Index of Harm. Annals of the ICRP Publication
45, Oxford: Pergamon Press, 1985.
Ingamells CO, Pitard FF. Applied Geochemical Analysis. New York: WileyInterscience, 1986.
Isaaks EH, Srivastava RM. An Introduction to Applied Geostatistics. New York: Oxford University
Press, 1989.
Jarvinen M-L, Sipila H. In Cohen et al., eds. Advances in X-ray Analysis, Vol. 27. New York:
Plenum Press, 1984a, p 539.
Jarvinen M-L, Sipila H. IEEE Trans Nucl Sci NS-31:356, 1984b.
Jenkins R, Gould RW, Gedcke D. Quantitative X-ray Spectrometry. New York: Marcel Dekker,
1981a.
Jenkins R, Gould RW, Gedcke D. Quantitative X-ray Spectrometry. New York: Marcel Dekker,
1981b, p 94.
Jenkins R, Gould RW, Gedcke D. Quantitative X-ray Spectrometry. New York: Marcel Dekker,
1981c, p 132.
Jenkins R, Gould RW, Gedcke D. Quantitative X-ray Spectrometry. New York: Marcel Dekker,
1981d, pp 120, 129, 192.
Johnson G, Kalnicky D, Lass B. Ind Finishing 1, Jan. 1989.
Kelha V, Luukkala M, Tuomi T. Acta Polytech Scand Appl Phys Ser 138:90, 1983.
Knoll GF. Radiation Detection and Measurement. 3rd ed. New York: Wiley, 1999.
Law No. 99-483 of June 9, 1999, J.O., June 11, 1999, p 8544 (French Law).
Leppala A, Koskinen J, Leskinen T, Vanninen P. Trans Soc Mining Eng. AIME 250:261, 1971.
Lucas-Tooth HJ, Price BJ. Metallurgia 54(363):149, 1962.
Madden NW, Hanepen G, Clark BC. IEEE Trans Nucl Sci NS-33:303, 1986.
Makov VM, Losev NF, Pavlinski GV. Zavod Lab 34(12):1459, 1968.
Martin A, Harbison SA. An Introduction to Radiation Protection. London: Chapman & Hall, 1986.
McKnight ME, Byrd WE, Roberts WE, Lagergren ES. Methods for Measuring Lead Concentra-
tions in Paint Films. NIST Report NISTIR 89-4209, NIST, December 1989.
Niton Corp. Niton II Alloy Analyzer, Product Brochure. Billerica, MA: Niton Corp., 1999.
Outokumpu Mintec Oy. Beltcon 100 GS: Technical Description. Espoo, Finland: Outokumpu
Mintec Oy, 1986.
Piorek S. Adv X-ray Anal 32:239, 1989.
Piorek S. In Proceedings of the 1994 Symp. on Radiation Measurements and Applications, 1994,
p 528.
Piorek S. Field Anal Chem Technol 1(6):317, 1997.
Copyright 2002 Marcel Dekker, Inc.
Piorek S, Rhodes JR. In Proceedings of ISA-1986, Houston, 1986, p 1355.
Piorek S, Rhodes JR. In Proceedings of 15th Environmental Symposium, Long Beach, CA, 1987,
pp 292297.
PiorekS, Pasmore JR. InProc. of the 3rdInt. Symp. onFieldScreening Methods for Hazardous Wastes
and Toxic Chemicals. Pittsburgh: Air and Waste Management Association, 1993, p 1135.
Piorek S, Piorek E. Measurement of Sulfur in Diesel Fuel Oil in the Range Below 1000 mg=kg Using
Bench Top Energy Dispersive X-ray Analyzer. 1993 Gulf Coast Conference, Corpus Christi,
TX, 1993.
Piorek S, Pasmore JR, Lass BD, Koskinen J, Sipila H. In Breen JJ, Stroup CR, eds. Lead Poisoning,
Exposure, Abatement, Regulation. Boca Raton, FL: CRC Press, 1995, p 127.
Piorek S, Piorek E, Johnson G. In Proceedings of the Fifth International Symposium on Field
Analytical Methods for Hazardous Wastes and Toxic Chemicals. Pittsburgh: Air and Waste
Management Association, 1997, p 842.
Piorek S, Ojanpera J, Piorek E, Pasmore JR. In Green RE, ed. Nondestructive Characterization of
Materials VIII, New York: Plenum Press, 1998, p 461.
Pitard FF. Pierre Gys Sampling Theory and Sampling Practice. 2nd ed. Boca Raton, FL: CRC
Press, 1993.
Raab GA, Enwall RE, Cole III WH, Faber ML, Eccles LA. In Simmons MS, ed. Hazardous Waste
Measurements. Ann Arbor, MI: Lewis Publishers, 1991, p 159.
Rhodes JR. Analyst 91:683, 1966.
Rhodes JR. In X-ray and Electron Probe Analysis. Philadelphia: ASTM, 1971, p 243.
Rhodes JR, Rautala P. In Clayton CG, ed. Nuclear Geophysics Oxford: Pergamon Press, 1983,
p 333.
Rhodes JR, Pradzynski AH, Sieberg RD. ISA Trans 11(4):337, 1972.
Ro ntec. Xash Detector, Ro ntec Info, No 4. Berlin: Ro ntec GmbH, 1997.
Saarhilo K. Experiences of a New On-Stream X-ray Analyzer in a Metal Renery. In Proceedings of
IFAC Automation in Mining, Mineral and Metal Processing, Helsinki, 1983, pp 357367.
Schatzler HP. Int J Appl Radiat Isot 30:115, 1979.
Schlesinger TE, James RB, eds. Semiconductors for Room-Temperature Nuclear Detectors,
Semiconductors and Semimetal Series. Vol. 43. New York: Academic Press, 1995.
Scitech Corporation, MAP Portable Assayers: Specications and Technical Information, Kenne-
wick, WA: Scitech Corporation, undated.
Shefsky S. Proceedings of the Fifth International Symposium on Field Analytical Methods for
Hazardous Wastes and Toxic Chemicals. Pittsburgh: Air and Waste Management Associa-
tion, 1997, p 195.
Shefsky S, Pesce J, Martin K. Proceedings of the Fifth International Symposium on Field Analytical
Methods for Hazardous Wastes and Toxic Chemicals. Pittsburgh: Air and Waste Manage-
ment Association, 1997, p 490.
Siegbahn K. ed. Alpha-, Beta- and Gamma-ray Spectroscopy. Amsterdam: North-Holland, 1965.
Vol. 1, p 38.
Sowerby BD. Gamma, X-ray and Neutron Techniques for the Coal Industry. Vienna: IAEA, 1986,
p 131.
Spatz R, Lieser KH. Fresenius Z Anal Chem 288:267, 1977.
Spiegel FX, Horowitz E. Instruments for the Sorting and Identication of Scrap Metal. Baltimore,
MD: The Johns Hopkins University, Center for Material Research, 1981.
Waligora MK. Proceedings of the Fifth International Symposium on Field Analytical Methods for
Hazardous Wastes and Toxic Chemicals. Pittsburgh: Air and Waste Management Associa-
tion, 1997, p 815.
Watt JS. Int J Appl Radiat Isot. 23:257, 1972.
Watt JS. Practical Aspects of Energy Dispersive X-ray EmissionSpectrometry. Vienna: IAEA, 1978, p135.
Watt JS. In Clayton CG, ed. Nuclear Geophysics. Oxford: Pergamon Press, 1983, p 309.
Watt JS. Proc Aust IMM 290:57, 1985.
Watt JS, Stener EJ. Int J Appl Radiat Isot 36:867, 1985.
Copyright 2002 Marcel Dekker, Inc.
8
Synchrotron Radiation-Induced
X-ray Emission
Keith W. Jones
Brookhaven National Laboratory, Upton, New York
I. INTRODUCTION
Elemental analysis using the emission of characteristic x-rays is a well-established scientic
method. The success of this analytical method is highly dependent on the properties of the
source used to produce the x-rays. X-ray tubes have long existed as a principal excitation
source, but electron and proton beams have also been employed extensively. The devel-
opment of the synchrotron radiation x-ray source that has taken place during the past 40
years has had a major impact on the general eld of x-ray analysis. Even after 40 years, the
science of x-ray analysis with synchrotron x-ray beams is by no means mature. Im-
provements being made to existing synchrotron facilities and the design and construction
of new facilities promise to accelerate the development of the general scientic use of
synchrotron x-ray sources for at least the next 10 years.
The eective use of the synchrotron source technology depends heavily on the use of
high-performance computers for analysis and theoretical interpretation of the experi-
mental data. Fortunately, computer technology has advanced at least as rapidly as the
x-ray technology during the past 40 years and should continue to do so during the next
decade. The combination of these technologies should bring about dramatic advances in
many elds where synchrotron x-ray science is applied.
A short summary of the present state of the synchrotron radiation-induced x-ray
emission (SRIXE) method is presented here. Basically, SRIXE experiments can include
any that depend on the detection of characteristic x-rays produced by the incident x-ray
beam from the synchrotron source as they interact with a sample. Thus, experiments done
to measure elemental composition, chemical state, crystal structure, and other sample
parameters can be considered in a discussion of SRIXE. It is also clear that the experi-
mentalist may wish to use a variety of complementary techniques for study of a given
sample. For this reason, a discussion of computed microtomography (CMT) and x-ray
diraction is included here.
It is hoped that this present discussion will serve as a succinct introduction to the
basic ideas of SRIXE for those not working in the eld and possibly help to stimulate new
types of work by those starting in the eld as well as by experienced practitioners of the art.
Copyright 2002 Marcel Dekker, Inc.
The topics covered include short descriptions of (1) the properties of synchrotron
radiation, (2) a description of facilities used for its production, (3) collimated micro-
probes, (4) focused microprobes, (5) continuum and monoenergetic excitation, (6) de-
tection limits, (7) quantitation, (8) applications of SRIXE, (9) CMT, and (10) chemical
speciation using x-ray absorption near-edge structure (XANES) and extended x-ray
absorption ne structure (EXAFS). An eort has been made to cite a wide variety of
work from dierent laboratories to show the vital nature of the eld.
There are many review articles and books that cover all aspects of the production
and use of synchrotron radiation (Kim, 1986; Kunz, 1979; Margaritondo, 1988; Winick
and Doniach, 1980). The early article on SRIXE by Sparks (1980) is a useful source of
information on various aspects of the use of SRIXE as part of a high-resolution x-ray
microscope (XRM) system. Several more recent review articles have covered new devel-
opments (Chen et al., 1990; Jones and Gordon, 1989; Kno chel, 1990; Vis, 1990; Vis and
van Langevelde, 1991).
Microscopy using x-rays can be carried out in several ways. A large eort is
aimed at producing extremely high-resolution maps of the linear attenuation coecient
for low-energy x-rays in biological materials (Sayre et al., 1988). Another approach is
based on the detection of electrons emitted from the specimen (Ade et al., 1990).
Elemental detection can be accomplished by mapping above and below the absorption
edge for the element. Rarback et al. (1987) have shown that this may be a preferable
method for elements lighter than calcium because of the small values of the uores-
cence yield. This approach has been used by Kenney et al. (1985) to study the
calcium distribution in a bone specimen using the Ca L absorption edge for the image
formation.
Image formation using uorescent x-rays is advantageous for the high-Z elements in
terms of attaining the best possible values for minimum detection limits (DLs) combined
with the best possible spatial resolution. It is not possible to use x-ray detection in some
situations where the thickness of the specimen is much larger than the absorption depth of
the uorescent x-rays. In this case also, the use of the above- and below-edge imaging
using the techniques of computed microtomography is possible. In the discussion that
follows, the use of all these methods is disscussed although the major emphasis is placed
on SRIXE.
Images may also be formed using other types of information obtained through
probing the sample. X-ray diraction can be used to determine crystal structure. In-
formation on the chemical state of the elements and position in the lattice can be obtained
by use of XANES or EXAFS. These are, of course, methods widely applied with both
conventional and synchrotron x-ray sources. A merging of these methods with technol-
ogies for producing micrometer-sized beams is a natural direction for the development of
the instrumentation to follow. This is beginning to happen now, and the rate of devel-
opment is bound to increase in the future.
No matter what mode is used for image formation or to characterize a single
volume element in a specimen, the work is dominated by the fact that only limited
numbers of atoms are present to produce a signal. As an example, consider the number
of zinc atoms present in an organic material as a function of the specimen volume
probed for a constant weight fraction of one part per million (ppm). Values are shown
in Figure 1, and it can be seen that there are only 10
4
atoms present in a 1-mm
3
volume.
The detection of such a small number is a technical challenge no matter what mode of
detection is used.
Copyright 2002 Marcel Dekker, Inc.
II. PROPERTIES OF SYNCHROTRONRADIATION
It has been known for almost 100 years that the acceleration of a charged particle will
result in the radiation of electromagnetic energy. The development of the betatron and
synchrotron electron accelerators 50 years later led to the experimental observation of
radiation from electrons circulating in a closed orbit (Elder et al., 1947) and to naming it
specically synchrotron radiation after the accelerator used to produce it. The rst re-
cognition of the unique properties of the synchrotron radiation by Tombulian and
Harman (1956) then brought about an explosive development of activity in constructing
improved sources for production of the radiation and using the radiation in experimental
science.
The original synchrotrons were designed for use in nuclear physics research. Later,
facilities were designed to optimize the conditions for production of x-rays. The compo-
nents of the new facilities include a source of electrons or positrons and an accelerator to
produce high-energy beams. This might be done by the use of a linear accelerator to
produce energies of around 100 MeV. These beams are then injected into a synchrotron
and boosted to energies in the GeV range. Finally, the beam is stored in the same accel-
erator used to attain the nal energy or in a separate storage ring. An acceleration eld in
the ring is used to supply energy to the beam to compensate for the radiation energy loss.
Figure 1 The number of zinc atoms contained in a given volume element is plotted as a function
of volume. A constant zinc concentration of 1 ppm contained in an organic matrix with a density of
1 g=cm
3
is assumed. This shows that the detection of trace amounts of an element at a given
concentration level becomes increasingly difficult as the probe decreases. (From Jones KW, Pounds
JG. Biol Trace Element Res 12:3, 1987.)
Copyright 2002 Marcel Dekker, Inc.
The lifetime of the stored beam is many hours, so that, in practice, the synchrotron source
is almost starting to become similar to a standard x-ray tube in use.
The synchrotron-produced x-ray beams have unique properties that make them
desirable for use. They have a continuous energy distribution so that monoenergetic
beams can be produced over a wide range of energies. The photons are highly polarized in
the plane of the electron beam orbit, which is extremely important for background re-
duction in SRIXE-type experiments in particular. The x-rays are emitted in a continuous
band in the horizontal direction but are highly collimated in the vertical direction. It is
therefore possible to produce intense beams with little angular divergence. The source size
is small and, as a result, the production of intense beams of small area is feasible. The
synchrotron source is a pulsed source because of the nature of synchrotron-type accel-
erators. The x-rays are produced in narrow bursts, less than 1 ns in width, and have a time
between pulses of around 20 ns or more.
The main parameters of interest in dening the synchrotron source are as follows:
1. Magnitude of the stored electron=positron current. Typically, currents are in the
range from 100 to 1000 mA. The lifetimes of the stored beam are many hours.
The lifetimes for stored electrons at the National Synchrotron Light Source
(NSLS) at Brookhaven National Laboratory, (Upton, NY) have typical values
of around 24 h. At Laboratoire pour LUtilisation de Radiation Electromagne-
tique (LURE), Orsay, France, where positron beams are used, the lifetimes are
even longer because the positive beam does not trap positively charged heavy
ions produced from the residual gases in the vacuum chamber.
2. Assuming that the size and angular divergence of the electron beam are not
important, the source brightness is dened as [in units of photons=s=mr
2
=(0.1%
bandwidth)]:
d
2
I
dy dc
1:327 10
13
E
2
GeViAH
2
o
o
c
1
where E is the energy of the electron beam (in GeV), i the current (in A), H
2
(o=o
c
) is a function tabulated in Kim (1986), o is the photon angular frequency,
and o
c
is the critical frequency which splits the emitted power into halves and is
given by the expression 3g
3
c=2r, where g is the electron energy in units of the
electron rest mass, c is the velocity of light, and r is the radius of curvature of the
electron path. The angles y and c are the angles of emission in the plane of the
electron orbit and perpendicular to that plane, respectively.
3. The total photon emission is found by integrating over c and is given by [in units
of photons=s
1
=mr
1
=(0.1% bandwidth)]
dI
dy
2:457 10
13
EGeV iA G
1
o
o
c
2
For electron beams with nonzero emittance (nite area and angular divergence), it is
necessary to dene another quantity, the brilliance, which is the number of photons
emitted into angular intervals dy and dc at angles y and c from an innitesimal source
area (in units of photons=s
1
=mr
2
=mm
2
=(0.1% bandwidth)]. The values of brilliance
and brightness are important in evaluating the performance expected for focused and
collimated x-ray microscopes. Kim (1986) provides a detailed discussion of the emittance
eects, polarization, and performance of wiggler and undulator insertion devices.
Copyright 2002 Marcel Dekker, Inc.
Values attained for these quantities at the 2.5-GeV x-ray ring of the NSLS are shown
in Figures 24. Its x-ray beams are typical of those produced by second-generation syn-
chrotron storage rings. The ring energy is high enough to produce x-rays over an energy
range sucient to produce K x-rays from elements to about Z40 with good eciency
and L x-rays throughout the periodic table. Thus, it is highly suitable for use as the basis of
a system for x-ray microscopy-based SRIXE.
The brilliance obtained from two third-generation synchrotrons is shown in Figures 5
and 6 as a function of the photon energy for several types of devices. The brilliance
obtained from an undulator insertion device at the 7-GeV Argonne National Laboratory
Advanced Photon Source is shown in Figure 5. The brilliance for dierent bending
magnet, undulator, and wiggler devices at the 8-GeV Spring-8 facility in Japan is given in
Figure 6. It can be seen that the third-generation storage rings have gained approximately
three to four orders of magnitude in brilliance when compared to that found from bending
magnets at the NSLS. Similar gains are found for third-generation rings operating at lower
energies although the spectrum of x-rays is naturally aected by the energy of the stored
electron=positron beam. This increase translates to a gain in x-ray intensity that enables
new types of x-ray uorescence experiments.
The high degree of linear polarization of the x-rays from the synchrotron source is a
major factor in making the synchrotron XRM a sensitive instrument. The physics describing
the interaction of polarized x-rays with matter is therefore an important topic. Hanson (1990)
Figure 2 The brilliance of the NSLS x-ray ring is plotted as a function of the x-ray energy
produced. The brilliance is important in defining limitations on production of x-ray images using
focusing optics. (From Jones KW, et al. Ultramicroscopy 24:313, 1988.)
Copyright 2002 Marcel Dekker, Inc.
has carried out an extensive examination of the scattering problem and given methods for
assessing particular geometries used in the XRM. A second point is the continuous x-ray
energyspectrumextending fromenergies inthe infraredtohardx-rays withenergies of several
hundred kiloelectron volts. The tunable energy is important for many types of experiments.
III. DESCRIPTIONOF SYNCHROTRONFACILITIES
There are nowmany synchrotron facilities located around the world that are suitable for use
in various types of x-ray spectrometry measurement. They fall roughly into three classes.
First-generation synchrotrons were built primarily as high-energy physics machines
and were used secondarily for synchrotron radiation production. The Cornell High En-
ergy Synchrotron Source (CHESS) at Ithaca, NY is an example.
Second-generation synchrotrons were optimized as radiation sources, and as a result,
produce x-ray beams with superior brilliance and brightness characteristics. The two rings
of the Brookhaven NSLS fall in this category.
Finally, the third generation of synchrotrons is now being designed. They will be the
rst sources intended to incorporate insertion devices, wigglers, and undulators in the design
phase. In some cases, the ring energy is increased to give better performance with the in-
sertion devices. Synchrotrons designed with these features will begin operation in the latter
part of the 1990s. The European Synchrotron Radiation Facility (ESRF) at Grenoble,
Figure 3 The brightness [defined in Eq. (1)] of the NSLS x-ray ring as a function of the x-ray
energy produced. The brightness is of importance in defining the usefulness of x-ray beams produced
by use of a collimator. (From Jones KW, et al. Ultramicroscopy 24:313, 1988.)
Copyright 2002 Marcel Dekker, Inc.
France, the Super Photon Ring-9 GeV (SPring-8) in Kansai, Japan, and the Advanced
Photon Source (APS) at Argonne National Laboratory, Argonne, IL, illustrate this case.
A listing of a number of synchrotron laboratories producing high-energy x-ray
beams suitable for use in SRIXE is given in Tables 13 (Fuggle, 1990; Jackson, 1990;
Winick, 1989, 1990). The number of XRM beam lines is growing rapidly and their em-
ployment for research as a result is becoming widespread.
IV. APPARATUS FOR X-RAYMICROSCOPY
The apparatus that is used for x-ray microscopy measurements varies substantially from
laboratory to laboratory. A schematic diagram of the components of a very compre-
hensive system is shown in Figure 7. All the components are not necessarily used in a
specic instrument. The most important dierences between systems lies in the treatment
of the incident beam. The simplest approach is to use the white beam (full-energy
spectrum) and a collimator. The more complex systems use focusing mirrors to collect
more photons and demagnify the beam and monochromators to produce monoenergetic
beams. At present, the performance of the various systems is quite comparable in terms of
spatial resolution and detection limits (DLs). Thus, a versatile and exible approach to the
Figure 4 The x-ray flux produced by the NSLS x-ray ring is plotted as a function of the x-ray
energy produced. The values are given after integration over the vertical opening angle of the beam.
The definition of the flux is given in Eq. (2). (From Jones KW, et al. Ultramicroscopy 24:313, 1988.)
Copyright 2002 Marcel Dekker, Inc.
Figure 5 The brilliance obtained from an undulator insertion device at the 7-GeV Argonne
National Laboratory Advanced Photon Source. (From Advanced Photon Source Research,
ANL=APL=TB-31 1(1):5, 1998. Argonne National Laboratory, managed and operated by The
University of Chicago for the U.S. Department of Energy under Contract No. W-31-109-ENG-38.)
Figure6 The brilliance for different bending magnet, undulator, and wiggler devices at the 8-GeV
SPring-8 facility in Japan. (From Kitamura H. SPring-8 Annual Report 1994, JASRI, 1995, p 47.)
Copyright 2002 Marcel Dekker, Inc.
choice and design of the components is advisable. A number of the dierent instruments
are described briey in the following subsections.
A. Collimated X-Ray Microscopes
A highly eective XRM system can be made by simply collimating the white beam
(continuous energy) of x-rays produced by the synchrotron. This approach has been fol-
lowed mainly by groups at the Hamburg storage ring (Kno chel et al., 1983), HASYLAB,
and at the Brookhaven NSLS (Jones et al., 1988). The use of white radiation is feasible
because the high brightness gives a high ux of photons in a small area, and the high
polarization of the synchrotron beams minimizes scattering from the sample into the de-
tector. White radiation also makes possible ecient multielement detection over a very
broad range of atomic numbers.
The collimators for the instrument can be made from a set of four polished tantalum
strips. The strips can be spaced apart with thin plastic or metal foils to produce apertures
usable to a beam size of 1 mm. Alternatively, the slits can be attached to individual stepper
Table 2 Second-Generation Synchrotron Light Sources
a
Storage ring (Lab) Energy (GeV) Location URL
CAMD (LSU) 1.4 Baton Rouge, USA www.camd.lsu.edu
DELTA (DU) 1.5 Dortmund, Germany prian.physik.uni-dormund.de
NSLS x-ray (BNL) 2.5 Upton, USA www.nsls.bnl.gov
Photon Fac. (KEK) 2.5 Tsukuba, Japan www.kek.jp
Synch. Rad. Source (SRS) 2.0 Daresbury, UK www.dl.ac.uk
Note: Lists of synchrotron sites can be found from many of the home pages listed.
a
This is the rst generation of machines built to be dedicated synchrotron radiation facilities. The use of bending
magnet ports is emphasized although some straight sections have insertion devices (e.g., undulators and wigglers)
[see Winick (1989, 1990), Jackson (1990), Fuggle (1990)].
Table1 First-Generation Synchrotron Light Sources
a
Storage ring (Lab) Energy (GeV) Location URL
ADONE (LNF) 1.5 Frascati, Italy
DCI (LURE) 1.8 Orsay, France www.lure.u-psud.fr
VEPP-3 (INP) 2.2 Novosibirsk, USSR
BEPC (IHEP) 2.22.8 Beijing, China
SPEAR (SSRL) 3.03.5 Stanford, USA www-ssrl.slac.stanford.edu=
welcome.html
ELSA (Bonn Univ.) 3.5 Bonn, Germany www-elsa.physik.uni-bonn.de
DORIS II (HASYLAB) 3.55.5 Hamburg, Germany www-hasylab.desy.de
VEPP-4 (INP) 5.07.0 Novosibirsk, USSR
CESR (CHESS) 5.58.0 Ithaca, USA www.chess.cornell.edu
Acc. Ring (KEK) 6.08.0 Tsukuba, Japan www.kek.jp
Tristan (KEK) 2530 Tsubuka, Japan
Note: Lists of synchrotron sites can be found from many of the home pages listed.
a
Most of these early facilities were built to do research with the primary electron or positron beams. Synchrotron
radiation research was parasitic. Today, all have become at least partly dedicated as synchrotron light sources [see
Winick (1989, 1990), Jackson (1990), Fuggle (1990)].
Copyright 2002 Marcel Dekker, Inc.
motors for producing variable-sized beams of size greater than about 10 mm. The col-
limation approach is ultimately limited by beam-spreading related to nite source and
pinhole dimensions and to diraction.
The collimated XRM (CXRM) was rst operated at the now defunct Cambridge
Electron Accelerator by Horowitz and Howell (1972). They used a pinhole made by the
evaporation of a thick gold layer around a 2-mm quartz ber, followed by subsequent
etching away of the quartz. The x-ray energy was about 2.2 keV, and image contrast was
achieved using determination of the linear attenuation coecient. The specimen was
moved past the incident beam, and uorescent x-rays were detected with a proportional
counter. A spatial resolution of 2 mm was measured with this pioneering apparatus.
Later versions of the CXRM have been put into operation at Hamburg and
Brookhaven. Some of the details of the operations are given to illustrate some of the more
important operational details.
It is often best to maximize the photon ux on the sample by employing a white
beam of x-rays. The ux available at a point 9 m from a NSLS bending magnet source is
shown in Figure 8. The ux from this source, integrated over the entire energy spectrum, is
about 3610
8
photons=s=m
2
with a 100-mA stored electron current.
For best operation of the CXRM, care must be taken in shaping the incident x-ray
spectrum using lters. The inuence of the beam lters has been studied by the Hamburg
group (Kno chel et al., 1983). Their results are shown in Figure 9. By varying the eective
energy, it is possible to tailor the beam to give the best possible DL for a given atomic
number. Filters on the detector can be used in some cases to reduce the eect of a major
Table 3 Third-Generation Synchrotron Light Sources
a
Light source Energy (GeV) Location URL
ALS 1.01.9 Berkeley, USA www-als.lbl.gov
ASTRID 0.58 Aarhus, Denmark www.dfi.aau.dk
BSRF 1.41.55 Beijing, China solar.rtd.utk.edu=$china=ins=
IHEP=bsrf=bsrf.html
CLS 2.52.9 Saskatoon, Canada sal.usask.ca=cls=cls.html
SLS 2.1 Villigen, Switzerland www.psi.ch=sls
SRC 0.81.0 Stoughton, USA www.src.wisc.edu
SRRC 1.3 Hsinchu, ROC 210.65.15.200=en=index.html
INDUS II 1.4 Indore, India
MAX II 1.5 Lund, Sweden www.maxlab.ln.se
BESSY II 1.52.0 Berlin, Germany www.bessy.de
ELETTRA 1.52.0 Trieste, Italy www.elettra.trieste.it
LNLS 2.0 Campinas, Brazil www.lnls.br
PLS 2.0 Pohang, Korea
SIBERIA II 2.5 Moscow, USSR
ESRF 6.0 Grenoble, France www.esrf.fr
APS 7.0 Argonne, USA www.epics.aps.anl.gov
SPring-8 8.0 Kansai, Japan www.spring8.or.jp
Note: Lists of synchrotron sites can be found from many of the home pages listed above.
a
This is the newest generation of dedicated synchrotron radiation facilities. The use of insertion devices (e.g.,
undulators and wigglers) is emphasized, but bending-magnet ports will also be available [see Winick (1989, 1990),
Jackson (1990) Fuggle (1990)].
Copyright 2002 Marcel Dekker, Inc.
element. In experiments that examine the distribution of trace elements in bone, a lter of
polyimide can reduce the high rates caused by the calcium in the bone, but it will not have
a large eect on the x-rays from iron, copper, and zinc.
An equally critical task is the alignment of the energy-dispersive x-ray detector.
Kwiatek et al. (1990) have reported on this phase of the optimization procedure. The
importance of aligning the detector in the horizontal plane can be seen by reference to
Figure 10. In addition, the energy spectrum and degree of polarization change as a
function of the vertical distance from the plane of the electron orbit in the storage ring.
Figure 11 shows the relative photon ux for the two polarization states. Examination of
the curves shows that the alignment needs to be made to an accuracy of better than a few
hundred micrometers to get the best reduction of scattered background.
Results of the experimental background=peak ratio determination as a function of
vertical displacement are shown in Figure 12 for elements from calcium to zinc in a gelatin
matrix and for palladium in a pyrrhotite matrix. The dependence on scattering angle in the
Figure 7 A comprehensive synchrotron beam line designed for use as a x-ray microscope. Not all
the components would be utilized at a given time in practice. The rather varied uses of the system
require a flexible approach to the design of the equipment.
Copyright 2002 Marcel Dekker, Inc.
Figure 9 Values for DLs obtained for a white-light XRM are shown as a function of atomic
number. The change in the DLs as a function of filtering of the incident beam using aluminum filters
is an important feature of this type of arrangement. (From Kno chel et al., 1983.)
Figure 8 Energy distribution of photon flux produced by a bending magnet on the NSLS x-ray
ring at a electron current of 100 mA. (From Jones et al., 1990b.)
Copyright 2002 Marcel Dekker, Inc.
Figure 10 Dependence of incoherent scattering cross sections for x-rays polarized parallel and
perpendicular to the plane of the stored electron orbit on the scattering angle. Observation at a
scattering angle of 90
Z
1
E
ab
NE exp fmEws
fl
EgdE 3
where YE
Z
is the count rate for an element with a characteristic x-ray energy E
Z
; N
A
is
the number of target atoms within a beam spot, e
k
is the detector eciency, dO=4p is the
solid angle, NE is the photon ux [number of photons=(s mm
2
0.1-keV bandwidth)], mE
is the linear attenuation coecient for Al and mE
Z
that for polyimide, w is the Al
thickness, w
d
is the polyimide thickness, s
fl
E is the uorescent cross section, E
ab
is the
energy of the absorption edge, and E is the photon beam energy.
Results of a comparison of measured and calculated rates for a section of gelatin
with known amounts of iron and zinc are shown in Table 4 for three dierent incident
beam spectra. The results of the comparison are excellent for iron, less so for zinc. It is
clear, however, that it is feasible to make determinations of the abundances of the elements
without reference to standards.
B. Focused X-Ray Microscopes
Focused x-ray microprobes (FXRM) have been the subject of great interest over the years.
The rst version appears to have been developed by Sparks (1980) and was used at
Figure 13 The background-to-signal ratio is plotted as a function of scattering angle in the
horizontal plane. The curve was obtained by displacement of the detector and the equivalent angular
range spanned was roughly 4
and 3
. In 1971, Yoneda and Horiuchi (1971) rst took advantage of this eect for x-ray
uorescence (XRF). They proposed the analysis of a small amount of material deposited
on a at totally reecting support. This idea was subsequently implemented in the so-
called total-reection x-ray uorescence (TXRF) analysis which has spread out worldwide.
It is now a recognized analytical tool with high sensitivity and low detection limits, down
to the femtogram range. TXRF is basically an energy-dispersive technique in a special
excitation geometry. This geometry is achieved by adjusting the sample carrier inclined
under angles of about 1 mrad (0.06
cos j
T
cos j
0
6
For x-rays, the real part of n
reflector
is slightly less than 1, whereas the refraction index for
vacuum is 1 (in good approximation also valid for air). Consequently, the reector is the
optically thinner medium than vacuum (air), and the reected beam will be deected
toward the boundary, which is completely dierent as compared to usual light optics. Due
to the fact that the angles for which the total reection eect is observed for x-rays are
small, sine functions can be replaced by their argument sin j
0
!j
0
, and cosine by
1 j
2
0
=2.
The electromagnetic waves for the incoming, reected, and refracted beams can be
dened in the usual way for light optics, and Fresnels formulas will give the ratios of, for
example, the reected E
R
and incoming E
0
amplitudes:
E
R
E
0
j
0
j
2
0
2d 2ib
q
j
0
j
2
0
2d 2ib
q 7
For exact calculations, the polarization state of the primary radiation and the propagation
of the refracted beam as the so-called inhomogeneous wave have to be considered. It
appears, however, that the eects for MoKa primary radiation and a quartz reector are
negligible and can be ignored in the calculation of reection (R) and transmission (T)
coecients as well as of refraction angles j
T
:
Figure 3 The incident (I
0
), the reflected (I
R
), and the refracted (transmitted, I
T
) beams at the
interface of two media. The refraction index of the medium from the radiation comes (usually air or
vacuum) is greater than the one of the reflector.
Copyright 2002 Marcel Dekker, Inc.
R I
R
E
R
E
0
4x
2
T
x
0
x
T
2
y
2
4x
2
T
x
0
x
T
2
y
2
8
T
E
T
E
0
2
x
T
j
0
16x
3
T
x
0
4x
2
T
x
0
x
T
2
y
2
9
The parameters x
T
; x
0
, and y are dened as follows:
x
T
j
T
j
crit
10a
x
0
j
0
j
crit
10b
y
b
d
10c
The reection and transmission coecients fulll the following condition (as a direct
consequence of energy conservation):
R T 1 11
The refraction angle j
T
, given by
j
T
d
p
x
2
0
1
2
y
2
h i
r
x
2
0
1
s
12
is a function of the incidence angle and coincides with it for the angles well above the
critical angle. The refraction angle is very small, but does not vanish for an (hypothetical)
incidence parallel to the interface j
0
0 (see Fig. 4).
Figure 4 The refraction angle j
T
as a function of the incident angle j
0
for MoKa and a quartz
reflector. The critical angle for this configuration is at 1.8 mrad. Even for very small incident angles
close to 0
2d
p
13
and is closed to the position of the respective inection points of the transmission and
reection coecient curves and also of the refraction angle. For energies well above the
reector materials absorption edges, a simple dimensional equation can be derived,
j
crit
mrad
28:8
E keV
Zr g=cm
3
A
s
14
showing that the critical angle is inversely proportional to the energy of the incident ra-
diation, which means that the higher the energy of the incident photons, the smaller the
angle of the reector has to be adjusted to observe total reection. Z is the atomic number,
r is the density, and A is the atomic mass.
The dependence of the reectivity on the critical angle of the reector material is
demonstrated in Figure 5: The lower the atomic number of the reector, the lower its
critical angle and the more step-function-like is its shape. For the combination of MoKa
incident radiation reected on quartz, j
crit
1.8 mrad ( 0.10
).
The Fresnel equations are based on classical dispersion theory and have been derived
assuming a perfectly at and smooth interface between homogeneous media. Even though
a real surface is, in general, rough on a microscopic scale, the experimental results have
shown sucient agreement.
The penetration depth z
P
for total reection, in accordance with the denition for
conventional XRF, is the distance measured normally to the interface at which the
Figure 5 Calculated reflectivities of several reflector materials for MoKa radiation. The values of
the critical angles j
crit
for C, Si, Ge, Nb, and Ta (1.75, 1.8, 2.5, 3.1, and 4.2 mrad, respectively) are
indicated by circles.
Copyright 2002 Marcel Dekker, Inc.
intensity of the (refracted) beam is reduced by a factor of e (i.e., to 37% of its value). It is
directly proportional to the refraction angle (for conventional XRF, this angle is equal to
the angle of incidence):
z
P
1
rt
m
j
T
15
As will be shown later, the penetration depth has no direct inuence on the background
intensity.
Only some years after TXRF was introduced as a spectrometric tool, the interference
eect on top of the reector between the incoming and reected beam was recognized. The
undisturbed coherent superposition of the plane electromagnetic waves results in a vari-
ation of the intensity pattern, depending on the distance above the surface, called a
standing wave.
Figure 6 displays the fundamental facts; the length D gives the distance between two
maxima of the standing wave with wavelength l,
D
l
2j
0
16
and is typically in the range of 10100 nm. The intensity of the standing wave is a function
of the height z above the reector surface and the incident angle:
Ij
0
; z I
0
1 Rj
0
2
Rj
0
p
cos Fj
0
2p
z
Dj
0
!
17
This intensity can vary between zero and four times the primary intensity I
0
. The phase
factor Fj
0
cos F 2x
2
0
1 18
was found by Bedzyk et al. (1989) for the case b 0 (i.e., when absorption was ne-
glected). For standard TXRF where granular residues deposited on a reector are
analyzed, a problem might arise that samples with dierent thickness (but smaller than
D) can give dierent uorescence intensities. For homogeneous residues for which many
of these maxima and minima of the standing waves occur within the sample thickness,
this eect will level out to a large extent. If an internal standard is homogeneously mixed
with the sample material, the ratio of the respective uorescence intensities will lead to
Figure 6 The interference (standing wave) zone between the incident (I
0
) and reflected (I
R
) plane
waves with wavelength l shows nodes and antinodes with a period of D.
Copyright 2002 Marcel Dekker, Inc.
an acceptable precision in any case. Therefore, the use of an internal standard is in-
evitable for TXRF analysis of the residues. (See Figure 7.)
For a broad range of elements, TXRF provides x-ray spectra characterized by high
sensitivities and low background. The background generated by the sample support can be
assessed from the transmitted part of the primary radiation.
One of the main features of TXRF is a strong dependence of both the scattered and
uorescence signals, which originate from the substrate itself, on the incidence angle. The
intensity of the scattered radiation I
scatter
(j
0
) is given by
I
scatter
j
0
/ j
0
Tj
0
ds
dO
1
m
m
E
0
19
The following factors had been considered:
For standard TXRF, the area on the reector, seen by the detector, is generally
smaller than the area hit by the primary radiation; therefore, a geometry factor
proportional to sin j
0
has to be used (de Boer, 1989) (and the sine can be
replaced by its argument).
Only those primary photons that are not totally reected on the surface are able
to penetrate and induce interactions. This leads to the transmission
coecient T.
Considering the phenomena of elastic and inelastic scattering, the dierential
scattering coecients ds=dO, which depend on the scattering angle (for TXRF
in the range of 90
for TXRF.
Copyright 2002 Marcel Dekker, Inc.
The emitted spectrum consists of the continuum (bremsstrahlung) and superimposed
are the characteristic lines of the anode material (e.g., MoKa and MoKb) (see Fig. 10).
Insertion of spectral modication devices in the path of the primary radiation beam im-
proves the background (see Fig. 11).
A so-called cuto reector, acting as a low-pass lter, suppresses the high-energy
photons (above the K lines of the anode material), whereas the low-energy part of the
primary radiation, including the characteristic lines, should be reected, improving the
backgroundin particular, in the low-energy region. High-energy photons of the con-
tinuum (although not of direct inuence to the background under the lines of interest) can
get to the front side of the detector and can be inelastically backscattered (Compton
backscattering). By-products of such events are the recoil electrons of the scattering atom,
which are registered in the detector. Their energy is a function of the scattering angle and
the primary energy. As a consequence, the spectrum shows a shelf (Compton edge),
starting from zero up to several kiloelectron volts.
The eect of inserting a cuto reector into the beam path between the x-ray tube
(collimation unit) and the sample reector can be seen from the dierence for the high-
energy parts of the spectra in Figures 12a and 12b; almost all photons above 20 keV are
suppressed for the latter. A proper alignment is fairly critical, because, otherwise, either
some high-energy photons can pass the device or, even worse, a part of the needed ex-
citation spectrum is suppressed (see Fig. 12c, where MoKb is drastically reduced in
intensity).
A typical resulting sample spectrum for a well-adjusted TXRF spectrometer (sample
reector also adjusted) is displayed in Figure 13. Unknown samples can be quantied
(referring to a given internal standard element) with sensitivity factors [for denition, see
Eq. (22)] (Fig. 14).
Figure 9 The angular dependence of the background intensity for TXRF measurements (MoKa
primary radiation under total reflection on a quartz reflector; the critical angle is 1.8 mrad). The
dashed line represents the case where no total reflection is observed.
Copyright 2002 Marcel Dekker, Inc.
Monochromators also can modify the primary radiation and they are usually set to
the energy of the most intense characteristic line of the anode material. For a Mo-anode
x-ray tube MoKa or for a W-anode x-ray tube WLb are selected, but a part of the
continuum can be monochromatized as well. Commonly used crystal monochromators
Figure 10 Measured primary spectrum of a fine-focus Mo diffraction x-ray tube (45 kV
acceleration voltage) as typically used for TXRF. The characteristic MoKa and MoKb lines are
superimposed on the bremsstrahlung background. In the energy range between 5 and 10 keV, several
additional lines resulting from anode contaminations and so on were present for this particular tube.
Figure11 Major components of a TXRF spectrometer.
Copyright 2002 Marcel Dekker, Inc.
Figure 12 (a) Scattered spectrum of a Mo x-ray tube (40 kV acceleration voltage) as used in a
TXRF module spectrometer; (b) scattered spectrum, under the identical conditions as above, but
with a cutoff reflector insertedadjusted to suppress primary photons between 20 and 40 keV.
(c) scattered spectrum, under the identical conditions as in (a) and (b), but with the cutoff reflector
slightly de-adjusted. As a consequence, the MoKb line is also suppressed.
Copyright 2002 Marcel Dekker, Inc.
have the disadvantage of a very narrow energy band transmitted (usually in the range of
few electron volts), whereas synthetic multilayer structures are characterized by higher
DE=E and reectivities of up to 75% for premium quality materials. (See Fig. 15.)
The excitation of an element is the most ecient when the energy of the exciting
radiation is close to its absorption edge. The search for anode materials for special ap-
plications has led to the development of Al- or Si-anode x-ray tubes which are suitable for
the determination of elements below Si, such as F, Na, Mg, and Al.
A spectrometer has been developed with a variable double multilayer pair used as a
tunable monochromator combined with an x-ray tube with an anode made of a homo-
geneous alloy of molybdenum and tungsten (Knoth et al., 1997). Three discretely ad-
justable excitation energies (9.7, 17.5, and 35 keV) are supplied by this combination and
cover a much wider range of the elements than the detectable with a single spectrometer.
The detection limits were determined to be 0.4 pg for Ni using WLb excitation, 0.6 pg
for Pb with MoKa excitation, and 8 pg for Cd (determined by using its K lines) with the
35-keV bandpass excitation.
The applicability of three congurations of curved multilayer mirrors as mono-
chromatizing and focusing elements has been studied theoretically (Stoev et al., 1998) and
the predicted theoretical results were compared with those obtained with a prototype
spectrometer (tunable between 9.7 and 17.5 keV, with the above-mentioned W=Mo alloy
anode) for the element Ni (Knoth et al., 1999).
Currently, synchrotron radiation is the most brilliant photon beam for XRF and it
has almost ideal features for TXRF. It is several orders of magnitude brighter than the
output of x-ray tubes and is naturally collimated and linearly polarized in the plane of
the orbit of the electrons of positrons. The spectral distribution is continuous; therefore,
Figure12
Copyright 2002 Marcel Dekker, Inc.
monochromatization is necessary and selective excitation is possible. Because scattering
and absorption of primary photons in air should be avoided, the use of a vacuum
chamber is recommended. As a result, the ArK line disappears from the spectrum and
the absorption of primary and uorescent photons is also avoided. If for any reasons
the measurements in vacuum are not possible, ushing the chamber with helium is
advantageous.
B. Sample Reflectors
For the trace analysis of granular residues, a carrier with high reectivity that serves as a
totally reecting sample support is required. Therefore, the mean roughness should be in
the range of only a few nanometers and the overall atness should typically be less than
l=20 (l 589 nm, the mean wavelength of the visible light). Furthermore, reectors
should be free of impurities so that the blank spectrum should be free from con-
tamination peaks and the carrier material must not have uorescence peaks in the
spectral region of interest. In addition, the carrier material must be chemically inert (also
against strong chemicals, which are often used for the sample preparation), easy to be
cleaned for repeated use. They should be commercially available and inexpensive.
Typically, they are disk shaped with a 30-mm diameter and a thickness of 35 mm. The
carrier materials commonly used are quartz glass in the synthetic form as Synsil, or
Suprasil, fused silica, and elemental Si. Heavier elements like Ti, Nb, and Ta were
Figure13 Spectrum of a multielement standard (25 ng of each element) prepared as residue sample
on a quartz sample reflector (well-adjusted cutoff reflector). Conditions: TXRF module, Mo tube,
50 kV, 40 mA, 100 s measuring time. S
Sr
and DL
Sr
are the sensitivity factor and detection limit for Sr,
respectively.
Copyright 2002 Marcel Dekker, Inc.
recently tested and showed a good performance. Boron nitride, glassy carbon, Plexiglas,
and Perspex (for single use) are suitable (e.g., for the determination of light elements).
All components of a TXRF spectrometer need to be adjusted (at least once) by
translation and rotation stages and tilters, either manually, or by using remote controllers.
To visualize the eect of total reection, one can use a proper charge-coupled detector
(CCD) camera where the x-rays are falling directly on the chip and become visible on a
monitor. In a more traditional approach, the x-rays can be observed on a ZnS screen.
C. Detectors
Total-reection XRF is an energy-dispersive XRF method (see also Chapter 3); the
radiation is measured mainly by Si(Li)-detectors. A good detector oers a high-energy
resolution [Full width at half-maximum (FWHM) in the range of 140 eV at 5.89 keV],
intrinsic eciency close to 1 for the x-ray lines of interest, symmetric peak shapes, and low
contribution to the background. Primarily, incomplete charge collection at the electrodes
leads to low-energy tailing. The detector escape eect creates escape peaks and thus an
increased background in certain spectral regions. An inherent advantage of semiconductor
detectors is the possibility of bringing the detector crystal very close to the sample, which
results in a large solid angle. Light elements emit uorescent lines in the range from 100
to 1000 eV. The usually used Be entrance window would completely absorb them, so new
window materials, oering better transmission characteristics, are used instead.
Figure14 Experimentally determined sensitivities (total) for the K lines of several elements for a
TXRF spectrometer with a cutoff filter. The excitation can be split (theoretically) into the
contributions from the characteristic MoK lines (MoKa Kb) and bremsstrahlung (continuum). The
ratio of the two sensitivity factors depends, among others, on the applied high voltage (in this case
45 kV). Zr and Nb (Z40, 41) have the K absorption edges already above the energy of MoKa.
Copyright 2002 Marcel Dekker, Inc.
The development of Peltier-element-cooled detectors is a step in the direction of
being independent of liquid-nitrogen supply. Energy-dispersive detectors with a Si p-i-n
diode and Peltier cooling are used as light hand-sized units with an acceptable resolution
of 160200 eV.
D. Manufacturers of TXRF Spectrometers
Atomika Instruments (Germany), spectrometers for chemical analysis (ch ) and
wafer surfaces (w); e-mail: sales@atomika.com
Atominstitut (Austria) (ch ); e-mail: wobi@ati.ac.at
Diraction Technology (Australia) (ch ); e-mail: rod@pcug.org.au
Italstructures (Italy) (ch ); e-mail: isinfo@italstructures.com
I.U.T. (Germany) (ch ); e-mail: i.u.t@t-online.de
Rigaku (Japan) (w); e-mail: rinttyo@rigaku.co.jp
Technos (Japan) (w); Fax: 0720-20 2002
IV. CHEMICAL ANALYSIS
In 1971, Yoneda and Horiuchi (1971) published an article with the rst experimental
data from an energy-dispersive x-ray uorescence measurement with a Ge detector in
Figure15 Two spectra of a sample with a total mass of 700 ng, containing Mg, Cl, Sc, and so forth
excited with a Cr tube, obtained with a cutoff reflector inserted into the beam path (Cut-off), as a
spectrum-modifying device, and a multilayer monochromator (ML). The spectral background and
the characteristic lines are smaller for the latter case; also the scattered CrKb line is suppressed.
Copyright 2002 Marcel Dekker, Inc.
total-reection geometry. The sample was a dried spot of 100 ng of Ni from a water
solution of NiCl
2
on an optical at. In 1974, Aiginger and Wobrauschek (1974)
published results using a Si(Li) detector and a Suprasil reector where 5 mL of Cr salts in
aqueous solution were dried and analyzed with a Cu-anode x-ray tube. A more detailed
publication (Wobrauschek and Aiginger, 1975) contained theoretical considerations,
details of the setup, calibration curves, steps for the quantication, and detection limits
of 4 ng for the element Cr. In 1977, Knoth et al. (1977) presented their rst prototype of
a total-reection XRF spectrometer and analytical results for a blood serum analysis.
From that time on, several groups worldwide participated in the further development of
this technique. From 1986 on, regular biannual TXRF user meetings have been held
(Proceedings of the TXRF Conference, 1989, 1991, 1993, 1995, 1997). The rst book
dealing exclusively with TXRF was written by Klockenka mper (1997).
A. Sample Reflectors and Their Cleaning
Various sample carrier materials have been used for chemical analysis with TXRF, as
mentioned in the previous chapter. Nonreecting residues on top of suitable surfaces give
a doubled uorescence intensity, when measurements are made at angles of incidence
smaller than the respective critical angle for the substrate, owing to the twofold excita-
tionby the direct beam (1) and the reected (2) one (see Fig. 16). The reector must be
long enough so that the latter can fully contribute.
The high sensitivity makes cleaning of the sample carriers critical, particularly in
routine operations when large numbers of them are in use. The following cleaning pro-
cedure has proven to be eective.
1. Mechanical removal of sample residues from previous analysis using tissue or a
brush.
2. Rinsing with water or, in case of organic residuals, with acetone.
3. Gentle boiling for 1 h in a detergent bath (e.g. diluted Extran neutral or acidic),
preferably in special carrier supports from Teon, in order to avoid scratching or
even breaking.
4. Gentle boiling for 1 h in diluted nitric acid (p.a. grade).
5. Rinsing with ultrapure water and subsequent drying.
Figure16 Twofold excitation of the sample for TXRF by the primary beam (1) and the reflected
beam (2), for a case where the reflector is long enough so that beam (2) is observed.
Copyright 2002 Marcel Dekker, Inc.
6. Nonhydrophobic surfaces, such as quartz, must be coated with a silicone
solution (e.g., available from Serva) and subsequently dried at 100
C for about
1 h to keep aqueous samples within a small spot.
7. After cleaning, a blank spectrum should be measured for each reector.
B. Special Sample PreparationTechniques for TXRF
Numerous samples have been prepared for TXRF by using sample preparation tech-
niques, which are basically the same as those applied for atomic absorption spectro-
metry (AAS) or ICPOES. There are, however, special cases where suitable sample
preparation techniques are required to take full advantage of the particular features of
TXRF (e.g., the capability of analyzing small samples and the advantage of an inert
sample carrier).
An example is the direct collection of air particulates on the sample carrier with the
aid of size-separating samplers (impactors). Plexiglas carriers can be used. If the collected
air is wet, the aerosols are reliably deposited on the carrier; when the air is dry, they can
be bounced o. This drawback can be prevented by coating the reectors with a thin lm
of suitable grease (Salva et al., 1993). The reectors loaded with collected aerosols can be
directly analyzed after adding an internal standard. Detection limits are in the range of
0.1 ng for a 1-h collection time and a sampling volume of 0.5 m
3
. Of course, aerosols also
can be analyzed after collection on a lter and subsequent digestion of the loaded lter
material (see Fig. 17). Due to direct analysis, systematic errors are avoided (Injuk and
Van Grieken, 1995).
Figure 17 Spectrum of an aerosol sample, which was collected in a Batelle impactor (stage 1)
directly on a polycarbonate sample carrier. On top 30 ng of Sc were pipetted as internal standard
element (Cr anode, 30 kV, 30 mA, multilayer monochromator, 500 s measuring time).
Copyright 2002 Marcel Dekker, Inc.
One of the inherent advantages of TXRF is the small sample volume required for
analysis; Only a few nanograms of sample mass (or a few microliters of sample volume)
are required, which might be considered as an advantage compared to some other sensitive
analytical methods for trace analysis.
For liquid samples, the easiest way of performing TXRF is the direct analysis.
A few microliters (of the sample solution mixed with the internal standard) are pipetted
on the sample reector and dried, either in a desiccator, with an infrared lamp, on a hot
plate, or in a (dedicated) oven. The sample support can be utilized to allow chemical
reactions in microliter volumes, for example, in the following way: Up to 100 mL solution
loaded with alkali and alkaline-earth salts are placed on the quartz carrier, previously
rendered hydrophobic by means of a silicone solution. The droplet on the carrier is
spiked with about 5 mL of a 1% sodium dibenzyldithiocarbamate solution. After allowing
the sample to dry for about 30 min, the reector is rinsed with ultrapure water. The
insoluble metal carbamates, including that of the internal standard element, remain xed
on the hydrophobic surface, whereas the soluble matrix is dissolved and removed by the
water. The resulting specimen gives substantially improved detection limits compared to
the unprocessed sample.
The properties of the sample support in combination with the high detection per-
formance of TXRF are also utilized in the analysis of thin sections of tissue (microtome
section). A small piece of tissue is frozen and cut by a microtome in thin sections of about
15 mm thickness. A section is then placed on a sample reector and spiked with an internal
standard. This method was applied to tissue of kidney, liver, and lung. Detection limits of
10 ng=g have been reported (von Bohlen et al., 1988).
C. General Sample Preparation
The detection limits obtained for a special sample depend very much on the sample pre-
paration. Figure 18 gives an overview of various common methods for sample preparation
in TXRF, depending on the kind of sample to be analyzed. Of course, one has to be aware
that sample preparation can cause loss of elements as well as contamination by other
elements, and the sample taken for analysis must represent the whole specimen; therefore,
homogenization might be required.
Solid samples can be crushed and then ground to a ne powder of micrometer grain
size. This powder can be mixed with a liquid to produce a suspension, which can be pip-
etted after adding an internal standard on the sample reector. The pulverized sample can
also be dissolved in a suitable solvent, and after adding the internal standard, an aliquot is
pipetted on the sample reector and dried.
For the decomposition of biological and environmental materials, various methods
have been utilized (e.g. with a low-temperature oxygen plasma asher, followed by dis-
solving the ash in an acid). The most popular method of decomposition of biological and
environmental samples like plants, tissue, sediments, and so forth is the wet digestion in
Teon vessels (Teon bombs) with acids like HNO
3
, HF, HNO
3
HCl, HNO
3
H
2
O
2
,
and so forth, in dierent proportions. Using the hydrouoric acid might be a problem if
quartz glass reectors are used. The use of a microwave oven for heating the Teon
bomb reduces the time of digestion to less than 1 h.
The volume of some sample solutions or any sample containing water can be
reduced by freeze-drying. The sample is frozen and the solvent is evaporated under
vacuum conditions. The dried residue can be dissolved in small volume of acid or wet
digested.
Copyright 2002 Marcel Dekker, Inc.
It is also possible to extract traces of certain elements by phase separation. To
a given volume of sample water solutions at appropriate pH and spiked with an internal
standard, an organic solvent is added and mixed thoroughly. Then, the two phases
are separated. The traces of metal ions stay in the organic phase, whereas the matrix
Figure18 Sample preparation methods for TXRF.
Copyright 2002 Marcel Dekker, Inc.
Table1 Inuence of Sample Preparation Methods on Detection Limits
Preparation
Sample Drying
Freeze-
drying
Chemical
matrix
separation
Open
digestion Ashing Suspension Solution
Pressure
digestion
Freeze-
cutting
Rain, river water 0.13 ng=mL 20100 pg=mL 320 pg=mL 13 ng=mL
Blood, serum Digestion:
230 ng=mL
2080 ng=mL 40220 ng=mL
Air dust, ash,
aerosols
5200 mg=g 10100 mg=g 0.13 mg=g
Air dust on lter 0.620 ng=cm
2
0.26 ng=cm
2
Suspended matter 325 mg=g 10100 mg=g
Sediment 10100 mg=g 15300 mg=g
Powdered
biomaterial
110 mg=g 0.22 mg=g
Fine roots 110 mg=g Digestion:
0.11 mg=g
High-purity acids 550 pg=mL
Tissue, foodstu,
biomaterial
0.55 mg=g
Mineral oil 115 mg=g
Mussel, sh 0.11 mg=g
High-purity water 1 pg=mL
Source: Data from Klockenka mper, 1997.
Copyright 2002 Marcel Dekker, Inc.
elements are left in the inorganic solution. The organic liquid can be directly pipetted
onto the reector. Also, the separation of traces by adding a chelating agent and pre-
cipitating the metal ions is a commonly used technique. The metal complexes are ltered
through a membrane lter and dissolved in a suitable organic solvent.
Generally, the selected sample preparation method has a tremendous inuence on
the obtainable detection limits, which is documented in Table 1. Details on sample pre-
paration procedures can be found in the works of Prange et al. (1989), Klockenka mper
and von Bohlen (1999), Holynska et al. (1996), Dargie et al. (1997).
D. Some Applications
Three main advantages characterize TXRF: simultaneous multielement capability, low
detection limits for many elements, and small sample volume. Additional advantages are
the absence of matrix eects, easy calibration, fast analysis, and comparatively low costs.
Table 2 gives an overview of various kinds of sample that have been already analyzed with
TXRF. Generally all kinds of aqueous or acidic liquids where the liquid matrix can be
evaporated, leaving a small amount on a sample reector, can be analyzed. Oils, alcohols,
whole blood, and blood serum can be analyzed after special treatment.
Typical for TXRF are water samples (Fig. 19). Pure water like rainwater or tap
water can be directly analyzed (Holynska et al., 1998; Barreiros et al., 1997); for riverwater
or seawater as well as wastewater, sample preparation is usually required to remove
Table 2 Applications of TXRF
Environment
Water: Rain water, river water, sea water, drinking water, waste water
Air: Aerosols, airborne particles, dust, fly ash
Soil: Sediments, sewage sludge
Plant material: Algae, hay, leaves, lichen, moss, needles, roots, wood
Foodstuff: Fish, flour, fruits, crab, mussel, mushrooms, nuts, vegetables, wine, tea
Various: Coal, peat
Medicine=biology=pharmacology
Body fluids: Blood, serum, urine, amniotic fluid
Tissue: Hair, kidney, liver, lung, nails, stomach, colon
Various: Enzymes, polysaccharides, glucose, proteins, cosmetics, biofilms
Industrial=technical
Surface analysis: Water surfaces
Implanted ions
Thin films
Oil: Crude oil, fuel oil, grease
Chemicals: Acids, bases, salts, solvents
Fusion=fission research: Transmutational elements Al Cu, iodine in water
Geology=mineralogy
Ores, rocks, minerals, rare earth elements
Fine arts=archeology=forensic
Pigments, paintings, varnish
Bronzes, pottery, jewelry
Textile fibres, glass, cognac, dollar bills, gunshot residue, drugs, tapes, sperm, finger-prints
Copyright 2002 Marcel Dekker, Inc.
suspended matter or the salt content. Detection limits are in the low nanograms per
milliliter range. To reduce the detection limits down to the 10-pg=ml level, freeze-drying
with additional leaching of the residue with HNO
3
has to be applied. To remove the salt
matrix from seawater, complexation, chromatographic adsorption, and subsequent elu-
tion have to be performed. This method was developed by Prange et al. (1985). Suspended
matter separated from ltrates can be analyzed by digesting the lter with HNO
3
(Prange
et al., 1993). For the analysis of sediments, microwave digestion is recommended
(Koopmann and Prange, 1991).
Environmental monitoring can also be done by analyzing appropriate plants like
moss or lichen (biomonitoring). The analysis of plants requires a proper sample pre-
paration. After cleaning, shredding, freeze-drying, and nally pulverizing, the powder is
mixed with HNO
3
and digested in a Teon bomb.
Vegetable oil is recommended to be diluted with toluene and an oil-based standard
should be added as internal standard (Reus, 1991). After pipetting an aliquot on the
sample reector, it is heated up to evaporate the volatile parts of the matrix. Detection
limits are in the range of 320 ng=g. Also, low-temperature oxygen plasma ashing is
suitable; the resulting residue has to be dissolved in HNO
3
.
Both depletion and accumulation of trace elements can inuence the biological
functions of human beings. Whole blood and blood serum, amniotic uid, organ tissue,
hair, and dental plaque have been analyzed by TXRF. Detection limits down to 20 ng=mL
for body uids after microwave digestion (Prange et al., 1989) were reported.
Quality control of ultrapure reagents, like acids, bases and solvents can be performed
by using TXRF (Prange et al., 1991). Crude oils (Ojeda et al., 1993), lubricating oils
(Bilbrey et al., 1987), motor oils (Freitag et al., 1989), and diesel fuel (Yap et al., 1988)
have also been analyzed. Light oils can be diluted with chloroform or toluene, the volatile
Figure 19 Spectrum of 10 mL of the NIST water reference standard SRM 1634c, analyzed in an
Atominstitut TXRF spectrometer (Mo anode, 40 kV, 50 mA, multilayer monochromator, vacuum
conditions, 1000 s measuring time).
Copyright 2002 Marcel Dekker, Inc.
matrix then removed by evaporation, and detection limits in the range of micrograms per
milliliter are achievable. If detection limits of nanograms per milliliter are required, low-
temperature oxygen plasma ashing is necessary. For forensic applications the micro-
analytical capability of TXRF is appreciated, because microsamples are usually available.
In recent years, several specic problems have been investigated, including analysis of hair
samples, glass particles, tape fragments, drug powder, semen traces, gunshot residues, and
textile bers (Ninomiya et al., 1995; Prange et al., 1995). Oil paints were investigated by
analyzing the pigments, which are characterized by a limited number of major elements.
Cotton tips can be used to remove an amount of about 1 mg from the paint and deposit it
on a sample reector (Klockenka mper et al., 1993).
E. Quantification
One of the inherent advantages of TXRF is the fact that the sample forms a thin lm on
the sample reector; thus, no matrix eects have to be considered and the so-called thin-
lm approximation is applicable (both absorption and enhancement eects can be neg-
lected). The intensity of the uorescent radiation for, the Ka line, for example, of an
element i with concentration c
i
in the sample (with mass m) is then given by
IE
i
Ka
Z
E
max
EE
i
abs
I
0
EG
m
sin j
s
i
Ka
Ec
i
f E
i
Ka
eE
i
Ka
dE 20
where E
abs
is the energy of the absorption edge of element i, E
max
is the maximum energy of
the excitation spectrum, E
i
Ka
is the energy of the Ka line of element i, I
0
E is the spectral
distribution of the exciting radiation, G is the geometry factor, s
i
Ka
is the uorescence cross
section for the K shell of element i, fE
i
Ka
is the absorption factor for the uorescence
radiation between the sample and the detector, and eE
i
Ka
is the relative detector eciency
for the energy E
i
Ka
. It is assumed that the sample is always completely irradiated by the
primary radiation.
In this special case, the relation between concentration and uorescence intensity is
linear. The so called sensitivity [counts=(second)(sample mass)] can be dened:
S
i
I
i
c
i
m
21
and S
i
depends only on fundamental parameters and the measuring conditions, which
usually can be assumed to be constant. When an element St is used as the internal standard
the relative sensitivity (sensitivity factors) for elements i dened as
S
i
rel
fE
i
Ka
eE
i
Ka
R
E
max
EE
i
abs
I
0
Es
i
Ka
E dE
fE
St
Ka
eE
St
Ka
R
E
max
EE
St
abs
I
0
Es
St
Ka
E dE
22
can be established experimentally with articially prepared standards (Fig. 20) or calcu-
lated theoretically. The determination of the concentration c
i
of the element i in an un-
known sample spiked with the same internal standard element is then simple:
c
i
I
i
I
St
1
S
i
rel
c
St
23
There is a linear correlation between intensity I
i
and concentration c
i
. The addition of
one element as the internal standard of known concentration to the sample is necessary
Copyright 2002 Marcel Dekker, Inc.
primarily because of the inhomogeneous excitation due to the standing-wave eld above
the sample reector surface, where the sample is positioned. Also, geometric and vol-
umetric errors can be canceled.
A sample is thin if its thickness does not exceed the critical thickness
(Klockenka mper, 1997), which is about 4 mm for organic tissue, 0.7 mm for mineral
powders, and 0.01 mm for metallic smears. Under the assumption that the matrix ab-
sorption for the analyte diers only slightly from that of the internal standard element,
these values can generally be higher by a factor of 10400. For the calculation of these
values, the standing-wave eld was not taken into account. This eect and the sample self-
absorption can lead to contradictory requirements for the sample thickness (de Boer,
1991a).
Figure 21 shows a comparison of detection limits for various analytical methods in
the trace element range, following the work of Klockenka mper (1997). Inductively coupled
plasmamass spectrometry (ICPMS) and INAA are more macro than micro methods. In
comparison, TXRF requires only very small sample volumes (in the range of microliters).
ICPMS provides lower detection limits, but both spectral interferences and matrix eects
make quantication more complex.
V. SURFACE ANALYSIS
In order to remain competitive, the semiconductor industry is being constantly forced to
increase the performance and reduce the cost of integrated circuits by shrinking device
Figure 20 Fitted calibration curve (relative sensitivities) for K lines (Atominstitut TXRF module,
Mo anode) with Ga as internal standard element. The circles represent the measured elements
actually used for the calibration (standard solutions with various concentrations). Other elements in
a sample can also be determined by interpolation (in this case, S, K, Mn, Br, and Rb).
Copyright 2002 Marcel Dekker, Inc.
Figure 21 Relative detection limits of INAA, TXRF, ET-AAS, and ICPMS, applied to trace
analysis of aqueous solutions. A 50 mL specimen was used for TXRF and ET-AAS; 3 mL were
needed for INAA and ICPMS. The individual values are approximated after the work of
Klockenka mper (1997).
Copyright 2002 Marcel Dekker, Inc.
dimensions and increasing the number of devices per unit area on chips. A strong
correlation has been found between the presence of metal contamination on the wafer
surface and process yields. Even very low concentrations of dispersed metals can create
defects at the atomic scale, which lead to leakage current, gate insulator breakdown, or
poor threshold voltage control. All of these can result in device failure or reliability
problems. With decreasing size of devices, the acceptable level of metal contamination
decreases as well. The ability to measure low levels of contamination is crucial for the
development of techniques to be used to remove these contaminations. The Semi-
conductor Industry Association (SIA) 1997 National Technology Roadmap for Semi-
conductors (NTRS; http:==www.sematech.org) projects that the required sensitivities for
the transition metals will be at the level of 3610
8
atoms=cm
2
in the year 2001.
One of the most important applications showing the analytical power of TXRF is
the analysis of impurities on the surface of Si wafers (Na, Al, Ca, Ti, Fe, Ni, Cu, Zn, etc.)
for the semiconductor industry (Knoth et al., 1989; Berneike et al., 1989; Weisbrod et al.,
1991; Schwenke et al., 1992). The sample is the wafer itself, with its polished plane surface
having already the quality required for total reection of x-rays (sometimes except at-
ness). With the attributes of TXRF [nondestructive analysis, multielement capacity,
mapping ability (Fig. 22), and excellent detection limits], this technique dominates over
Figure 22 Wafer mapping for the element Na of 7610 spots (1-mm scan steps, relative
coordinates). Synchrotron radiation excitation (SSRL, Beam line III-4), measuring time for each
pixel: 10 s.
Copyright 2002 Marcel Dekker, Inc.
others in this eld. Presently, the 10
9
-atoms=cm
2
level is reached by TXRF using standard
tube excitation (Atomika TXRF8030W, http:==www.atomika.com) or rotating anode tube
excitation (Philips TREX 630 T=S, http:==www.analytical.philips.com; Rigaku 3750
TXRF). The possibility to map a wafer [i.e., scan several points on the surface and de-
termine the level of contamination as a function of the coordinates (Berneike, 1993)], can
help to locate sources of contaminations uniquely.
To measure surface contaminations on wafers, a special setup is required, without
any front surface contact and with a possibility of making measurements at various
angles (Fig. 23). Correct quantication is strongly related to the glancing angle setting.
Measuring the angle dependence of the uorescence signal allows one to distinguish
the form of the contamination (i.e., if it is lm or particulate type). Figure 24 shows the
uorescence intensities of Si, Sc, and Ni as a function of the incident angle around the
critical angle. The steplike function (Sc) is obtained if the contamination is found in
particles on the surface of the wafer (equivalent chemical analysis TXRF). The peaking
curve (Ni) is obtained when the atoms are evenly distributed within a layer of a few
nanometers thickness placed on the wafer surface. Most of the real samples do not show
one of these extreme cases. Prange and Schwenke, (1992) rst gave some examples of
applications of TXRF for surface analysis. If one does not want to measure the complete
angle-dependent behavior, only a single measurement performed at the operating angle,
where the two curves (Sc, Ni) are crossing, allows accurate quantication. For com-
pleteness, the typical curve for the bulk material (silicon of the wafer) or bulk con-
tamination is given. Usually, the bulk Si signal is used for control of the angular
adjustment.
Figure 23 Typical components of a TXRF spectrometer for wafer analysis with all necessary
degrees of freedom for surface mapping (x and y movement), height control (z movement), and
rotation around the x axis (angle of incidence). The special detector and the evacuable sample
chamber allow also the detection of low-Z elements.
Copyright 2002 Marcel Dekker, Inc.
A. Quantification
For the case of granular residues on a substrate (particulate type), which is equivalent to
chemical analysis by using TXRF, the intensity above the critical angle is constant because
the thin, small sample is completely excited. The intensity doubles at the critical angle in
a steplike fashion and remains at the twofold value down to very small angles due to total
reection,
I
particle
i
j
0
k
particle
I
0
c
i
1 Rj
0
24
following the angular behavior of the reection coecient Rj
0
. This intensity is pro-
portional to the primary intensity I
0
and the interesting area-related concentration c
i
. For
incidence angles j
0
below j
crit
, Eq. (24) can be simplied to
I
particle
i
2 k
particle
I
0
c
i
24a
For buried layers (lm like) in a substrate, the intensity far above the critical angle becomes
constant. The asymptotic behavior of the intensities for the particulate and lmlike con-
taminations can even be equal if both concentration values are the same and the appro-
priate scaling factors k
particle
and k
lm
are chosen. However, the intensity for the buried
layers steadily increases with decreasing angle and can reach (theoretically) the fourfold
value at the critical angle. For the smaller incidence angles, the intensity is going to zero,
according to
I
film
i
j
0
k
film
I
0
c
i
1 Rj
0
j
0
z
P
k
film
I
0
c
i
Tj
0
j
0
j
T
j
0
rt
m
25
Figure24 Fluorescence intensities of Sc, prepared as a residue sample (conventional TXRF), Si of
the sample refractor substrate, as representative of bulk signal, and Ni, prepared as a thin-film
sample. The angles of incidence are normalized to the respective critical angles.
Copyright 2002 Marcel Dekker, Inc.
Z
P
is the penetration depth as dened in Eq. (15). The angular behavior is caused by the
countereect of the transmission coecient Tj
0
and the refraction angle j
T
.
The intensities of the characteristic x-rays for homogeneously distributed con-
taminations and major constituents(s) in an innitely thick and at substrate can be
described by
I
bulk
i
j
0
k
bulk
I
0
c
i
1 Rj
0
j
0
k
bulk
I
0
c
i
Tj
0
j
0
26
which is related to Eq. (19) with respect of the angular behavior. Due to the geometry
factor, represented by the direct proportionality to j
0
, there is a monotonous increase in
intensity for angles above the critical angle.
For practical purposes, the divergence of the primary beam should be considered.
The scaling (calibrating) factors k must be determined during the calibration of an in-
strument. Layer-type external standards produced by immersion or spin coating of a wafer
with a spiked solution are recommended (Torcheux et al., 1994; Mori et al., 1995);
however, particulate-type (Fabry et al., 1995) and bulk-type standards have been used
(Gutschke, 1991; Schwenke and Knoth, 1995). All types of standards are commercially
available. Calibration by internal standard is not permissible.
A common preconcentration technique used for the analysis of Si-wafer surfaces is
VPD (vapor-phase decomposition). The wafer is exposed to HF vapor, leading to dis-
solution of the SiO
2
surface layer of the wafer (Neumann and Eichinger, 1991). The
impurities can then be collected by scanning the surface systematically with a droplet of
a special liquid or water from the whole surface. The droplet can then be dried and
measured by TXRF. This leads to a detection capability of 10
8
atoms=cm
2
; the im-
provement is given by the area of the wafer (100 mm, 200 mm, 300 mm ) divided by the
area seen by the detector. The disadvantages of the wafer analysis with VPD pre-
concentration are the loss of the local information and the destructive character (Fabry
et al., 1994). Care must be taken to get a residue within an acceptable small area and to
place it just under the detector.
VI. THINFILMS ANDDEPTHPROFILES
In many technological applications, thin-layered materials and implantations have an
increasing importance. A characterization of such structures can be achieved with angle-
dependent x-ray uorescence analysis in the region around the critical angle of total re-
ection. As the penetration depth of the incoming x-rays changes from the nanometer to
the micrometer region when the angle of incidence passes the critical angle, information
can be obtained on depth distributions and thin lms in that range.
A. Depth Profiles
The uorescence intensity I
i
j
0
of an element i, implanted in a at smooth substrate, with
a depth prole c
i
z showing a variation of its concentration over depth z (normal to the
surface), is a function of the incident angle j
0
:
I
i
j
0
/ Tj
0
j
0
j
T
j
0
Z
1
z0
c
i
z exp
m
1
j
T
j
0
m
2
rz
!
r dz 27
Copyright 2002 Marcel Dekker, Inc.
which is the generalized form of Eq. (26). The mass-attenuation coecients m
1
and m
2
describe the absorption of the exciting and the uorescence radiation of the element i,
respectively; r is the density of the substrate. The transmission coecient Tj
0
takes into
account the refracted part of the incoming radiation and j
T
is the refraction angle inside
the substrate. Examples are given for Ni (Fig. 25) and Al (Fig 26) depth proles.
In practice, the general shape of a depth prole has to be given and some parameters
can be tted (Kregsamer et al., 1999) or dierent functions can be tested (Weisbrod et al.,
1991; Schwenke et al., 1997; Mori et al., 1997), although, theoretically, a Laplace trans-
formation is possible.
B. Thin Films
The composition, density, and thickness of single as well as periodic and nonperiodic
multiple lms can be analyzed (Weisbrod et al., 1991; Schwenke et al., 1992). The uor-
escence intensity I
i
j
0
of an element i (with concentration c
i
) embedded in a single thin
lm with thickness d on top of a at smooth substrate, is a function of the incident angle
j
0
and is described by (de Boer, 1991a, b)
I
i
j
0
/ c
i
jE
T
j
2
1 expm
1
=j
T
m
2
rd
m
1
=j
T
m
2
&
jE
R
j
2
1 expm
1
=j
T
m
2
rd
m
1
=j
T
m
2
2 Real part E
T
E
R
1 exp fi4pj
T
=lr m
2
grd
i4pj
T
=lr m
2
'
28
Figure 25 Calculated NiKa fluorescence intensities as a function of the incidence angle for Ni
depth profiles in quartz excited by MoKa (normalized to values at 10 mrad). As a depth distribution
for the Ni atoms, a step function was assumed. The maxima of the angle distributions with less than
50 nm are in the vicinity of the critical angle of quartz ( 1.8 mrad).
Copyright 2002 Marcel Dekker, Inc.
The electric elds E
T
(transmitted by the lm surface) and E
R
(reected from the substrate)
are derived from Fresnels equations and are complex quantities. The wavelength of the
exciting radiation is l. The formation of standing wave above a reecting surface is al-
ready taken into account by this formalism. An example is given for a Ni thin lm in
Figure 27.
The layer(s) and the substrate must be homogeneous, at, and smooth, with perfectly
sharp interfaces. With some eorts, even a certain roughness of the interfaces can be taken
into account in the calculations (van den Hoogenhof and de Boer, 1993; Nevot and Croce,
1980). A review on grazing incidence X-ray spectrometry and reectometry had been given
by Stoev and Sakurai (1999).
VII. SYNCHROTRONRADIATIONEXCITATION
There are dierent strategies to improve the detection limits, either by increasing the
sensitivity, by reduction of the spectral background, or extending the measuring time
(which, however, is limited for practical reasons).
One approach to reducing the spectral background it to use total-reection geo-
metry. A further possibility for reducing the background is the use of polarized primary
radiation (Ryon and Zahrt, 1993; Aiginger and Wobrauschek, 1974). Due to the ani-
sotropy in the intensity of scattered radiation, it is advantageous to place a detector in
such a position that almost only the isotropic emission of the uorescence signal is
detected. It is possible to combine TXRF with polarized primary radiation excitation.
Moreover, the use of monochromatic primary radiation improves the background
conditions, because only photons with one energy are scattered (incomplete charge
Figure 26 Angular dependence for the intensity of the AlK line obtained for 10
16
atoms=cm
2
Al
ions implanted in silicon at 80
~
E
0
;
~
E;
~
S 0
~
E
0
~
S
0
0
~
E
~
S 0
14
The rst equation in Eq. (14) is also valid for Compton scattering, as g approaches zero
in Eq. (10). Using Eqs. (13) and (14), the mean electric eld vector is dened as
hE
2
i hE
2
x
i hE
2
y
i hE
2
z
i as a function of E
0;x
; E
0;y
; j, and d (see Fig. 1) (Brumme,
1990):
hE
2
i r
2
e
hE
2
0;x
i sin
2
d hE
2
0;y
i1 sin
2
d cos
2
j 15
With Eqs. (11) and (13), this leads to a description of the polarization-dependent scattering
cross section:
ds
dO
d;j
r
2
e
2
1 sin
2
d sin
2
j P
0
sin
2
d cos
2
j cos
2
d 16
Neglecting the atomic form factor and the scattering amplitude, the following is
obtained from Eq. (16) for the scattering cross section sP
0
; d
max
of a photon with the
direction of propagation
~
S
0
and the polarization P
0
in the solid-angle component
O 2p1 cos d
max
(see also Fig. 1):
s
O
P
0
; d
max
r
2
e
2
p 31 cos d
max
1
3
1 cos
3
d
max
P
0
cos d
max
cos
3
d
max
17
When a series approximation of Eq. (17) is conducted around d
max
0 for a scattering
angle Y p=2 to the second order in d
max
, the following is obtained:
s
O
P
0
; d
max
r
2
e
2
p1 P
0
d
2
max
18
drawing on Eq. (18) for the derivation of Eq. (2). For a polarization P
0
1, which ac-
cording to Eq. (11) is equivalent to
~
E
0
E
0;x
~e
x
, and a scattering angle of Y p=2; there is
no scattering in the x direction. With P
0
1, which is equivalent to
~
E
0
E
0;y
~e
y
; scat-
tering in the x direction reaches its maximum. In addition, it follows from Eq. (18) that the
radiation scattered in the solid-angle component O 2p1 cos d
max
is proportional to
the square of the scattering angle d
max
. The scattering angle d
max
is dened by the colli-
mator system.
The polarization P of the scattered radiation measured in the xz plane according to
Figure 1 is
P
hE
2
y
i hE
2
x
i hE
2
z
i
hE
2
y
i hE
2
x
i hE
2
z
i
19
When scatter in the solid angle O is also considered, in accordance with the above
observations the following is obtained:
Copyright 2002 Marcel Dekker, Inc.
P % 1
d
2
max
1 P
0
20
In the ideal case of scatter exactly around p=2 d
max
0, the polarization P of the
scattered radiation is independent of the primary polarization and has a value of unity.
The electric eld vector has exactly one component in the y direction. For greater scat-
tering angles (i.e. greater collimator diameters), the polarization is reduced as described in
Eq. (20). The primary polarization has a smaller inuence than d
max
. The best values of P
are achieved, as expected, with a primary polarization P
0
1
~
E
0
E
0;y
~e
y
.
C. Cartesian Geometry
In Barklas experiment (Compton and Hagenow, 1924) a source of x-rays is incident on a
scatterer S
1
, as shown in Figure 2. The scattered rays from S
1
intercepted by S
2
are nearly
linearly polarized, as the angle S
0
S
1
S
2
is p=2 plane. The electric eld vector is perpendi-
cular to the S
0
S
1
S
2
plane. Radiation from S
1
scattered by S
2
into I
z
also undergoes p=2
scattering, but it is in the same plane as the rst scattering and no further annihilation takes
place. However, scattering of the radiation from S
1
by S
2
into I
y
(also through p=2) anni-
hilates all remaining radiation. Thus, I
z
receives radiation, by I
y
does not. This discussion is
idealized; the measured value I
y
=I
z
depends on the tightness of collimation of the paths S
0
to
S
1
, S
1
to S
2
, S
2
to I
z
, and S
2
to I
y
, and also on the materials and thickness of S
1
and S
2
.
For the experimental determination of the degree of polarization (Brumme et al.,
1990) according to Figure 2, it is necessary to realize that the value P
eff
calculated with
Eq. (21) represents an eective polarization.
Figure 2 Schematic geometry for polarized beam XRF (PXRF). S
0
is the x-ray source, S
1
is the
polarizer, S
2
is the specimen, and I
y
is the intensity at the detector. I
z
is the intensity at an alternate,
in plane position for the detector, which does not eliminate source radiation.
Copyright 2002 Marcel Dekker, Inc.
P
eff
I
z
I
y
I
z
I
y
21
Multiple scattering in the scatterer S
1
and in the specimen S
2
, in particular, but also the
nite apertures of the two collimators C
3
and C
4
, lead to worsening of the degree of
polarization P > P
eff
. If the rst scatter occurs at some angle other than 90
, there will be a
component which can scatter a second time at a right angle relative to the initial incident
beam. The degree of polarization of such multiple scattering events is very low (P580%).
It is, therefore, necessary to consider the polarization unit (S
0
, S
1
, C
1
, and C
2
) as well as
the excitation unit (S
2
, C
4
, and the detector in the y direction) during optimization of
EDPXRF system.
The scattering target, S
1
in Figure 2 can, of course, be replaced by a secondary
target. In this case, the x-ray radiation emitted form the source, S
0
, is also polarized by p=2
through scattering on the target, S
1
. Because the ratio between scattering on the target and
emission of the characteristic radiation of the metal is very small for most metals, po-
larization of the source radiation plays as less decisive role than for typical scattering
targets. When a low-Z Barkla scatterer is used in an orthogonal geometry, marked im-
provements in the detection limits are possible [detection in the xy plane instead of xz
plane in Figure 2 (Bisga rd et al., 1981)]. An intermediate case is when Mg, Al, or Si
secondary targets are used for trace analysis. The polarization of source radiation becomes
more important in comparison to a planar excitation geometry, as the scattering eec-
tiveness of these metals cannot be neglected. In multiple-element analysis, however, sec-
ondary targets are disadvantageous, because only a limited number of elements can be
detected with good sensitivity. In orde
r
to react exibly to analytical applications with
Cartesian geometry, secondary targets are indispensable, in addition to Barkla and Bragg
polarizers. The eectiveness of this technique is shown in Figure 3.
III. BARKLASYSTEMS
Proof-of-principle experiments in the early 1970s (Dzubay et al., 1974; Howell et al., 1975),
using single collimators to dene the three orthogonal beams (Fig. 2), demonstrated im-
proved signal-to-noise ratios, but at severely reduced intensities. A few years later, it was
demonstrated (Ryon, 1977) that the way to maximize analyte intensities while minimizing
background is to open the apertures of the collimators until maximum systems throughput
is achieved while maintaining the orthogonal beam geometry. Such nonzero beam
divergence causes a small decrease in the degree of polarization, which is more than
compensated for by the increase in x-ray ux. Detection limits are improved in comparison
Figure 3 Comparison of excitation methods. The sample is an oil standard which contains 21
elements (e.g., Ca, Ti, V, Cr, Mn, Fe, Ni, Cu, Zn, Mo, Ag, Cd, Sn, Ba, and Pb) with concentrations
of 30 mg=g. The spectra (log scale) are (top) nonpolarized, direct excitation by radiation from
rhodium anode x-ray tube, (middle) molybdenum secondary excitation, and (bottom) polarized
excitation by scatter from a HOPG=Al
2
O
3
target. The measuring time amounts to 200 s for each
spectrum. The counting rate was first adjusted by increasing the tube current to its maximum value
and then increasing the collimator aperture to give a total count rate approximately 60,000 cps using
a Si(Li) detector with an active area of 25 mm
2
and a energy resolution of 166 eV for MnKa. In all
cases, a Rh end-window tube was used.
"
Copyright 2002 Marcel Dekker, Inc.
Copyright 2002 Marcel Dekker, Inc.
to either direct excitation or secondary uorescence (Ter-Saakov and Glebov; 1984,
Bertin, 1975).
Energy-dispersive x-ray uorescence analysis is ideally suited to measuring a wide
range of elements simultaneously because all photon energies are separately binned. When
the analytical goal is to exploit this capability of measuring many elements with good
sensitivity and low detection limits, a polarized polychromatic source is preferred. This is
in contrast to nonpolarized sources, for which the increased sensitivity due to ecient
excitation is oset by the increased background under the peaks.
To optimize polarization and scattering intensity, all inuencing factors, such as the
scattering material and its dimensions and form, multiple scattering in the polarizer and in
the specimen, scope of the collimator, and the nite dimension of the x-ray source and the
detector, must be taken into consideration. The primary intensity of the x-ray tube and the
optimized polarization equipment must be tuned so that, depending on the required re-
solution, the highest possible input count rate is achieved. This means that the loss in
intensity due to the geometry and scattering loss are balanced. Brumme (1990) examined
the above-mentioned inuencing factors using Monte Carlo simulations.
A. Scattering Material
The scattering eectiveness, e, of a planar target, S
1
(see Fig. 2) is dened as
eE
s
S
E
mE
1 exp
rmEd
sin 45
!
22
where mE is the mass-attenuation coecient of the target material for the x-ray energy E
(in cm
2
=g), s
S
is the scattering coecient (coherent incoherent) for the target material
for a scattering angle of p=2 and an x-ray energy E (in cm
2
=g), r is the density of the target
(in g=cm
3
), and d is the thickness of the target (in cm).
The quotient s=m of scatter and absorption must be a maximum for the scattering
target material. Materials with a very small average atomic number are especially suited
for this. They have a relatively small mass-attenuation coecient (m < 1 cm
3
=g) in the
energy ranges of interest, so that the thickness of the scattering target must be increased in
order to obtain a suciently high scattering intensity according to Eq. (22). Thick scat-
tering targets enable great variations in the scattering angle, which logically leads to poor
polarization (Table 1). Additionally, the probability of multiple-scattering processes in-
creases by increasing the thickness of the scattering target and, with it, the depolarization
(see Table 1). Polarization losses have been calculated by Zahrt and Ryon (1981).
Table 1 Scattering Target Effectiveness e and Polarization Calculated with Monte Carlo
Simulations Without P
0
and With P
MS
Consideration of Multiple-Scattering Processes in the
Target as a Function of the Thickness, d
T
, of a Planar Carbon Target (E 19:6 keV
d
T
(mm) e P
0
(%) P
MS
(%)
0.5 0.011 94.1 92.8
1.0 0.021 93.7 91.3
2.0 0.040 93.2 89.8
5.0 0.084 91.3 86.3
10.0 0.131 90.4 84.2
Source: From Brumme, 1990.
Copyright 2002 Marcel Dekker, Inc.
According to Table 1, the thickness of the scattering target should not be increased
unlimitedly to get a high scattering eectiveness. Therefore, the density r of the scattering
material plays a decisive role. For a constant thickness, the materials with a higher density
have the advantage of having a higher scatter eectiveness [see Eq. (22)] for a nearly
equal fraction of multiple scattering and degree of polarization. Materials with density
r < 2 g=cm
3
are not ideal Barkla polarizers. Because the scattering eectiveness is a function
of the energy, there energy ranges, E < 10 keV, 10 < E < 25 keV, and 25 < E < 60 keV, are
considered in the following. In the energy range E < 10 keV, photoabsorption dominates
compared with scattering so that the scattering eectiveness is too small, even for poor
polarizers (P < 0:7). Barkla polarization is therefore too ineective even using x-ray tubes
with anode materials that emit characteristic x-rays in this energy range (Cr, Cu, W, etc.).
In the mid-energy range, it is advantageous to use Barkla polarizers if the characteristic
radiation of the tube anode material lies in this energy range (Mo, Rh, Pd, and Ag). Boron
carbide (Ryon, 1977), pyrolitic graphite (Kanngieer, 1990; Kanngieer et al., 1992), and
boron nitride are particularly suitable as scattering target materials. Beryllium oxide and
boron can also be used. In the upper energy range, the bremsstrahlung emitted from x-ray
tubes withanode materials with Z > 40 canbe used for polarization. Corundum(Svoboda et
al., 1993), boron nitride, beryllium oxide, silicon nitride, and boron carbide are especially
useful as scattering materials. Because the average penetration depth of the scatter radiation
and the fraction of multiple scattering dier in both energy ranges, special optimization
examinations are necessary. Systems using layers of dierent materials are suitable as a
compromise for the entire energy range 10 < E < 60 keV.
In addition to the above-mentioned parameters, other material properties are of
practical interest:
Purity: Because the uorescence radiation excited in the scattering target is
nonpolarized, the smallest impurities in the target material lead to additional
background components in the measured spectrum. Elements with atomic
numbers Z > 20 should not be present with concentrations C>10 mg=g.
Resistance to high temperatures (x-ray tube exit window), air humidity, and x-ray
hardness.
Mechanical stability.
Availability.
B. Geometry of the Polarization Unit
In order to minimize as much as possible the geometric loss in intensity, it is necessary
to keep the distances between the tube and the target, the target and the specimen, and the
specimen and the detector small. It is, however, very important to prevent the sample from
directly irradiating by the primary radiation of the x-ray tube and also the detector from
directly detecting of scattered x-rays from the target or other parts of the unit.
The dimensions of the collimator system C
1
, C
2
, and C
4
(see Fig. 2) exercise the
largest inuence on the properties of the polarization unit. As can be seen in Table 2,
the scattering eectiveness increases by a factor of approximately 50 with variation of the
collimator diameter (d
1
d
2
) of the primary collimator C
1
, d
1
, and the secondary colli-
mator C
2
, d
2
, from 4 to 12 mm with a simultaneous loss in polarization from 0.938 to 0.8.
The ratio between single and multiple scattering events and with it the degradation of P
0
caused by multiple scattering (P
0
P
MS
) is not inuenced signicantly by increasing
collimator diameters (Brumme, 1990).
Copyright 2002 Marcel Dekker, Inc.
With a diameter d
1
d
2
d
3
5 mm for C
1
, C
2
, and C
3
a tube voltage of 50 keV
(Rh anode) and a tube power of 280 W, an input count rate at the detector of approxi-
mately 10
5
cps is obtained for the excitation of 4-g base oil (Conostan) specimen using a
Barkla target. The detected radiation consists mainly of incomplete polarized radiation
scattered at the sample. A further increase in the total scattering eciency is possible
through an increase in the target thickness [see Eq. (22)] or through an increase in colli-
mator diameters. Monte Carlo simulations indicate (Brumme, 1990) that the use of larger
collimator diameters is preferred to the use of thicker targets (see the dierence (P
0
P
MS
)
in Tables 1 and 2). The increase of the ratio between multiple-scattering events and single-
scattering events within the target, and the depolarization (P
0
P
MS
), coupled with it,
due to growing target thickness is the decisive factor.
The Monte Carlo simulations demonstrate that for a given target thickness, the
target area and, with it, the target volume can be extended beyond the volume, in which a
single-scattering event of primary radiation from the x-ray tube into the direction of the
sample can occur (Brumme, 1990). The increase of depolarization caused by the described
extending of the target area is insignicant.
C. X-rayTubes and Filters
All results of Monte Carlo simulations described so far are based on the assumption that
the focal spot of the x-ray tube is innitely small. The consideration of a nite focal spot
leads to a reduction of the polarization (see Table 3).
When mounting a side-window x-ray tube, it should be noted that the brems-
strahlung is slightly polarized in the plane dened by electron beam and x-ray beam from
the focal spot to the Be window of the x-ray tube. For this reason, the electron beam
and the specimen-detector axis should be parallel. When using line-focus x-ray tubes, the
Table 2 Inuence of the Collimator Diameter d
1
(Collimator C
1
) d
2
(Collimator C
2
) on the
Polarization Without P
0
and With P
MS
Consideration of Multiple Scattering and on the Total
Scattering Effectiveness G for a Planar Carbon Target with a Thickness of 2 mm
d (mm) G610
6
P
0
(%) P
MS
(%)
4 0.31 93.8 90.3
5 0.65 92.9 89.6
6 1.38 90.7 87.4
8 4.21 87.8 84.6
12 17.1 79.6 76.7
Note: Point Source S
0
, Distance S
0
S
1
: 36 mm, collimator length C
1
, l
1
2 mm; distance S
1
S
2
: 35 mm, collimator
length C
2
, l
2
12 mm.
Source: From Brumme, 1990.
Table 3 Inuence of the Dimensions of a Quadratic Anode Focal Spot on the Polarization
(Target Thickness: 1 mm)
Focus (mm) Point 161 363 666 868
P
0
0.957 0.957 0.954 0.936 0.918
Source: From Brumme, 1990.
Copyright 2002 Marcel Dekker, Inc.
line-shaped focal spot should lie in the scattering plane. For end-window x-ray tubes, the
diameter of the ring-shaped focal spot should be as small as possible. In practice, a
compromise among tube power, focal spot size, and distance between target and tube
anode should be found.
Except for the Compton shift, the spectral distribution of the tube primary spectrum
is approximately unchanged after Barkla scattering. Thus, the form of the spectral
background that is generated by incompletely polarized tube primary radiation approxi-
mately corresponds to the known background from direct excitation (see also Fig. 3). For
this reason, x-ray lters can be used to optimize the excitation of a limited number of
elements also in EDPXRF similarly to direct excitation EDXRF. Polarization measure-
ments show that lters between the x-ray source and target do not lead to worsening of the
eective polarization. In contrast, lters in the radiation path between the target and the
specimen cause a depolarization of up to 2%. Nonpolarized radiation that results from
contamination in the Barkla target can be absorbed by lters between the target and the
specimen. For example, Cr, Fe, and Ni contamination occurs in B
4
C due to the pressing
procedure; the uorescence radiation of these elements can be absorbed by a 30-mmRh
lter on its way to the specimen.
D. Multiple Scattering in the Specimen
In Sec. II.C, it was mentioned that the eective polarization, P
eff
, must be considered in
the description of the suppression of polarization. The dierences between P
MS
from Table
1 and P
eff
were caused either by x-ray lters between target and specimen or by multiple-
scattering processes in the specimen. The fraction of multiple scattering is dependent on
the matrix. Monte Carlo simulations (Brumme, 1990) show that polarization is reduced by
multiple scattering in a 0.5-mm-thick SiO
2
specimen from P
MS
1:0 to P
eff
0:96. In the
case of fully polarized radiation (P
MS
1:0), the ratio between multiple-scattering events
and single-scattering events within a 3-mm-thick SiO
2
specimen is 40%. This ratio
amounts to only 1520% for a typical polarization of 0:8 P
MS
0:9 and causes a de-
crease of polarization P
MS
P
eff
of 12%. Considering multiple scattering within the
sample and target, the eective polarization P
eff
amounts to 0.800.85 in the energy range
E < 25 keV and to 0.760.83 in the energy range E > 25 keV.
E. Multiple-Layer Scatterers
As later indicated (see Table 4), if a single element is determined, a monochromatic source
gives the lowest detection limits. However, a polychromatic source is desirable when
multielement analysis is performed. If polychromatic radiation is to be polarized by
scattering, high-Z scatterers absorb the low-energy portion of the incident spectrum,
whereas low-Z scatterers do not interact eectively with the high-energy components of
the spectrum (Zahrt and Ryon, 1986). In both cases, this limits the intensity over a broad
range of energies. In both of the examined energy ranges, E < 25 keV and
25 < E < 60 keV, B
4
C is well suited as the polarizer. However, optimized polarization
systems for both energy ranges dier from each other. Compared with B
4
C target, slight
improvements can be achieved using a combination of HOPG (highly oriented pyrolytic
graphite) on Al
2
O
3
. An HOPG target (d 1:2 mm) optimized for the energy range
8 < E < 25 keV is inecient in the upper energy range, due to its low density
(r 2:2 g=cm
3
). For the collimator diameters of 5 mm used here, the Bragg reection is
negligible (Kanngieer et al., 1992). When the target is glued to an Al
2
O
3
substrate,
Copyright 2002 Marcel Dekker, Inc.
the polarization decreases imperceptibly due to multiple scattering in the lower energy
range E < 25 keV. However, in the energy range E < 25 keV, the Al
2
O
3
substrate improves
the scattering eciency at the cost of polarization. On balance, such a multiple-layer
scatterer is useful for determining a wide range of elements simultaneously with very good
detection limits and sensitivities.
F. Applications
Based on Eqs. (3) and (4), Barkla targets are especially suited for multiple-trace-element
analysis in a light matrix. Typical applications include traces in organic substances
[polymers, oils, pharmaceuticals, biological materials (Heckel, 1995), etc.], traces in sili-
cates [rocks (Heckel et al., 1991), soils (Heckel et al., 1992), sludges, cinders, etc.], as well
as traces in Mg and Al alloys.
In a rst application, detection limits and sensitivities for traces in an oil matrix are
compared among direct excitation, secondary target excitation, and polarization excita-
tion (see Fig. 3). As mentioned several times earlier, the count-rate limitation of the de-
tection system is the deciding factor for optimization of the polarization system. It must
also be noted that the electronic noise and, with it, the energy resolution of the spectro-
meter are functions of the pulse peaking time of the pulse-processing electronics and,
through it, of the maximum possible pulse throughput rate. For multiple-element analysis
in the energy range between 1 and 50 keV, the energy resolution for MnKa should be
better than 180 eV. For the spectra displayed in Figure 3, a detection system with an
energy resolution of 166 eV at an input count rate of about 60,000 cps was used. It must be
noted that no x-ray lters are used to keep the goal of a multielement analysis in a short
measuring time. Especially in the case of polarized or direct excitation, the detection limits
of a limited number of elements can be improved by using x-ray lters signicantly. The
detection limits, DLs (see Chapter 2; condence level: 99.86%), for the excitation types
displayed in Figure 3 are summarized in Table 4; the analytical sensitivities, are
summarized in Table 5. A comparison of the numerical values in Table 4 indicates that
Table 4 Detection Limits (n.d. Not Detectable) for Some Elements in a Base Oil Using Direct
Excitation, Monochromatic Excitation Using a Mo Secondary Target, and Direct Excitation
with Linearly Polarized X-rays
Detection limits (in mg=g)
Element Direct excit. Mo second Polarization
Ca 13.0 8.8 4.1
Ti 3.8 2.9 1.6
Cr 3.1 2.0 0.78
Mn 2.6 1.2 0.51
Cu 1.7 0.31 0.34
Zn 1.7 0.30 0.33
Mo 2.3 n.d.
a
0.95
Cd 18.0* n.d. 1.6
Sn 12.0 n.d. 2.0
Pb 3.9 0.31 0.79
Note: 175 W for a Measuring Time of 200 s and an Incident Pulse Density of About 60,000 cps.
a
Overlapping with RhK
b2
.
Copyright 2002 Marcel Dekker, Inc.
the detection limits determined with excitation using polarized x-rays are, on average ve
times better than those determined with direct excitation. An improvement of only a factor
of 2.4 was achieved for the element Mo. In contrast to direct excitation using RhK ra-
diation, the Compton scattered RhKa line is unable to excite Mo. The detection limits for
the elements 46 < Z < 60 are worse than those for the elements 23 < Z < 41, because
these elements are only excited by bremsstrahlung. An improvement is possible when a
voltage >50 kV is used. The Rh tube is operated with a power of 175 W for polarization.
If a 2-kW tube is used, the eective polarization can be improved from P
e
%82% for the
Compton-scattered RhKa line to P
e
%87% for the same detector input count rate
through reduction of the collimator diameter. This clear increase in polarization results in
a 40% improvement of the detection limits displayed for polarization in Table 4.
If one compares the values for polarization equipment with those obtained with Mo
secondary targets, an improvement in the detection limits of a factor up to 2.5 is obtained
for all of the elements with absorption edges that are only slightly below the energy of the
MoKa line (e.g., Pb in Table 4). Similar improvements are expected for the elements Sr and
Zr. Detection limits of the polarization setup become better compared to those obtained
with Mo secondary target if the energy dierence between absorption edge of the elements
and the MoKa line becomes greater. Of course, not all of the elements can be determined
with the Mo secondary target. This reects an additional advantage of Cartesian excita-
tion geometry. A change in the targets is sucient to switch from a multiple-element
application with a polarizer to a single-element application when analyzing single
elements, such as Pb in oil, or a couple of elements with absorption edges that are only
slightly below the energy of the characteristic radiation of the secondary target. Changing
of the targets is computer controlled in commercially available instruments.
A comparison of the numerical values in Table 5 shows that, on the average, a
factor-of-4 improvement in the sensitivity is obtained using the polarization unit as
compared to direct excitation. The sensitivities are also improved by a factor of between 2
and 7 compared to those of the Mo secondary target.
Table 5 Analytical Sensitivity for Some Elements in a Base Oil Using Direct Excitation,
Monochromatic Excitation Using a Mo Secondary Target, and Direct Excitation with Linearly
Polarized X-rays
Sensitivity [in cps=(mg=g)]
Element Direct excit. Mo second Polarization
Ca 0.064 0.037 0.25
Ti 0.29 0.11 0.49
Cr 0.49 0.16 0.98
Mn 0.64 0.25 1.70
Cu 1.71 1.35 4.25
Zn 1.73 1.54 4.28
Mo 1.89 0.0 5.56
Cd 0.28 0.0 2.52
Sn 0.22 0.0 2.15
Pb 0.9 1.57 2.90
Note: 175 W for a Measuring Time of 200 s and an Incident Pulse Density of About 60,000 cps.
Copyright 2002 Marcel Dekker, Inc.
If the detection limits given in Table 4 are not sucient for the elements with atomic
numbers in the range 45 <Z<61, for example, then an improvement can be obtained by
optimizing the polarization unit for the respective energy range. For example, the de-
termination of Cd, Sn, and Sb with a concentration range C>0.1 mg=g in rocks represents
an interesting geological application. From the above discussion, it follows that an ad-
ditional improvement in the peak-to-background ratio can be achieved using an x-ray
lter. The result of the optimization is displayed in Figure 4 for the spectrum of the
standard granite AC-E. As can be seen in Figure 4, the peak-to-background ratios for Ag
and Cd are sucient for the determination of the elements in sub-ppm concentration
range. The root-mean-square deviation for the Cd calibration [see Eq. (28)], which is based
on 22 international rock standards, is 0.2 mg=g. The DLs are summarized in Table 6.
In contrast to the analysis of oil, which has been discussed earlier, when analyzing
geological samples with input count rates of 60,000 cps or more, it is necessary to take
pileup peaks into consideration at energies E<25 keV. This is because the concentration
of the elements K, Ca, Ti, and Fe may amount to % levels. A pileup peak from FeKa
FeKa at 12.8 keV near the PbLb line at 12.6 keV and near the ThLa line at 12.95 keV or a
pileup peak from FeKa CaKa at 10.09 keV near the HgLa line at 9.99 keV are typical
examples. These analytical problems can be contained using lters between the target and
the specimen or between the specimen and the detector, by reducing the input count rate
(see Fig. 5) or by improving the pileup rejection. With the given analytical conditions,
detection limits at sub-ppm levels were obtained for most of the trace elements in geo-
logical specimens (see Tables 6 and 7).
Because the iron content in many geological specimens amounts to the % levels,
an increase in the background in the energy region E<6.4 keV occurs due to incomplete
charge collection in the Si(Li) detector. This background worsens the detection limits,
particularly for V, Cr, and Mn. With a Co secondary target and an Fe lter between the
target and the specimen, only the CoKa line excites the specimen. This line is not able to
excite Fe. The detection limits of the elements V, Cr, and Mn are improved by using the
Co secondary target in an Fe-rich matrix (e.g. Fe ores). The secondary target is again
successfully combined with polarizers.
G. Limitations of Barkla Polarization
As established in Sec. III.A, the scattering eectiveness in the energy region E<10 keV
continually decreases in favor of photoabsorption. Thus, assuming a Cartesian excitation
geometry, it is necessary to examine other excitation possibilities. Using secondary targets,
the number of elements that can be excited with a sucient sensitivity decreases with
falling energy. Therefore, many dierent secondary targets must be used. In practice,
the measuring times would be unacceptably high. Bragg polarizers are an alternative to
secondary targets and are discussed in the following section.
IV. BRAGG SYSTEMS
A. Orthogonal System
Diraction oers an excellent means of obtaining both monochromatic and polarized
radiation. Synthetic multilayer materials that have very high reectivities can have a
d-spacing chosen virtually at will to give a diraction at 2y of p=2 for energies up to about
1.0 keV. For higher energies, crystalline materials must be used. Candidate materials for
Copyright 2002 Marcel Dekker, Inc.
Figure 4 Spectrum (log scale) of the standard specimen AC-E, excited using an Al
2
O
3
polarizer. A 180-mm Ta filter was placed between the
specimen and the target. Some of the certified concentrations are as follows: Ag, 0.1 mg=g; Cd, 0.6 mg=g; In, 0.1 mg=g; Sn, 12.0 mg=g.
Copyright 2002 Marcel Dekker, Inc.
Figure 5 Spectrum (log scale) of the standard specimen AC-E, excited using an B
4
C polarizer. A 50-mm Pd filter was placed between the
specimen and the target. Some of the certified concentrations are as follows: Ni: 1.5 mg=g; Cu: 4.0 mg=g; Sr: 3.0 mg=g; Tl: 0.9 mg=g.
Copyright 2002 Marcel Dekker, Inc.
use in a polarized-beam XRF experiment should have several characteristics. The rst
criterion is a material that has a reection of high diraction power (integral reectivity
>10
4
rad) at 2y near p=2 for the characteristic wavelength of the anode material used. In
addition, the material should be stable, must be free of contamination with characteristic
lines in the energy region 1 <E<15 keV, easy to orient and cut to the desired orientation,
obtainable at reasonable cost, and perhaps nontoxic.
The orthogonal triaxial geometry for Bragg-diraction polarization is similar to that
for Barkla scattering, shown in Figure 2. What is important here is the physics of the
diraction process and the parameters of the diraction crystal. There are two theories of
x-ray diraction; the wave kinematics (K theory) and the dynamic (D theory). The K
theory is valid for small crystals or mosaic blocks. The D theory must be used when
absorption and=or the interaction between incident and diracted beams becomes im-
portant. According to the theory of Zachariasen (1967), the integrated reectivity R
i
of an
ideal mosaic crystal in the symmetrical Bragg case has the following form:
R
i
r
0
V
2
1 cos
2
2y
B
2 sin 2y
B
jF
hkl
j
2
l
3
1
2m
0
exp
2B
sin
2
y
B
l
2
23
where
R
i
Integrated intensity compared to the incident beam
r
0
Classical electron radius
y
B
Bragg diraction angle
Table6 Detection Limits for Various Elements in Rocks Using Excitation with Linearly Polarized
X-rays
Element DL (mg=g) Element DL (mg=g) Element DL (mg=g)
Mo 0.32 Sb 0.25 La 1.4
Pd 0.24 Te 0.38 Ce 1.9
Ag 0.17 I 0.55 Pr 2.4
Cd 0.13 Cs 0.82 Nd 3.3
In 0.16 Ba 1.1 U 0.8
Sn 0.18
Note: Voltage: 59 kV; current: 33 mA; measuring time: 900 s; Al
2
O
3
polarizer; 180-mm Ta lter.
Table7 Detection Limits for Various Elements in Rocks Using Excitation with Linearly Polarized
X-rays
Element DL (mg=g) Element DL (mg=g) Element DL (mg=g)
Co 1.1 Se 0.2 Hf 1.9
Ni 0.9 Br 0.1 Ta 1.5
Cu 0.8 Rb 0.2 W 1.1
Zn 0.6 Sr 0.3 Tl 0.6
Ga 0.3 Y 0.3 Pb 0.5
Ge 0.2 Zr 0.3 Bi 0.5
As 0.2 Nb 0.4 Th 0.6
Note: Voltage: 38 kV; measuring time: 900 s; B
4
C polarizer; 50-mm Pd lter.
Copyright 2002 Marcel Dekker, Inc.
F
hkl
P
N
n1
f
n
exp2pihu
n
kv
n
lw
n
(distance, tube crystal: 26 mm; distance, crystal detector: 28 mm; collimator diameter:
1 mm each; measurement in air; tube voltage: 30 kV). Logarithmic scale!
Copyright 2002 Marcel Dekker, Inc.
polarization unit. Equation (23) is valid assuming that the mosaic spread Z is much greater
than the critical value for secondary extinction:
Z4
2
p
r
R
i
26
A greater mosaic spread is advantageous for the collimator diameter (C
1
, C
2
, and C
4
in Fig. 2) 145 mm, normally used in EDXRF, as a larger angular acceptance is obtained.
On the other hand, the peak reectivity is reduced with increasing mosaic spread. In the
case of an HOPG crystal (R
i
0.0032 rad), the FWHM mosaic spread according to Eq.
(26) is much greater than 0.34
and 2.1
.
The large angular acceptance desired for the HOPG (002) reection leads to an
increase in intensity of the (004) and (006) reections, as bremsstrahlung spectrum of
larger energy range can be reected. An eective excitation (K radiation) of the elements K
to Fe is possible with these reections so that the spectrum from Na to Fe, which can be
simultaneously excited with the three reections, becomes extremely interesting for many
applications. The product K
n, j
according to Eq. (27) is interesting for the analytical eva-
luation of the three reections (n 1, 2, 3):
K
n; j
R
i; n
o
j
r 1
r
j
27
where R
i,n
is the integral reectivity of the nth reection, n {1, 2, 3}, o
j
is the uorescent
yield of the element j, excited with reection n, and r
j
is the jump ratio for the element j.
When observing the K values for typical elements from the three excitation regions,
for example, Si, Ca, and Fe, similar values are found: K
1;Si
% K
2;Ca
% K
3;Fe
; that is, the
element specic sensitivities for the given excitations are similar. The very small uores-
cence yield for the light elements, such as Al and Si, is compensated for by higher integral
reectivity.
B. Curved Crystals
To increase the intensity at the sample, one may bend (Johann) or bend and grind (Jo-
hansson) the crystals (see Chapter 2). With a at crystal in a narrow-beam geometry, the
angular acceptance is essentially the rocking curve, generally !0.01
10
P
M
i
C
certified
i
C
calibrated
i
2
M f
s
28
where M is the number of standards used (M35) and f is the number of regression
coecients determined.
V. BARKLA^BRAGGCOMBINATIONSYSTEMS
The HOPG crystals are characterized as very good Barkla targets in Sec. III and as ex-
cellent Bragg polarizers in Sec. IV. Polarization of the entire spectrum emitted from a tube
has been successfully performed with these crystals. Excitation by monochromatic Bragg
reections dominates the energy range E<10 keV; polychromatic Barkla polarized ra-
diation dominates in the energy range E>10 keV. Optimization of this BarklaBragg
combination system is matrix dependent due to the count-rate limitation of the detection
system and possible pileup lines. Using modern Si(Li) detectors or Si-drift chambers, it is
possible to process 10
4
10
5
cps so that optimization is facilitated by taking advantage of
the ability to control the tube current. Additional targets are only required for special
applications, because it is possible to excite all of the elements between Na and U with this
combination target. Of course, it is possible to optimize the excitation conditions using
radiation lters between the target and sample, analogously to direct excitation with
nonpolarized primary tube radiation. Being able to use air-cooled low-power x-ray tubes
(50 W) as the source of primary radiation is an important advantage of BarklaBragg
combination targets. The spectrum shown in Figure 8 was measured with an integral input
count rate of 25,000 cps (dead time: 18%) using a Peltier-cooled Si-drift chamber with an
active area of 5 mm
2
and an energy resolution of 159 eV for MnKa.
VI. SECONDARY TARGETS
The basic principles when working with secondary targets are to generate nearly mono-
chromatic radiation and to use it for excitation. Because the characteristic radiation of the
target material is nonpolarized, the detection system is burdened by backscattered char-
acteristic radiation from the target material, particularly when analyzing light matrices.
The advantageous application of secondary targets in the Cartesian excitation
geometry required for polarization is described by Bisga rd et al. (1981). The suppression of
Table10 Root-Mean-Square Deviation and Range of the Calibration Lines for Different Elements
in Lithium Tetraborate Glasses Using Excitation with Linearly Polarized Radiation
Element
Range
(g=kg)
RMS
(g=kg) Element
Range
(g=kg)
RMS
(g=kg) Element
Range
(g=kg)
RMS
(g=kg)
Na
2
O 070 0.34 P
2
O
5
010 0.23 TiO
2
01000 0.75
MgO 0970 1.1 SO
3
036 0.28 Cr
2
O
3
0160 0.58
Al
2
O
3
0890 1.0 K
2
O 0130 0.32 MnO 050 0.3
SiO
2
0990 1.1 CaO 0680 0.71 Fe
2
O
3
0560 0.95
Note: Rh tube; voltage: 15 kV; current 11 mA; measuring time: 300 s; at HOPG polarizer; no lter; vacuum.
Copyright 2002 Marcel Dekker, Inc.
primary tube radiation scattered on the target plays an important role, especially when
using secondary targets with low atomic numbers, such as Al. This is because radiation
lters cannot be placed between the target and the sample in this case. The detection limits
listed in Table 4 show that, compared to polarization targets, better results can be ob-
tained with a given x-ray tube for a limited number of elements. For example, a detection
limit of 0.2 mg=g was achieved for Hg in an organic solution using a measuring time of
500 s with a Zr secondary target.
Secondary targets in the Cartesian excitation geometry are often utilized to de-
termine trace elements with atomic numbers Z
trace
<Z
major
without exciting the major
element Z
major
. For example, a Co secondary target is used for the detection of Mn traces
in an iron matrix; an Fe lter is used to absorb the CoKb line that can excite Fe. A Si
secondary target is used for the trace detection of Na, Mg, and Al in a SiO
2
matrix.
A detection limit of 11 mg=g for Al in glass sand prepared as a pressed powder pellet is
achievable.
Thus, secondary targets are an ideal complement to Barkla and Bragg targets in a
Cartesian excitation geometry.
VII. CONCLUSION
In the fundamental publications in the 1980s and 1990s about the application of polarized
radiation in EDXRF by Ryon and Zahrt for Barkla polarization and Aiginger and Wo-
brauschek for Bragg polarization, tube powers in kilowatt ranges are a condition for
presentation of the advantages of EDPXRF over direct excitation with nonpolarized x-ray
radiation. In recent years, the development of Cartesian excitation geometries has been
Figure 8 Spectrum (log scale) of a geological standard sample, GnA, excited with the polarized
primary spectrum of a 50-W low-power Pd x-ray tube, measured with a Peltier-cooled Si-drift
chamber. Some of the concentrations are as follows: Si, 334 mg=g; K, 21.8 mg=g; Ca, 4.4 mg=g; Ti,
120 mg/g; Mn, 1.3 mg=g; Fe, 41.4 mg=g; Cu, 18 mg=g; Zn, 78 mg=kg; Rb, 2020 mg=g; Zr, 70 mg=g; Nb,
94 mg=g; Mo, 100 mg=g; Sn, 1900 mg=g; Bi, 220 mg=g.
Copyright 2002 Marcel Dekker, Inc.
propelled forward so that when coupled with powerful semiconductor spectrometers that
enable the processing of about 100,000 cps with resolution FWHM<200 eV, it is now
advantageous to use 50-W tubes in PXRF.
In practice, PXRF has asserted itself in many industrial laboratories and research
institutes as a powerful method for elemental analysis. The following typical applications
of EDPXRF instruments* have been identied:
Determination of heavy metals in soils and sludges
Major, minor, and trace element determination in geological samples
Precise major and minor element determination in ceramics and refractory materials
Trace elements determination in biological and organic samples
Analysis of aerosols
Analysis of additives in oils
Determination of trace elements in polymers and pharmaceutical products
Rapid determination of heavy metals and halogens in wastes
Analysis of slag
In the future, the adjustment of tube and crystal parameters will be additionally
improved so that new applications with EDPXRF will be possible (e.g., major, minor and
trace element determination within sample areas 1 mm
2
). It will be a long time before the
developments in the analytical elds become completed, especially the use of curved
crystals in EDXRF.
REFERENCES
Aiginger H, Wobrauschek P. J Radioanal Chem 61:281, 1981.
Aiginger H, Wobrauschek P, Brauner C. Nucl Instrum Methods 120:541, 1974; also in measurement,
Detection and Control of Environmental Pollutants. IAEA: Vienna, 1976.
Barkla CG. Trans Roy Soc (London) 204A:467, 1905; Barkla CG. Proc Roy Soc (London) 77:247,
1906.
Beckho B, Kanngieer B, Scheer J, Swoboda W. Adv X-ray Anal 35:1083, 1992.
Bertin EP. Principles and Practice of X-ray Spectrometric Analysis. 2nd ed. Plenum Press: New
York, 1975, p 117.
Bisga rd KM, Laursen J, Schmidt Nielson B. X-ray Spectrom 10:17, 1981.
Brumme M. Die BARKLA-Polarisationseinrichtung in der energiedispersiven Ro ntgenuoreszen-
zanalyse. PhD dissertation, TU Dresden, Dresden, Germany, 1990.
Brumme M, Heckel J, Irmer K. Isotopenpraxis 26:341, 1990.
Champion KP, Whittem RN. Nature 199:1082, 1963.
Compton AH, Hagenow CF. J. Opt Soc Am 8:487, 1924.
Dzubay TG, Jarrett BV, Jaklevic JM. Nucl Instrum Methods 115:297, 1974.
Evans RD. In: Handbuch der Physik. Springer-Verlag: Berlin, 1958, Vol 34.
Friedrich W, Knipping P, von Laue M. Ann Phys. 41:971, 1913.
Hanson AL. Nucl Instrum Methods A243:583, 1986.
Heckel J. J Trace Microprobe Tech 13:97, 1995.
Heckel J, Brumme M, Weinert A, Irmer K. X-ray Spectrom 20:287, 1991.
Heckel J, Haschke M, Brumme M, Schindler R. J Anal Atomic Spectrom 7:281, 1992.
Howell R, Pickles W. Nucl Instrum Methods 120:187, 1974.
Howell R, Pickles W, Cate Jr J. Adv X-ray Anal 18:265, 1975.
*Manufactured by Spectro A.I. GmbH & Co. KG, Kleve, Germany.
Copyright 2002 Marcel Dekker, Inc.
Kanngieer B. Die Anwendung von HOPG-Kristallen in der energiedispersiven Ro ntgenuor-
eszenanalyse. Diploma thesis, University of Bremen, 1990.
Kanngieer B, Beckho B, Scheer J, Swoboda W. X-ray Spectrom 20:331, 1991.
Kanngieer B, Beckho B, Scheer J, Swoboda W. Adv X-ray Anal 35:1001, 1992.
Klein O, Nishina Y. Z Phys 52:853, 1928.
Ryon RW. Adv X-ray Anal 20:575, 1977.
Ryon RW, Zahrt JD, Wobrauschek P, Aiginger H. Adv X-ray Anal 25:63, 1982.
Strittmatter RB. Adv X-ray Anal 25:75, 1982.
Svoboda W, Beckho B, Kanngieer B, Scheer J. X-Ray Spectrom 22:317, 1993.
Ter-Saakov AA, Glebov MV. Atomnaya Energiya 58:260, 1984.
Thomson JJ, Thomson G. The Conduction of Electricity Through Gases. 3rd ed. Cambridge
University Press: Cambridge, 1933.
Wobrauschek P, Aiginger H. Adv X-ray Anal 28:69, 1985.
Wobrauschek P, Aiginger H, Owesny G, Streli C. J Trace Microprobe Tech 6:295, 1988.
Zachariasen WH. Theory of X-ray Diraction in Crystals. Dover: New York, 1967.
Zahrt JD. Adv X-ray Anal 26:331, 1983.
Zahrt JD. Nucl Instrum Methods A242:558, 1986.
Zahrt JD, Ryon RW. Adv X-ray Anal 24:345, 1981.
Zahrt JD, Ryon RW. Adv X-ray Anal 29:435, 1986.
Copyright 2002 Marcel Dekker, Inc.
11
Microbeam XRF
Anders Rindby
Chalmers University of Technology and University of Go
te
borg, Go
tebo
rg, Sweden
Koen H. A. Janssens
University of Antwerp, Antwerp, Belgium
I. INTRODUCTIONANDHISTORICAL PERSPECTIVE
Up to the end of the 1980s, elemental microanalysis based on the emission of characteristic
x-rays was largely restricted to charged-particle-beam techniques (electrons, protons, and
heavier nuclei). Although the use of primary x-rays for inducing the production of the
uorescent radiation was widespread in bulk x-ray uorescence (XRF) instrumentation,
the lack of brilliance of conventional x-ray tubes and the inherent diculties of focusing
x-rays in the 130-keV energy range hampered the development of the microscopic variant
of XRF as a microanalytical tool.
Nevertheless, Glockner and Schreiber (1928) quite early demonstrated the possibility
of doing chemical analysis by uorescence spectroscopy with small x-ray beams. In the
mid-1950s, Long and Cosslett (1957) reported detection limits at the picogram level from
an x-ray microbeam device. Zeits and Baez (1957) predicted a theoretical detection limit
from an optimal instrument to be below 10
14
g. They concluded that compared to the
electron microprobe, the x-ray microprobe was superior in relative sensitivity, but not in
terms of absolute detectability. Localization was also poor for x-rays; however, the energy
impact is much smaller in comparison to charged-particle beams.
In the early 1980s, the availability of highly intense radiation beams from syn-
chrotron sources increased the interest in using microscopic x-ray beams for performing
m-XRF experiments (Chapter 8); also around this period, capillary optics were being used
as simple, compact, and relatively inexpensive means of concentrating a beam of x-rays
down to a smaller size. This innovation lead to the development of microbeam XRF in-
strumentation employing (conventional) x-ray tubes as the source of primary radiation.
During the last two decades, monocapillary and polycapillary focusing devices have been
incorporated into a variety of m-XRF spectrometers; an overview of their characteristics is
provided in Table 1.
More recently, the introduction of commercially available compact minifocus and
microfocus tubes has considerably increased the potential of microbeam XRF, making
possible the use of microbeams in small-sized and easily transportable XRF equipment.
Another signicant development of recent years is the maturation of the techniques
Copyright 2002 Marcel Dekker, Inc.
Table1 Overview of Laboratory-Scale m-XRF Instruments Described in the Literature
Focal size
on anode
Distance
anodeoptics Optics
b
Tube power Brilliance
Anode X-ray source specication
a
(mm
2
) (mm) (diameters given) (W) (kW=mm
2
)
Diffraction tube (a) n.s. n.s. 10, 30, 100 mm coll. 1,000 n.s. Cu
W tube Mo secondary target (b) n.s. n.s. 200 mm str. cap n.s. n.s. W
Philips interchangeable tubes (c) n.s. n.s. 30 mm coll. 2,500 n.s. Cu
Fine-focus diffraction tube (d) n.s. n.s. 10, 30, 100 mm coll. 5001,000 n.s. Cu, Mo
n.s. (e) n.s. n.s. 22, 7.5 mm par. cap. n.s. n.s. Mo
Low-power tube (f) <0.25 0.25 n.s. 10, 30, 100 mm coll. 50 >0.8 Mo
Long, fine-focus diffr. tube (g,h) 0.4 12 $40 200 mm str. cap. 1,9003,000 0.40.6 Cr, W
Radiographic microfocus tube (i) 0.030 $5 10 mm str. cap. 1521 0.50.7 Mo, W
Diffraction tube ( j) 0.4 8 n.s. 18, 23, 27 mm str. cap. n.s. n.s. Cu, W, Mo
Low-power tube (k) 0.25 0.25 n.s. 70 mm coll. 50 0.8 Mo
Long, fine-focus 0.4 8 $40 18 mm con. cap. 1,000 0.3 Cr
diffraction tube (l,m) 7.5 mm ell. cap.
Idem (n) 15, 65, 70 mm con. cap. 1,500 0.5 Mo
Idem (o) 7.5 mm ell. cap. 1,000 0.3 Cr, Cu
n.s. (p) n.s. $40 15 mm con. cap. n.s. n.s. Mo
Radiographic microfocus 0.015 0.1 $5 4, 8, 29, 100 mm str. cap. 1.6 (Cu, W) 1.0 Cu, W, Mo
tube (q) 1.4 (Mo) 0.9
Rotating anode tube (r) 0.5 10 $70 10, 15 mm ell. con. cap. 18,000 3.6 Cu, Mo
Rotating anode tube (s) 0.5 10 $70 160 mm polycap. 18,000 3.6 Mo
Minifocus tube (t) 0.25 0.25 $40 50, 100 mm polycap. 60 0.96 Mo
Radiographic microfocus tube (u) 0.03 0.006 22 21 mm polycap. 12 67 Mo
a
(a) Nichols and Ryan, 1986; (b) Rindby, 1986; (c) Boehme, 1987; (d) Nichols et al., 1987; (e) Yamamoto and Hosokawa, 1988; (f ) Wherry et al., 1988; (g) Engstro m et al.,
1989; (h) Rindby et al., 1989; (i) Carpenter, 1989; Carpenter et al., 1989; Carpenter and Taylor, 1991; (j) Furata et al., 1991; Furata et al., 1993; Hosokawa et al., 1997;
(k) Pella and Feng, 1992; (l) Larsson et al., 1990; (m) Shakir et al., 1990; (n) Rindby, 1993; (o) Attaelmanan et al., 1995; (p) Holynska et al., 1995; (q) Carpenter et al., 1995b;
(r) Janssens et al., 1996a; (s) Vekemans et al., 1998; (t) Bichlmeier et al., 2000; (u) Gao et al., 1997.
b
str.cap., con.cap., par.cap., ell.cap.: resp. straight, conical, paraboloidal, ellipsoidal capillary; coll.: collimator; polcap.: monolithic polycapillary lens; n.s.: not specied.
Copyright 2002 Marcel Dekker, Inc.
for manufacturing monolithic polycapillary x-ray lenses. In addition to being used for
focusing tube-generated x-rays into microscopic spots, at synchrotrons, these devices are
now also being employed for performing m-XAS (x-ray absorption spectroscopy), m-XRD
(x-ray diraction), and x-ray radiography experiments.
The state of the art of m-XRF has recently been summarized in a special issue of the
journal X-ray Spectrometry (Carpenter, 1997) and in a book (Janssens et al., 2000a).
A. Development of Microscopic X-ray Emission Analysis
With the availability of an intense electron microbeam and an ecient detection system for
x-rays, a scanning x-ray microscope was quite an obvious development; one of the very
rst scanning x-ray microscopes was described by Duncumb (1957). The system used
electron excitation and was constructed as a microfocused x-ray tube. A proportional
counter was used for detecting the emitted x-rays. Proportional counters are very ecient
devices but are not able to separate the K x-rays of adjacent elements unless the counter is
combined with some kind of Bragg-reecting dispersion element. With the introduction of
the Si(Li) detectors in the late 1960s, the construction of XRF spectrometers in general
was simplied. One of the rst x-ray excitation scanning uorescent microbeams was
constructed by Long and Cosslett (1957) using a microfocused transmission tube in
combination with an aperture.
Next, by using highly brilliant sources of hard x-rays as provided by synchrotron
storage rings, microbeam spectroscopy became possible at a much more sophisticated
level. The basic instrumental arrangement employed during m-XRF experiments is shown
in Figure 1. The very rst synchrotron-based x-ray microbeam setup was described by
Horowitz and Howell (1972). Tabletop m-XRF devices have developed rapidly over the
last decade and the number of applications is growing. Today, m-XRF instruments are
used in forensic science, industrial quality control, and environmental science. They are
used in material science for analyzing polymers, composite materials, ber materials, soft
tissues from plants, and so forth. Several instruments are now commercially available and
most of these setups have a similar set of components.
A standard microbeam spectrometer consists of an x-ray source, some kind of fo-
cusing device (or just an aperture), a sample holder, and a detection system. The rst two
Figure1 Basic experimental scheme employed during m-XRF investigations.
Copyright 2002 Marcel Dekker, Inc.
components will dene the spectral properties of the primary beam (large or narrow
energy bandwidth). The sample holder should allow for an accurate remote control of the
sample position and the detection system for intensity measurement of uorescent,
transmitted, or=and diracted radiation. A conventional microscope (or just a camera and
a lens system) can be used as a monitoring system for identifying the part of the sample
that is to be analyzed. Normally, uorescence spectroscopy can be done in dierent
modes: point analysis, line scan, and area scan. In addition to uorescence spectroscopy,
absorption and diraction measurements can be done in the same modes. Other types of
spectroscopy can also be performed such as XANES (x-ray absorption near-edge spec-
troscopy) or EXAFS (extended x-ray absorption ne structure spectroscopy) (Bertsch
et al., 1994); however, these types of spectroscopy require the possibility to tune the pri-
mary energy to the absorption edges of element(s) of interest and, in most cases, are
performed only at synchrotron beam lines.
With the development of capillary optics, a considerable number of tabletop
microbeam instruments have been developed and described in the scientic literature
(see Table 1). These instruments normally use capillary optics in combinations with
conventional ne-focus x-ray tubes, where the virtual brilliance can be fully utilized, or
with micro-focused tubes, where the (actual) brilliance is very high. The sensitivity of these
instruments can be quite high: Larsson and Engstro m (1992) reported a detectable limit
(DL) for such an instrument of about 40 fg (4 10
14
g) of calcium in a paper specimen,
which is in agreement with the prediction of Zeitz and Baez (1957).
B. Development of Microfocusing X-ray Optics
The reason why x-ray spectroscopy has not been applied on the microscopic level as has
been the case with electron-probe x-ray microanalysis is mainly attributed to the di-
culties of optically controlling an x-ray beam as compared to an electron beam. Although
dierent types of optics for focusing and imaging in the x-ray region have been developed,
their eciencies are still inferior to those of conventional electron optical systems. How-
ever, even with poor optical eciency, the use of x-ray microbeams oers many ad-
vantages in comparison to electron beam excitation.
Also, for x-ray microscopy, there has been a great need for microfocusing optics. As
early as 1936, Sievert (1936) used a conventional aperture of a fewmicrometers to generate a
small x-ray source for x-ray microscopy. Although the resolution was good, the intensity
was poor and von Ardenne (1939) proposed an electron lens to generate small x-ray sources
for projection x-ray microscopy. A microfocus x-ray tube based on the idea of von Ardenne
was actually constructed by Cosslett and Nixon (1952) for this purpose. Although the
electron lens system constructed in the early 1950s was the starting point for a very rapid
development of electron microscopy, x-ray microscopy and microbeam spectroscopy did
not develop at the same rate due to the inability to generate ecient optical systems for
x-rays. With the introduction of synchrotrons in the late 1960s, a newinterest in x-ray optics
was created and several synchrotron-based x-ray microbeam setups are now in operation
(Iida, 2000). In recent years, also tabletop x-ray microbeam setups have been constructed in
which both conventional and nonconventional optics have been applied (Rindby, 2000).
1. First-Generation Focusing Systems
One of the very rst focusing devices used for x-rays was probably the von SeemanBohlin
(von Seeman, 1919) camera, which was a Rowland circle setup. The practical problem of
Copyright 2002 Marcel Dekker, Inc.
x-ray optics was recognized early when Compton (1923) demonstrated that although
x-rays have measurable optical properties, the refractive power of most materials in
the x-ray region is very low. This means that a conventional lens system for x-rays was very
impractical. However, in 1924, Larsson, et al. (1924) used the edge of a prism block to focus
x-rays. The system, which was rather inecient and had severe chromatic aberration, was
later used by Hink and Petzold (1958), but only to measure refractive indices. The technical
possibilities left for x-ray optics were Bragg reection and total external reection.
Bragg reection was used in the Johan (1931) and Johansson (1933) cameras. An
advantage with Bragg reection is that normal incidence can be used, thus reducing as-
tigmatism and aberrations. However, the system only works within a very small energy
band width and the rest of the radiation energy is completely lost. In order to construct
broad-band x-ray optics, total-reecting devices need to be employed.
Jentzsch (1929) established the essential geometrical principles for x-ray focusing
from total-reecting spherical surfaces. He pointed out the importance of using highly
nished surfaces but also concluded that glancing-angle incidence will always generate
severe astigmatism. Attempts were also made to form images from the inside of a glass
tube. Experiments were carried out by Ehrenberg and Jentzsch (1929) and by Na hring
(1930a, 1930b) and imaging properties were also discussed by Kellerman (1943), but no
result was ever reported. One of the very rst practical focusing devices was a total-
reecting curved crystal developed by Ehrenberg (1947).
In order to overcome some of the problems with astigmatism, another approach to
the imaging requirements was investigated by Wolter (1952a) and Herrnring and Wiedner
(1956). They suggested the use of cylindrical mirrors (i.e., the Wolter microscope). A more
practical method for astigmatism correction is the crossing of two cylindrical mirrors at
right angles. In general, large aberrations in cylindrical and spherical mirrors can always
be reduced by a second mirror. Pioneering work with this kind of compound system was
done by Kirkpatrick and Baez (1948). Other compound systems were proposed by
Herrnring and Wiedner (1956) and Wolter (1952b), suggesting combinations of
hyperboloid and paraboloid surfaces.
2. Second-Generation Focusing Systems
Today, the KirkpatrickBaez compound system as well as the Wolter mirror are in use at
dierent synchrontron-based microbeam setups. Due to the improved technique for de-
veloping highly nished and perfect surfaces, ellipsoidal and paraboloidal mirrors can now
be manufactured with high perfection. The multilayer technique has also improved the
reecting quality of modern mirrors.
The LBL (Lawrence Berkeley Laboratory) microbeam setup uses a Kirkpatrick
Baez system with two mutually perpendicular spherical mirrors coated with a WC
(tungstencarbon) multilayer. The system, when operating at the X-26C beamline at
NSLS (Upton, NY), provides a 50-fold demagnication and generates a 10-mm
2
x-ray
focus. At the NSLS X26A beamline, an elliposoidal mirror is used for focusing purposes.
A Wolter focusing mirror combined with a hyperboloidal and an elipsoidal surface is used
at the Photon Factory microbeam setup (Tsukuba, Japan). A bent crystal is used for the
SRS microprobe (Daresbury, UK). A comparison between dierent microprobe setups
and their sensitivity was given by Larsson and Engstro m (1992). Reviews of synchrotron
microprobes are given by Jones and Gordon (1989) and Rivers and Sutton (1991); Rindby
et al. (2000) provides an up-to-date overview of the technology that is currently used to
focus x-rays. See also Chapter 8.
Copyright 2002 Marcel Dekker, Inc.
3. X-ray Concentrators
Although improvements have been made in the fabrication of optical elements such as
mirrors and zone plates, the technique is limited by its basic imaging properties. Whatever
imaging system is used, the focal spot is nothing else than the image of the source; thus,
in order to produce an x-ray beam with small focal spot, one needs a small x-ray source.
Variation and instability of the primary source can induce a corresponding instability of
the focal spot, and small variation in the geometry of the reecting mirrors, due to thermal
gradients, can cause a large uctuation in the focal length and focal spot size. Many of
these problems can be circumvented by going beyond the conventional imaging optics and
apply nonimaging optics (NIO). If the optical system is only used to focus the beam, NIO
technology is normally superior to conventional imaging optics.
Nonimaging optics was born in the mid-1960s when Baranov, Ploke, and Winston
(see Winston, 1991) independently designed the rst compound parabolic concentrator. In
nonimaging optics, no focal plane can be dened, but the optical elements are designed
only to generate maximum (spatial) concentration of the radiation. These optical elements,
which are normally called concentrators, can generate intensities far better than any
conventional imaging system. A typical NIO technique for x-rays is the application of
small conical glass capillaries used for generating micrometer-sized x-ray beams.
The advantage of the x-ray concentrators (mostly referred to as capillary optics) is
their source independence. The beam size is dened by the capillary inner diameter
rather than the source size. As it is a NIO technique, a high virtual brilliance source (such
as the point focal output from a line-focus x-ray tube) can be used in an ecient way.
Instability and uctuations in the source position will not aect the position or the size of
the focal spot. As the beam is dened by the capillary opening, the beam position can
easily be established by an ordinary optical microscope. Due to its cylindrical symmetry,
thermal gradients will not aect the beam position even at high radiation exposure. Be-
cause it is a total-reection device, capillaries also have broad-band properties. They have
submicrometer capabilities, and compared to conventional x-ray optics, are very in-
expensive. One disadvantage, however, is that the highest intensity is achieved at the ca-
pillary opening so that the sample needs to be very close to the capillary output in order to
utilize the high photon density completely. Although the divergence of a capillary beam is
normally very small (it is of the order of the critical angles for total reection), this pro-
blem can be severe for near-mircometer or submicrometer capillary beams.
After the attempts of Ehrenberg and Jentzsch (1929) to form x-ray images from the
inside of a glass-capillary tube, Jentzsch and Na hring (1931) used straight glass capillaries
of dierent dimensions, of length 1075 cm and an opening of 0.12 mm in diameter, to
generate x-ray microbeams. Practical applications of glass capillaries began at the end of
the 1940s. Chesley (1947) designed a microcamera that used a short, straight glass capil-
lary; in the same year, Kreger (see Hirsch, 1955) used a 20-mm beam from a 1-cm-long
conical glass capillary to study the structure of the wax-rod coating of a sugar cane stem.
Hirsch and Keller (1951) used photographs of the emergent beam from a 35-mm capillary
to properly align the capillary. In the period 19521954, capillary-generated microbeams
were applied to the study of the deformation, recovery, and recrystallization of metals
(Hirsch, 1955). The smallest beam used was 8 mm. Mosher and Stephanakis (1976) de-
scribed the ecient transport of soft x-rays through hollow glass and metal tubing. A
bundle of such tubes was used to remotely image a weak plasma source of soft x-rays.
Comprehensive theoretical analysis of x-ray guides was conducted by Chung and
Pantell (1977) and by Pantell and Chung (1978, 1979), who investigated the transmission
Copyright 2002 Marcel Dekker, Inc.
and power enhancement of x-rays in cylindrical wave guides, the transmission of x-rays
through curved wave guides, and also studied the inuence of surface roughness on the
propagation of x-rays through capillaries.
Rindby (1986) measured the critical angle for total reection for x-rays inside a 200-mm
glass capillary. Inthe same year Nozaki andNakazawa (1986) describeda conical x-ray guide
tube (XGT). The XGT was used with the smaller opening nearest to the x-ray tube to obtain
an almost parallel x-ray beam. Carpenter et al. (1987) carried out several experiments to
determine the variation in beam prole, intensity, and spectrum with capillary size and
position. The rst focusing XGT was reported by Yamamoto and Hosokawa (1988), who
developed a parabolic inner-wall x-ray guide tube made of glass that focused the x-ray beam
into a 5.7-mm spot. The rst scanning x-ray microprobe with glass-capillary collimation,
which had a beam size of 10 mm, was described by Carpenter et al. (1988).
The rst capillary-generated submicron x-ray beam was reported by Engstro m et al.
(1991), from a synchrotron radiation source. Homan et al. (1994) reported a 50-nm x-ray
beam from a conical capillary, which was measured with energies up to 8 keV, using
synchrotron radiation.
Kumakhov (1986, 1990) and Kumakhov and Komarov (1990) performed extensive
research, both theoretical and practical, on the multireaction properties of glass capil-
laries, which led to the design of the Kumakhov or polycapillary x-ray lens, a type of
device now manufactured by several laboratories worldwide (Kardiawarman et al., 1995;
Ullrich et al., 1995; Kumakhov, 1998; Ding et al., 1998; Vekemans et al., 1998; Gao et al.,
1998). This optical technology possesses real imaging properties and has found broad
application areas in both pure and applied sciences. In combination with microfocus x-ray
tubes, beam sizes down to 1530 mm can now be achieved.
C. Chapter Overview
In this chapter, after providing some information on the basic x-raymatter interactions
which are employed for microfocusing x-ray and comparing the way in which x-ray and
charged-particle microbeams interact with matter, the instrumentation required for per-
forming microbeam XRF is discussed, with some emphasis on recent developments in this
eld (tubes, optics, detectors). Next, data acquisition strategies for point, line, and image
XRF analysis are described and the appropriate procedures for evaluating, segmenting,
and quantifying (large amounts of) spectral data are discussed. Finally, an overview of
applications of the microbeam XRF technique in various elds is presented: the latter
include the use of m-XRF in the industrial, environmental, cultural, and forensic sectors of
activity.
II. THEORETICAL BACKGROUND
A. PhotonVersus Charged-Particle-Induced X-ray Emission
When considering the (dis)advantages of microbeam XRF relative to other micro-
analytical methods, in the rst instance, it is relevant to consider the factors determining
the sensitivity of the microbeam XRF technique and its closest analogs, EPXMA (elec-
tron-probe x-ray microanalysis) and m-PIXE (proton-induced x-ray emission). For all
x-ray emission techniques that employ some form of photon counting in the detection
chain (mostly when energy-dispersive detectors are used), the lowest detectable con-
centration level c
DL
(DL: detection limit) for a given chemical element can be estimated
Copyright 2002 Marcel Dekker, Inc.
from the net photopeak intensity N and the corresponding background intensity B, which
are collected from a sample containing a concentration c of a given element, as:
c
DL
c
k
B
p
N
k
s
B
p
s
N
1
It
p 1
where k is usually a number in the range 15 (often k 3), s
N
N=cIt is the net x-ray yield
for this element and s
B
B=It is the corresponding background yield. The sensitivity of
any x-ray emission technique therefore can be increased by the following:
1. Decreasing the background abundance s
B
2. Increasing the elemental yield s
N
3. Increasing the beam intensity I and=or the collection time t
It is useful to consider the relative and absolute magnitudes of the above-mentioned
quantities when comparing the analytical characteristics of various x-ray emission tech-
niques. Because, in the present context, dierent microanlytical techniques are compared,
the achievable beam size S and the beam ux density F I=S (expressed, e.g., in photons,
electrons, or protons per second and per square micrometer) are also important gures of
merit.
a. Penetration Depth
In Table 2, the penetration depth of 20-keV x-ray photons and electrons and 2.5-MeV
protons are compared. The much larger penetrative power of photons in light matrices
(Z<20) relative to charged particles is immediately clear; protons take an intermediate
position between electron and photons. The penetration range for charged particles is
determined by the gradual energy loss that occurs as a result of many inelastic scattering
interactions with the sample material [quantitative described by the stopping power
function SE] and by the curvature of the trajectories (mainly resulting from many elastic
collisions between projectile and sample atoms). On the other hand, for x-ray photons in
the range 150 keV, the most probable interaction mechanism (for many sample materials)
is the photoelectric eect. For example, for 10-keV photons in Al, the probabilities for
photoionization and Compton (i.e., inelastic) and Rayleigh (elastic) scattering are 97.5%,
0.4% and 2.1%, respectively. Because the photoelectric eect causes the primary photon
to be annihilated, the majority of the primary photons will only be subject to a single
Table 2 Penetration Depth of Electrons, Protons, and X-ray Photons in Various Materials
Element
Atomic
number
Atomic
mass
Density
(g=cm
3
)
m
a
L
(cm
71
)
Penetration range (mm)
X
b
20 keV
e
7c
20 kV
p
c
2.5 MeV
C 6 12 1.9 0.79 12,000 5 55
Si 14 28.1 2.3 10 1,000 5 68
Fe 26 55.9 7.8 200 50 1.6 27
Ag 47 107.9 10.5 189 50 1.7 28
Pb 82 207.2 11.3 971 10 2 37
a
m
L
linear mass absorption coefcient.
b
1=e range.
c
Bethe range.
Copyright 2002 Marcel Dekker, Inc.
(scattering or photelectric) interaction; only a very small fraction will rst be (in)elastically
scattered by a sample atom before ionizing a second atom.
b. Phenomena Contributing to the Background
The consequence of this dierent behaviour is that samples irradiated with monochro-
matic x-ray photons of energy E
0
, next to the characteristic radiation of the sample atoms,
will only emit scattered photons in a limited energy window from a few kiloelectron volts
below up to the primary energy E
0
. In contrast, the irradiation of a material with charged
particles of energy E
0
will result in a bremsstrahlung continuum (primarily due to decel-
erating electrons) spanning the entire energy range from 0 to E
0
. This continuum is
especially noticeable in EPXMA spectra, where the continuum is generated at relatively
shallow depths (Reed, 1975) (typically $5 mm; see Table 2 and Fig. 2a). In case of PIXE,
the abundance of the continuum is supressed because it is caused by secondary electrons of
relative low energy generated by the primary protons (Johansson and Campbell, 1988)
(Fig. 2b). When a sample is irradiated with a polychromatic x-ray beam, also in XRF
spectra, a scatter continuum spanning the entire energy range can be observed (Fig. 2c). In
contrast to SR-based microbeam XRF instruments, in which linearly polarized radiation
is employed in order to reduce the scatter background level (Fig. 3a), laboratory m-XRF
instruments operate with unpolarized polychromatic radiation, giving rise to an appre-
ciable background continuum in the EDXRF spectra (Fig. 3b).
c. Production of Characteristic Radiation
In Figure 4, the cross sections for characteristic x-ray production by x-ray photons,
electrons, and protons are compared (Vis, 1990). Two important observations can be
made: (1) for the three projectile types, the overall probability for ionization is compar-
able; (2) for the charged particles, the ionization cross sections decrease with increasing
atomic number of the target atomsthe photoionization process favors heavier over
lighter elements.
Considering Eq. (1), it can therefore be concluded that for equal beam intensity I and
collection time t, the lower background levels that are recorded with monochromatic
photon-induced x-ray emission, together with the comparable or higher ionization cross
sections, will result in lower detection limits (DLs) for m-XRF than for EPXMA and
m-PIXE in the case of heavy elements (Z>25).
Unfortunately, whereas in conventional electron microprobes routinely currents of,
for example, 1 nA (i.e., about 6 10
9
electrons=sec) are focused in an area smaller than
0.001 mm
2
and whereas in m-PIXE apparatus, usually 101000 pA are concentrated into a
110-mm
2
spot (resulting in ux densities in the range of 10
6
10
9
protons=s=mm
2
),
equivalent photon ux densities cannot be straightforwardly achieved using conventional
x-ray sources. The reasons for this are twofold:
1. The spot emitting x-rays on the anode of a conventional x-ray tube is
comparatively large (one to a few tens of square millimeters), whereas the
generated photon beam is emitted into a broad cone.
2. It is not possible to focus x-rays as easily as charged particles or as radiation of
longer wavelengths.
The problem is even aggravated by the fact that x-ray tubes (based on electron-induced
x-ray emission) produce a continuum of x-rays rather than monochromatic radiation.
Typical primary ux densities on a sample surface achievable with a standard x-ray tube
are situated in the range of 10
4
10
5
photons=s=mm
2
(Cu tube, 30 mA, 45 kV, sample at
5 cm from the anode spot) when only the principal anode line intensity is considered.
Copyright 2002 Marcel Dekker, Inc.
Therefore, in order to take advantage of the inherently better peak-to-background
ratio of photon-induced x-ray emission relative to excitation by charged particles, either
much more brilliant x-ray sources than conventional tubes must be used or the radiation
from an x-ray tube must be focused into a microspot on the sample or both.
Figure 2 EPXMA, m-PIXE, and white beam SR-XRF (NSLS X26A station) spectra collected by
irradiating NIST SRM K961 glass microspheres.
Copyright 2002 Marcel Dekker, Inc.
Figure 3 ED-XRF spectra of NIST SRM 1577 Bovine Liver obtained using (a) an 8 8 -mm
2
polychromatic SR beam (220 s counting time) at the NSLS beam line X26A station, (b) a Mo-anode
x-ray tube using a conical capillary (300 s counting time). Dots: experimental spectra, lines: predicted
spectral distributions obtained by Monte Carlo simulation (see text).
Copyright 2002 Marcel Dekker, Inc.
In addition to the use of synchrotron sources (Chapter 8), in the laboratory,
microfocus or minifocus x-ray tubes may be used, either in combination with an aperture
or with focusing optics. Of the latter, in practice, only compact devices such as single-
capillary concentrators or monolithic polycapillary lenses are being used in laboratory
m-XRF instruments.
B. X-ray^Matter Interactions Employed for Microfocusing
1. Optical Theoryof X-rays
The description of the propagation of x-rays in any refractive medium is based on the
same concepts as used for describing ordinary optical light propagation. These types of
models rely on the fact that most materials can be described in terms of a refractive index n
which is related to its dielectric properties
n
e
r
p
2
where e
r
is the relatively dielectric constant. As in general, e
r
is a complex number and the
index of refraction can be written under the form: n 1 d where d a ib. a and b can
be written as
a
2pNe
2
m
e
o
2
f
0
; b
2pNe
2
m
e
o
2
f
00
3
where f
0
and f
00
are the real and imaginary parts of scattering factor, respectively, o is the
frequency of the x-ray, N is the number of atoms per unit volume, m
e
is the electron rest
mass, and e is its charge. The imaginary part of the index of refraction is also related to the
linear absorption coecient m
L
:
b
cm
L
2o
4
Figure 4 Cross sections for K-shell x-ray production by x-rays, protons, and electrons as a
function of target atomic number for various projectile energies.
Copyright 2002 Marcel Dekker, Inc.
For a multielectron atom, f
0
can be written as a sum of scattering factors for all the
electrons. For energies far below any absorption edge, the corresponding electrons will
only contribute in a minor way to the atomic scattering factor. However, for energies far
above the absorption edge, the corresponding electron will contribute with unity. Thus, for
energies far above the K edge, the scattering factor will be equal to the atomic number.
Close to any edge, the scattering factor will have a resonance structure, as shown in
Fig. 5.
2. X-ray Reflectivity
If a plane wave propagating in a medium characterized by the refractive index n
1
1 d
1
is incident on a smooth-plane boundary of another medium with refractive index
n
2
1 d
2
then if d
1
6 d
2
, a reected and a transmitted wave is generated. The reectivity
(R) as well as the transmittance (T) can be calculated from the Fresnel formula that simply
assumes the electric and magnetic eld components to be continuous at the boundary
interface (Azaro et al., 1974). Using Snells law, the parallel and perpendicular compo-
nents of the reectivity R
p
and R
s
can be expressed in terms of the incident angle y
i
and the
indices of refraction of the two media:
r
p
1 d
2
2
sin y
i
1 d
1
1 d
1
2
1 d
1
2
cos
2
y
i
1=2
1 d
2
2
sin y
i
1 d
1
1 d
2
2
1 d
1
2
cos
2
y
i
1=2
; R
p
r
p
r
p
r
s
1 d
1
2
sin y
i
1 d
2
2
1 d
1
2
cos
2
y
i
1=2
1 d
1
2
sin y
i
1 d
2
2
1 d
1
2
cos
2
y
i
1=2
; R
s
r
s
r
s
5
Figure 5 The real part of the atomic scattering factor times the number of atoms per cubic
centimeter for silver and for SiO
2
glass.
Copyright 2002 Marcel Dekker, Inc.
By assuming that the glancing angle is small, the parallel and perpendicular components
will become almost identical and, furthermore, by putting d
1
0 (incidence from air or
vacuum) and d
2
d, we can simplify the formulas for the reectivity to
r
s
r
p
y
i
y
2
i
2d
1=2
y
i
y
2
i
2d
1=2
6
Note that this expression becomes unity if y
2
2d, provided that d is real. This situation
corresponds to total reection of the incident wave front and thus the critical angle
y
c
2d
1=2
/ 1=E, where E is the x-ray energy. Because d, in general, is complex (taking
the absorption into account) the reection coecient will not be exactly unity. However,
for most of the optical materials considered here, the reectivity will be close to unity
within a small angular range from zero degrees up to a few minutes of arc and then
suddenly drop to about $d
2
=4 (see Fig. 6).
Because d is complex d a ib, the following substitution will be convenient:
y
2
2d
p
y
2
2a 2ib
q
p iq 7
Thus, R can be written as
Ry; E
y p
2
q
2
y p
2
q
2
8
where
p
2
1
2
y
2
2a
2
4b
2
q
y
2
2a;
q
2
1
2
y
2
2a
2
4b
2
q
y
2
2a
9
3. Geometrical Aberrations
As the reectivity for x-rays is extremely small except within a small range from zero up to
critical angle y
c
, most x-ray optics are based on the principle of grazing incidence, although
optical devices based on Bragg diraction are also in use (Ice, 1997). The dierent types of
geometrical aberration that occur for such an optical system can be described in terms of
spherical aberration, astigmatism, curvature of the eld, and coma. However, in contrast to
the imaging system of light optics, there is no axial symmetry. Thus, the normal optical
treatment cannot be applied and the classication of aberration will be somewhat dierent.
Spherical aberration arises when rays (originating from a point source), which will be
incident on the mirror surface at dierent angle, are not intercepting at the
same point. Thus, the image of a point object formed by such a nonideal
reector will become a disk on confusion rather then a point.
Astigmatism will arise if the meridian and sagittal rays are not focussed at the same
distance from the mirror.
Coma is a kind of eld aberration that is due to dierent magnications being
produced by dierent regions of the mirror.
Curvature of the eld means the focal plane is slightly curved due to dierences in
focal lengths at dierent position in the object plane. This type of aberration
will determine the physical depth of focus for the system.
Copyright 2002 Marcel Dekker, Inc.
Figure6 Left: The reflectivity for glass as a function of the glancing angle at E5.411 keV; the absorption coefficient was chosen to
be m
glass
(i.e., the normal value), 10m
glass
and 0.1m
glass
. Right: Calculated reflectivity of glass at two energies (CrKa and MoKa) while
using the normal value for absorption coefficient.
Copyright 2002 Marcel Dekker, Inc.
III. INSTRUMENTATIONFOR MICROBEAMXRF
Table 1 summarizes the characteristic features of several tabletop instruments recently de-
scribed in the scientic literature. A m-XRF setup essentially consists of four major parts:
1. An x-ray source
2. A focusing or collimation device
3. A motorized sample stage microscope
4. One or more detectors
As we can see from Table 1, capillary optics is, by far, the most popular type of optics for the
tabletop instruments. Most of the tabletop setups are equipped with either microfocused
tubes in combination with collimators or straight capillaries or high-ux tubes with conical
capillaries. When simple pinhole or cross-slit systems are used to aperture the beam, the
problemis toget close enough to the actual source in order to pick up the highest ux density.
Straight capillaries have beenused by many authors as a means of doing so andguiding it out
to the sample position without the 1=r
2
geometrical loss that occurs when conventional
collimators are employed. As shown in Figure 7, these setups may utilize specially designed
microfocused tubes where the capillary optics have been integrated in the tube construction.
By using conical capillaries, one can operate further away from the source, as the
squeezing of the beam will compensate for the reduced ux density. As a nonimaging
device, the conical capillaries can utilize virtual brilliance that is achieved in standard
diraction tubes by the projection of long ne-focal spots. For any focusing device,
accurate alignment is important; thus, the focusing element has to be completely in-
tegrated and permanently xed to the source or be mounted on a exible stage so that
realignment can be done. A more recent development is the use of polycapillary x-ray
lenses in combination with compact low-power tubes for generating x-ray beams in the
size range above 1530 mm.
Figure 7 Microbeam XRF setup with the capillary focusing unit built into a microfocus x-ray
tube. (From Carpenter et al., 1988.)
Copyright 2002 Marcel Dekker, Inc.
A. X-ray Sources Suitable for MicrobeamXRF
The dierent types of x-ray tube commercially available at present are discussed elsewhere
in this volume (Chapter 5). General information on the x-ray continuum produced in x-ray
tubes can be found in Bethe (1930), Webster et al. (1993), Mott and Massey (1949),
Metchnick and Tomlin (1963), Robinson (1974), and Go rgl et al. (1992). In the framework
of microbeam XRF, basically two types of tube can be distinguished: high-power and low-
power tubes. High-power tubes have a water-cooled anode, are normally equipped with a
side-looking x-ray exit window, and can operate up to several kilowatts. The limitation of
the source brilliance is due to the heat load on the anode that is a function of the power
load, focal spot size, cooling eciency, thermal conductivity, and so forth. The standard
type of diraction tube can take a power load up to 0.6 kW=mm
2
. For diraction tubes,
the focal spot is formed as a ne line allowing for both a line-focal output and a point focal
output, where the virtual high brilliance, from the projection of the line, can be utilized.
The rotating anode technique is a way to improve the cooling eciency of the anode and is
used for very high-power loads.
Low-power tubes can be side-looking, end-window, or transmission tubes. The low
power means that the total ux of radiation is low; however, the brilliance can still be high
when all photons emerge froma small focal spot. For most types of X-ray focusing optics (see
next subsection), the brilliance is the vital parameter and not the total ux. Thus, low-power
tubes are also very interesting in m-beam XRF. For side-looking, microfocused tubes (50
100 mm spot size and smaller), the brilliance is comparable or better than that of the high-
power tubes, although the total power is only a fewwatts. In transmission tubes, the radiation
is taken out in a direction parallel with the electron beam. Here, the anode is just a thin foil
that acts as a radiation window. For this type of tube, the power load is further limited by the
reduced heat conductivity in the anode. In Table 3, the characteristics of some microfocus
and minifocus tubes are listed. A comparison of the brilliance values listed in Tables 1 and 3
shows that all of these newly available minifocus and microfocus sources are well (if not
better) suited for micro-XRF setups than the more conventional sources used up to now.
B. X-ray Optics
As can be seen from Table 1, in laboratory m-XRF setups, monocapillary or polycapillary
optics is employed almost exclusively for beam focusing or concentration. At synchrotron
Table 3 Characteristics of Some Commercially Available Minifocus and Microfocus Tubes
Manufacturer=
model Anode
Nominal
focal size
(mm
2
)
Maximum
voltage
(kV)
Maximum
current
(mA)
Maximum
power
(W)
Nominal
maximum
brilliance
(kW=mm
2
)
Distance
anode to
Be window
(mm)
Kevex=PXS4-613 Mo 250250 60 1 60 0.96 15.8
Oxford=XTF5011 Mo 150150 50 1 50 1.2 17
Oxford=XTF5000HP Mo 100100 50 1 50 5 17
Oxford=XTF5000HP W 5050 50 1 50 20 17
Kevex=PSX5-724 W 1010 70 0.1 7 70 12.2
Hamamatsu=L673101 W 88 80 0.1 8 125 12
Oxford Ultrabright W (1540)
2
90 2 80 128 1.6
Copyright 2002 Marcel Dekker, Inc.
beam lines (having lower natural divergence and higher intensity), in addition to these,
more expensive and elaborate microfocusing X-ray optics are in use also. These devices
were reviewed by Ice (1997). Capillary optics is one of the fastest growing x-ray optical
technologies because of its capacity of generating high-ux-density beams in the
micrometer and submicrometer range. Due to its broad-band characteristics, nonimaging
properties, and simplicity, it is a very attractive optical device for microbeam XRF and
x-ray scanning microscopy.
1. Different Capillary Types
There are basically four major types of X-ray capillary concentrator available today:
straight, conical, and ellipsoidal (or parabolic) monocapillary concentrators and mono-
lithic polycapillary lenses. The principle of x-ray propagation inside these dierent types of
capillary shapes is schematically shown in Fig. 8.
Straight capillaries can be used for transporting the x-ray intensity from close to a
point source to the sample thereby eliminating the 1=r
2
loss from isotropic point sources.
Radiation propagating inside such a device will have a constant angle of incidence and
these types of capillary can easily be manufactured and are easy to apply to almost any
kind of source.
Conical capillaries operate in a similar way as the straight tubes but in addition to
transporting the radiation from the source to the sample, they also squeeze the x-ray
beam down to the dimension of the capillary output diameter. Upon each reection, the
Figure8 The principle of x-ray propagation inside straight, conical, and ellipsoidal capillaries and
in monolithic polycapillary x-ray lenses. In the ellipsoidal and polycapillary cases, a true focus is
formed by rays emitted from the focal point of the device. Any ray emitted outside the optical axis
will propagate inside the ellipsoidal by multiple reflections similar to the conical capillary.
Copyright 2002 Marcel Dekker, Inc.
angle of incidence will increase by an amount 2g, where g is the capillary cone angle. Thus,
photons propagating inside the cone will gradually be absorbed by the walls, because their
angle of incidence will approach the critical angle. Thus, for conical capillaries the max-
imum angular divergence will always be equal to the critical angle (at least for energies
above a certain threshold). In this sense, conical capillaries always tend to maximize the
output intensity. For conical capillaries, the transmission spectrum is more complicated
with a typical bandpass structure, as compared with straight tubes that just have a cuto
energy. By varying the length and cone angle, it is possible to optimize the design of the
capillaries for various energies within a wide range.
One drawback with conical shapes is that the maximum intensity is reached at the
capillary opening, and after that, the beam divergence is of the order of y. For borosilicate
glass, y
c
(mrad) %30=E (keV). Thus, the sample has to be put close to the capillary
opening if the high intensity is to be retained. For medium-sized capillaries (5010 mm),
this is normally not a problem; however, for very small capillaries or irregularly shaped
samples, this can be a problem. If the capillary is too close to the sample surface, it might
block some of the uorescent radiation on its way to the detector.
A more recent development is the ellipsoidal capillary where the divergence can be
reduced due to the imaging properties of ellipsoidals. In a perfect ellipsoidal, any ray
coming from one of the focal points will eventually be focused in the other focal point,
thus small objects placed in one of the focal points will generate an image in the other
point. If an ellipsoid-shaped capillary is placed so that the x-ray source is positioned in one
focal point, part of that source will be focussed in the other focal point. However, for any
source point far away from the optical axis, the radiation will propagate through the el-
lipsoidal capillary much the same as in a conical capillary. If such a capillary has its
opening just in front of its focal plane, the radiation coming out will consists of two
components, one nonimaging divergent beam and one imaging convergent beam.
The sum of the two components will somewhat balance each other up to the focal point.
After that point, both of them will be divergent. This means that there exists a short
distance, for the capillary opening to the focal point of the ellipsoidal, where the diver-
gence of the beam will be substantially reduced.
Monolithic polycapillary x-ray lenses consists of a very large number of straight
channels (from a few thousands up to 250,000) which are rst bundled into monolithic
strands; then, the thick multiber is pulled in such a way that radiation originating from
a focal source point S, which is located typically 25 cm from the lens, is transported by
the channels, changed in direction, and refocused into a focal spot F at the other side of
the lens. In contrast to monocapillaries, the marked advantage of polycapillary devices is
that a large solid angle of the x-ray source is captured and that the radiation is focused
into a point which is several centimeters from the X-ray lens itself, allowing this type of
optics to be used for noncontact=nondestructive types of investigations. Early ap-
proaches (Kumakhov, 1986, 1990; Kumakhov and Komarov, 1990; Kardiawarman et al.,
1995) to the manufacture of polycapillary lenses consisted in the use of individual
straight capillaries which were bent and kept in place by a series of perforated metal
disks; this resulted in fairly large devices [e.g., see Kumakhov (1990) for details].
Monolithic polycapillary lenses are more compact and consists of one or more bundles of
closely packed straight capillaries that, after consolidation into a monolithic bundle, have
been shaped in the desired form. As these lenses are the result of a fairly sophisticated
production process, only a limited number of laboratories in the world are able to
produce them (Ullrich et al., 1995; Kumakhov, 1998; Ding et al., 1998; Vekemans et al.,
1998; Gao et al., 1998).
Copyright 2002 Marcel Dekker, Inc.
2. X-ray Transport Inside Capillaries
X-rays entering into a capillary under a very small glancing angle can propagate inside the
device by successive reections without losing almost any intensity. If the intensity of
x-rays from such a capillary is compared with the intensity when no reections would
occur (which can be determined by replacing the capillary with a collimator with the same
diameter as the nal capillary crosssection), the reectivity of the capillary wall material
can be studied and the intensity gain can be determined.
For a straight capillary, the gain factor can be dened as the ratio between the ef-
fective solid angle DO
eff
seen by the x-ray source (corresponding to the critical angle) and
the solid DO when no reections occur (defned by the exit end), assuming the reectivity to
be unity up to the critical angle (I
g
DO
eff
=DO, as shown in Figure 9. The gain factor
shows the behavior of a low-pass lter. In Figure 10, E
min
corresponds to the smallest
angle of y
c
or the angle that corresponds to the entrance opening, and E
max
corresponds to
the exit end of the capillary. Note that the gain factor is proportional to 1=E
2
in the range
from E
min
to E
max
.
When calculating the transmittance of x-rays through the capillary, one has to know
the grazing angle of incidence and the number of reections. For a straight capillary, the
angle of incidence for each reection and hence the reectivity will be the same. This is not
Figure9 The capillary gain factor (I
g
) of a straight capillary is defined by the ratio of DO
eff
to DO.
Figure 10 The capillary gain factor versus x-ray energy for a straight capillary without any
absorption.
Copyright 2002 Marcel Dekker, Inc.
the case for a conical geometry, where the angle will increase for each reection, and one
has to calculate the reectivity for each reection.
In the model presented in next subsection, a perfectly shaped capillary and an x-ray
point source situated at the capillary symmetry axis are assumed. All y
i
and g are su-
ciently small to allow for the approximations: tan y
i
% y
I
and tan g % g. For a conical ca-
pillary geometry (as dened in Fig. 11), the glancing angle of the ith reection can be
written as (Stern et al., 1988)
y
i
y
0
2i 1g 10
From the geometrical constraints shown in Figure 11, one can easily deduce that the
number of reections N is given by
N
y
0
l l
0
d
1
1
2
11
The total reection coecient R
tot
can be written as the product of each individual re-
ection coecient Ry
i
; E at each successive angle of incidence y
i
encountered by the ray
when passing through the capillary:
R
tot
y
0
; E
Y
N
i1
Ry
i
; E 12
As long as the nal angle of incidence is below the critical angle, the x-rays propagate
inside the capillary by successive reections without losing almost any intensity. If the
intensity of the x-rays from such a capillary is compared with the intensity where no re-
ections occur, which can be determined by replacing the capillary with an aperture
positioned at the site corresponding to the far end of the capillary, the reectivity of the
capillary wall material can be determined. The intensity gain factor, which can also be
dened as the ratio of the intensity with and without reection, can then be written as
I
g
1
DO
Z
y
max
y
min
R
tot
y
0
; E2p sin y
0
dy
0
1 13
Figure11 Conical capillary geometry.
Copyright 2002 Marcel Dekker, Inc.
where DO is the solid angle seen by the x-ray when no reections occur, corresponding to
the far end of the capillary. The limits of integration, y
min
and y
max
, are the minimum and
the maximum glancing angles, respectively, for the x-rays inside the capillary as dened by
the geometric dimensions of the capillary and the x-ray sourcecapillary distance (l
0
).
3. Ray-Tracing Models
Although explicit expressions can be obtained for calculating the reection of radiation
inside a capillary (Engstro m, 1991; Lechtenberg, 1994; Vincze, 1995c), they are only valid
for perfectly shaped conical capillaries and can only be applied for point sources. Any
model that would intend to describe real capillaries applied to real extended sources
have to rely on some kind of Monte Carlo model where all the source and capillary
parameters are used as input data.
There are many dierent factors that have an inuence on the capillary performance
and it is hard to quantitatively assess the signicance of each individual factor by inter-
pretation of simple test experiments. This problem is augmented by the fact that the
manufacture of capillary optics is still very much an art rather than a well-controlled
technique, making it dicult to produce series of capillaries where one parameter is sys-
tematically varied. The need to gain insight into the relative importance of the various
above-mentioned factors has prompted the development of a detailed ray-tracing code
(Vincze, 1995c) that is able to simulate the beam-forming properties of realistic capillary
devices assuming various experimental conditions.
A number of dierent factors such as capillary material, surface roughness, devia-
tions in shape that real devices assume in comparison to the ideal straight, conical,
or ellipsoidal shapes, need to be properly described in such a simulation calculation.
In addition to that, the size, divergence, and distance of the x-ray source and the energy
distribution of the photons that enter the capillary also strongly inuence the performance
of the capillary beam and thus also have to be specied in detail. Such a full-scale ray
tracing simulation model has been described by Vincze et al. (1995c).
Because the capillaries in such a simulation are treated as truly three-dimensional
devices having numerically dened shapes, the only restriction on the modeled shape is the
assumed circular cross section in the plane perpendicular to the capillary axis. This allows
overall distortions of the capillary shape, such as bending of the capillary axis, to be
included in the model. Also, the composition of the capillary material as well as the
roughness of the reecting surface can be freely chosen. The implemented surface
roughness model describes both the attenuation of the specular reectivity and the eects
of diuse scattering by the rough surface. An additional feature of the program is that the
possible transmission of photons through the capillary wall is also taken into account. In
the case of a perfectly conical capillary coupled to a point x-ray source, the simulation is in
perfect agreement with the analytical models, as is demonstrated in Figure 12. However,
the ray-tracing models, is also able to simulate all kinds of capillarysource combination
which cannot be simplied so that they may be described by simple analytical models.
As we can see from Figure 12, the transmission eciency versus energy will appear as
a steplike function, where the steps represents the maximum number of reections that
can occur inside the capillary before the critical angle is reached. This distinct feature
disappears for an extended source, as the initial inclination angle will vary over a wide
range corresponding to the source size. The graphs in Figure 12 also show that the e-
ciency goes down for an extended source and the energy of maximum transmission
eciency will be shifted to lower energies.
Copyright 2002 Marcel Dekker, Inc.
The two-dimensional distribution of the outcoming photons can also be simulated by
the ray-tracing code. For a monochromatic beam, the dierent reection orders will appear
as clearly visible rings in the intensity distribution observed at a distance from the capillary
opening (see Fig. 13, left panel). However, even here, the impact of an extended source will
blur this feature (Fig. 13, center panel). Geometrical distortions or misalignment of the
capillary will be manifested by the appearance of asymmetric rings or crescents in the an-
gular distributions. (The intensity distribution shown in the right panel of Figure 13 was
recorded at a signicant distance from the capillary tip.) Intensity distributions recorded
or calculatedclose to the opening will normally show a rather homogeneous and sharp
distribution corresponding to the zero-order radiation, imaging the actual cross section of
the capillary itself. Thus, for elemental maps, recorded close to the opening, the spatial
resolution can be improved by the application of deconvolution methods. The technique
and possibilities for this enhanced resolution have been discussed by Attaelmanan (1995).
4. Practical Considerations
According to most models of capillary optics, almost any gain factor could be obtained if
there are no geometrical constraints on the capillary dimensions. However, in practice, the
capillaries have to be designed within a narrow range of geometrical limits. In most cases,
the source characteristics and the beam diameter is xed or dictated by the analytical
Figure12 Comparison of the analytical model and the ray-tracing model as applied to a perfect
conical-shaped capillary and a point source. The solid curve shows the transmission efficiency (out
going flux=ingoing flux) calculated for a capillary 5 cm away from a point source with the
dimensions l 7 cm, d
0
31 mm, and d
1
5 mm. The symbols correspond to ray-tracing calculations
for a point and mm extended x-ray source (100 100 mm
2
).
Copyright 2002 Marcel Dekker, Inc.
requirements and, thus, length, shape and capillary material can be optimized. In some
cases, the sourcesample distance is determined by the mechanical arrangement so that
only the shape (i.e., the capillary cone angle) can be varied. Figure 14 shows an example
Figure 13 Intensity distribution from a conical capillary 5 cm away from the opening. left; ray
tracing calculation for a point source, center; ray tracing calculation for an extended source
(400 400 mm
2
) and right; recorded image from a synchrotron set-up. The central spot represents the
direct beam and the rings correspond to different reflection modes (first ring one reflection, second
ring two reflections etc.).
Figure14 Calculated effective solid angle as a function of entrance diameter for different capillary
lengths. The calculation was performed at an energy of 7 keV and a fixed output opening of 5 mm
assuming a point source and a sourcecapillary distance of 5 cm. The surface roughness was
characterized by assuming a slope error of 30 A
14
where wE
0
; E
i
mE
0
csca mE
i
cscb: mE
0
and mE
i
represent the mass
absorption coecients of the sample at the primary and uorescent energy. a and b are the
incidence and takeo angles, respectively (see Fig. 31). In the case of polychromatic
excitation, Eq. (14) needs to be integrated over the appropriate excitation energy range
and a value of d
99%,i
may be obtained numerically.
Copyright 2002 Marcel Dekker, Inc.
In Table 4, the information depth d
99%
is calculated for a few matrices, corre-
sponding to monochromatic MoKa excitation (E
0
17.4 keV); in addition to the energy
of the characteristic radiation, the major composition and density of the sample also
inuences d
99%
is a signicant way. This means that for a sample of given thickness, the
self-absorption can also vary considerably, the same sample being innitely thick for
low-energy radiation while being (partially) transparent for the more energetic lines. To
illustrate this, in Table 5, the absorption factors A
i
1 expw
i
rd =w
i
rd are listed for
a 100-mm-thick foil of the same materials as in Table 4. A
i
is the ratio of the elemental
intensity derived from this foil to that when no self-absorption would take place in it.
3. Self-Absorption Correctionin Heterogeneous Samples
In heterogeneous samples, consisting of several phases of dierent composition, instead of
modeling the attenuation by simple expressions as in Eq. (14), the total attenuation along
the path of the impinging and exiting x-rays must now be calculated as the line integral of
the absorption coecient m
L
. For example, for a sample which features a variable con-
centration c
i
(z) in the z direction (depth) but is homogeneous parallel to the surface,
Figure31 Size of area at beam impaction point required to be homogeneous to permit application
of conventional calibration models to heterogeneous samples.
Table 4 Information Depth (d
99 %
) for Different Elements in a Few Matrices (E
0
17.4 keV,
a b 45
) (in mm)
d
99%
ln100=wE
0
; E
i
r (mm)
Matrix
Density
(g=cm
3
)
CaKa
(3.7 keV)
FeKa
(6.4 keV)
PbLa
(10.54 keV)
SrKa
(14.16 keV)
MoKa
(17.54 keV)
Polypropylene 0.95 830 4,300 16,000 26,000 33,000
Soda-lime glass 2.5 43 150 540 1,000 1,500
Apatite, Ca
5
(PO
4
)
3
OH 3.2 42 60 200 400 550
Malachite, Cu
2
CO
3
(OH)
2
4 24 90 60 100 150
Brass, 70 % Cu, 30 % Zn 8.5 8 25 15 27 37
Lead 11.3 6 5 12 11 12
Copyright 2002 Marcel Dekker, Inc.
the concentration to the x-ray intensity of element i derived from an innitesimal sample
volume with thickness dz situated at a depth z would become (in rst order)
dI
i
dz
/ I
0
E
0
rzc
i
zt
i
E
0
exp
Z
z
0
m
L
E
0
; z
0
csca m
L
E
i
; z
0
cscbdz
0
!
I
0
E
0
rzc
i
zt
i
E
0
exp
Z
z
0
w
L
E
0
; E
i
; z
0
dz
0
!
15
where m
L
E; z mE; z rz is the (depth-dependent) linear absorption coecient. In the
case that this sample consists of a number of discrete layers (index k from 1 to N), each
homogeneous and which its own concentration c
i,k
, mass absorption coecient m
k
E,
density r
k
, and thickness d
k
, the contribution dI
i;k
to the intensity of element i from a slice
of thickness dz in the kth layer becomes
dI
i;k
dz
/ I
0
E
0
r
i
c
i;z
t
i
E
0
exp
X
k
j1
m
j
E
0
csca m
j
E
i
cscbr
j
z
j
" #
I
0
E
0
r
i
c
i;z
t
i
E
0
exp
X
k
j1
w
j
E
0
; E
i
r
j
z
j
" #
16
with
z
j
d
j
if j < k
z
P
k1
n1
d
n
if j k
8
<
:
For other heterogeneous samples with comparable (simple) heterogeneity, analogous ex-
pressions such as that of Eq. (16) may be established. As can be seen from this equation,
even for samples with limited heterogeneity, fairly complex expressions relate the observed
uorescent intensity I
i
to the characteristics of each layer. In the general case (sample
heterogeneous in three dimensions), the relation between I
i
and the sample structure can
only established by numerical integration of the sample concentration and absorption
characteristics over the interaction volume of the primary and uorescent x-rays.
What is important to note when considering Eqs. (15) and (16) (and their equivalents
for other cases) is that the magnitude of the exponential absorption factors depends very
Table 5 Absorption Factor (A
i
) for Different Elements in a Few Matrices (E
0
17.4 keV,
a b45
)
A
i
1 expw
i
rd=w
i
rd; d 100 mm
Matrix
Density
(g=cm
3
)
CaKa
(3.7 keV)
FeKa
(6.4 keV)
PbLa
(10.5 keV)
Sr-Ka
(14.2 keV)
MoKa
(17.4 keV)
Polypropylene 0.95 0.76 0.948 0.985 0.991 0.993
Soda-lime glass 2.5 0.09
a
0.3 0.67 0.8 0.86
Apatite (Ca
5
(PO
4
)
3
OH) 3.2 0.08
a
0.12
a
0.38 0.56 0.65
Malachite Cu
2
CO
3
(OH)
2
4 0.05
a
0.19
a
0.12
a
0.22
a
0.29
Brass (70=30 Cu=Zn) 8.5 0.02
a
0.05
a
0.03
a
0.06
a
0.08
a
Lead 11.3 0.01
a
0.01
a
0.03
a
0.02
a
0.02
a
a
d
99%
d (i.e., sample innitely thick).
Copyright 2002 Marcel Dekker, Inc.
strongly on how the total path of, for example, the uorescent x-rays [total length z csc(b)
in Eq. (16)] is distributed over the various phases. Small variations in this repartition may
cause large uctuations in the integrated absorption coecient along the total path. As an
example, in Table 6, for various thickness combinations the integrated attenuation
exp
P
j
m
j
E
i
r
j
z
j
cscb is listed of several characteristic line energies E
i
in a hypothe-
tical three-layered sample of 100 mm total thickness consisting of parallel layers of KAl-
Si
3
O
8
, NaAlSi
3
O
8
, and SiO
2
bottom layer), assuming that this radiation originates from a
depth of 90 mm.
The data in Table 6 illustrate that in order to reliably correct for strong self-ab-
sorption (CaKa, CrKa, and FeKa), the matrix composition and detailed spatial distribu-
tion of the various phases that make up a heterogeneous sample must be known with high
accuracy. Whereas, in many cases, the matrix of these phases may be approximately
known (e.g., from measurements on more homogeneous parts of the material or by using
other methods), the details on the spatial distribution of the various phases inside natural
samples (e.g., a rock sample composed of dierent minerals) is hard to nd. Table 6 also
shows that when the self-absorption is intermediate to weak (PbLa, SrKa, MoKa, in the
example), only an approximate knowledge of the repartition of phases is required.
In the former case (strong self-absorption), even when the sample geometry is known,
the position of the microbeam relative to this heterogeneous structure must also be known
in order to allow experimental spectra to be corrected for self-absorption. This is illustrated
for a very simple case in Figure 32. This gure shows experimental data collected
by scanning a NIST K961 glass microsphere of 32 mm diameter through a microbeam
of 8 8 mm
2
in the x and y direction( cf. Fig. 31). When the sample movement is done in the
beamsampledetector plane (x direction, see Figs. 32a and 32b), the characteristic ra-
diation of K, Ti, and so forth must pass through the body of the glass sphere on its way to
the detector and the intensity proles are strongly and assymetrically deformed by self-
absorption in the glass. [see Voglis and Rindby (1995) for a more detailed discussion]. Note
that the maxima of the K, Ti, and so forth proles do not coincide with each other or with
the geometrical center of the particle. Experimentally, it is therefore dicult to position the
beam exactly at the center of the particle because this position does not correspond with a
maximum in the detected count rate. The situation is simpler when the sample movement is
executed perpendicularly to the beamdetector plane ( y direction, see Figs. 32c and 32d);
in this case, the uorescent radiation leaves the particle sideways toward, the detector and a
Table 6 Variation of the Attentuation of Characteristic Radiation upon Leaving a Three-Layered
(SiO
2
, NaAlSi
3
O
8
, KAlSi
3
O
8
) Sample with Average Density of 2.5 g=cm
3
Characteristic
radiation
(keV)
25 mm KAlSi
3
O
8
45 mm NaAlSi
3
O
8
30 mm SiO
2
35 mm KAlSi
3
O
8
35 mm NaAlSi
3
O
8
30 mm SiO
2
45 mm KAlSi
3
O
8
25 mm NaAlSi
3
O
8
30 mm SiO
2
CaKa (3.7) 1.3 10
75
0.9 10
75
0.7 10
75
CrKa (5.4) 0.02 0.017 0.015
FeKa (6.4) 0.089 0.08 0.072
PbLa (10.5) 0.57 0.55 0.54
SrKa (14.2) 0.79 0.78 0.77
MoKa (17.4) 0.88 0.87 0.87
Note: The radiation takeoff angle b is 45
.
Copyright 2002 Marcel Dekker, Inc.
symmetrical response is obtained so that all prole maxima occur at the same position (i.e.,
when the beam irradiates the center of the spherical particle).
From the above considerations, it may be concluded that, with the exception of
samples having a simple and well-known spatial distribution and matrix composition of
the phases that constitute them, quantitative m-XRF analysis of truly heterogeneous
Figure 32 Experimental and simulated net intensity profiles resulting from scanning a 32-mm-
diameter K961 glass microsphere through a 8 8-mm
2
white SR beam (a, b) in the beamsample
detector plane (horizontal sample movement). (c, d) perpendicular to this plane (vertical sample
movement), (e) m-XRF spectra collected during scan (a) at 10 mm to the left (lower curve) and right
(upper curve) of the geometrical center of the particle.
Copyright 2002 Marcel Dekker, Inc.
samples is prone to error, especially when low-energy characteristic radiation is involved
for which the information depth is of the same order as the dimensions of the phases
present. Nevertheless, in many practical situations, quantitative measurements can be
performed provided these samples being analyzed are locally homogeneous.
Figure 32 Continued
Copyright 2002 Marcel Dekker, Inc.
4. Conditionsfor Local Homogeneity=Factors Determining Lateral Resolution
A sample may be dened as being locally homogeneous at a given location when both the
primary photons entering the sample at that point and the resulting uorescent x-rays
while leaving the sample on their way to the detector will pass through the same phase. In
this case, the spectral data derived from such a measurement can be quantitatively pro-
cessed as if they were obtained from a completely homogeneous sample. In Figure 31,
the three-dimensional size and shape of the area which must be homogeneous around the
beam-impact point is schematically drawn. Clearly, this area is much wider in the
beamdetector plane (L
x
) than in the direction perpendicular to it (L
y
). L
x
is determined
by the angles of incidence and takeo of the radiation and depends more on the in-
formation depth of the uorescent radiation being considered than on the actual beam size
d
beam,x
. For a parallel primary beam and a small detector far away from the sample, one
can derive that
L
x;i
L
z;i
cot a cot b d
beam;x
csc a
L
y;i
d
beam;y
L
z;i
mind
sample
; d
99%;i
17
In the case of radiation for which the sample is innitely thick, L
z;i
d
99%;i
; for more
penetrant radiation d
99%;i
> d
sample
, the sample thickness determines the lateral size L
x
of
the interaction volume in the x direction and therefore also the lateral resolution in this
direction. In the y direction and perpendicular to the beamdetector plane), only the beam
size d
beam,y
denes the size of the interaction volume, at least when the x-ray detector
Figure 32 Continued
Copyright 2002 Marcel Dekker, Inc.
subtends a small solid angle, as is usually the case at sychrotron m-XRF setups. In labo-
ratory instrumentation, sometimes large-area detectors are used (e.g., 80 mm
2
at a fairly
close distance to the sample (e.g., 2 mm) so that the detector solid angle O becomes quite
large. In this case, the conditions for local homogeneity are somewhat stricter and the
following expressions must used for L
x
and L
y
(with tan
2
g O=p):
L
x;i
L
z;i
cot a cotb g d
beam;x
csc a
L
y;i
L
z;i
2 tan gcsc b d
beam;y
L
z;i
mind
sample
; d
99%;i
18
Degradation of the lateral resolution due to spreading of the primary beam as a result of
scattering interactions (as is the case EPXMA) can be neglected in most cases.
From Eqs. (17) and (18), it follows that in order to maximize lateral resolution in
m-XRF, the highest possible incidence and takeo angles should be employed. Because
such an arrangement also corresponds to the shortest attentuation paths of both primary
and uorescent radiation and yields the smallest volume in which the sample must be
homogeneous, it is also recommended for obtaining the highest quantitative accuracy. The
only argument against the use of large incidence and takeo angles is that when their sum
is no longer 90
, the scatter peak intensity and spectral background will be higher than in
the 90
or 360
has been
covered.
In this manner, instead of performing a conventional XY scan, yielding multiple XY
elemental maps, an XY-scanning movement of the sample through the beam is executed,
which results in several elemental sinograms (see Fig. 36). Through the use of the ap-
propriate reconstruction algorithms, the latter can be back-projected into XZ maps of the
internal distribution of chemical elements in the horizontal plane through the sample at a
particular vertical position Y on the sample.
Figure 34 XRF maps of various elements in a large fly-ash particle obtained using two different
sample-detector orientations: (a) 0
, (b) 90
and a horizontal translation increment of 5 mm while the sample was kept at constant
height. In this way, the distribution of elements inside a plane situated at $35 mm above
the lower edge of the particle can be determined. From the elemental sinograms, by means
Figure 35 Photograph and schematic drawing of the sample environment during XRF
tomography experiments: The sample is placed at the end of a thin glass tube mounted in a
manually adjustable goniometer head. This allows the center of the sample to be aligned with the
rotation axis of the motorized XYZY stage and with the capillary.
Figure 36 Conventional elemental XY maps, elemental sinograms (or XY maps), and elemental
tomographs (or XZ maps) of the fly-ash particle shown in Figure 34. Size in sinograms: 30 120
pixels, pixel size: 5 mm3
, K
, and Ca
2
) were replaced by protons and other
Copyright 2002 Marcel Dekker, Inc.
cations dissolved in the groundwater. Due to a number of other physical and chemical
processes that have occurred inside the leached layer and in the surface crust on top of the
original surface, a chemically and physically complex multilayered structure was formed.
For example, in Figure 24a, a parallel corrosion front next to a hemispherical corrosion
body can be distinguished. In order to obtain a better understanding of the mechanisms
that lead to the formation of these complex structures, the distribution and migratory
patterns of various (trace) elements (e.g., Mn, Bi, Sr, Zr, etc.) must be determined and
correlated to their ionic size and bonding nature to the glass network structure in the
various layers (Newton and Paul, 1980). A comparison of m-XRF- and m-PIXE-generated
elemental maps obtained (Janssens et al., 1996a) showed that both techniques provide
approximately equivalent information on the distribution of trace constituents such as Cr,
Ni, Cu, Zn, Pb, Bi, Sr, and Mo, although, generally speaking, the noise level in the m-XRF
images is higher. Also, the achievable resolution in state-of-the-art m-PIXE setups is sig-
nicantly better than what is now possible in lab-scale m-XRF instruments. In cases,
however, where complementary information on matrix elements can be obtained by, for
example, EPXMA and a lateral resolution of $10 mm is sucient, the maps in Figure 24b
suggest that lab-scale m-XRF can be considered as a valuable (and less expensive) alter-
native microanalytical trace-level technique.
2. Local Analysis of Decorationsin Japanese18th-Century PorcelainVases
An area in which authentication of art(istic) objects is of considerable (economical)
interest in that of Oriental china ware. Since until the 17th century, high-quality por-
celain was exclusively manufactured in China and Japan; imported chinaware were
highly valued and expensive. In the later periods up to now, an extensive production
system of counterfeit Chinese and=or Japanese porcelain both in Europe and in Asia has
come into existence. Some of these reproductions are very sophisticated and cannot
easily be distinguished from the original counterparts by visual inspection alone.
However, objective ways of distinguishing between original period pieces and more
contemporary reproductions have up to now focused predominantly on the porcelain
base material itself (Yap and Tang, 1984, 1985). An alternative approach is to con-
centrate on the variation of the composition of the enamel paints used to decorate the
objects. In view of the size and shape of the pieces, this type of analysis is dicult to
perform using SEMEDX (scanning electron microscopy equipped with energy-dis-
persive x-ray detection) and the intricacy of the decoration patterns usually preclude the
use of conventional XRF except when only averaged analysis results are required (Yu
and Miao, 1997, 1999); m-XRF analysis of individual decorations is feasible however. As
detailed in Janssens et al. (1996a), by employing a monocapillary-base setup, it is pos-
sible to record large-scale m-XRF maps of these decorations and to determine which
transition elements (e.g., Mn, Fe, Cu) were employed to color the PbO-based enamel
paint; it was also possible to distinguish between originally present and add-on dec-
orations of the same color on the basis of the local XRF spectra. Also, polycapillary
lenses, yielding a beam size of approximately the same dimensions as the details of the
decorations, have been used to perform local analysis of this type of material (Vekemans
et al., 1998).
3. Local Analysis of Bronze Statues
In the 19th century, a great interest arose for the ancient Egyptian civilization; many
museums in the world currently own a collection of small Egyptian statues of religious
Copyright 2002 Marcel Dekker, Inc.
nature in stone or bronze (Oost, 1995). To a varying extent, most metal statues have
developed a corrosion layer which prevents reliable analysis of the Cu-based alloys of the
objects were made of (Robbiola et al., 1988; Robbiola and Fiaud, 1993). Nevertheless, this
information is of interest because it can be used to trace the evolution of metallurgy in the
period 30000 BC (Wuttmann, 1986, 1996). Analyses of this type can also be used to
distinguish between original pieces and contemporary forgeries (Vekemans et al., 1998) or
to establish, for example, whether all parts of a statue originate from the same workshop.
By employing microbeam or millibeam XRF, several parts of a statue can be examined in
reproducible conditions (e.g., by avoiding strongly curved surfaces) whereas for quanti-
tative analysis of the underlying metal, only small areas of the corrosion layer (e.g., on the
base of a statue) need to be removed.
As an example, Table 9 lists quantitative analysis results obtained from a statue of
the god Nefer-Hotep represented in Figure 37 (about 40 cm high, from the collection of
the Museum Vleeshuis, Antwerp, Belgium (Oost, 1995) of exceptional quality, dated to
the XXIIXXIII dynasty ($1000 BC) (Vittiglio et al., 1999)]. It was nondestructively
analyzed by means of a simple milli-XRF setup consisting of a collimated low-power
minifocus tube (250 mm focal spot on Mo anode), yielding a pencil beam of $300 mm in
diameter and a 30-mm
2
Si(Li) detector. A fundamental parameter method was used to
obtain quantitative results; analyses of brass and bronze standards revealed that the
elements Cu, Zn, Sn, and Pb (the elements most informative of Cu metallurgy in this
period) could be determined with an accuracy of $10%. Whereas the body of the statue is
copper-colored, has a very smooth surface, and shows only a very thin corrosion layer,
the double crown and base of the statue are darker in color and have a rougher texture,
suggesting the presence of a thicker corrosion crust and possibly the use of different alloy
composition. Both crown and base were made separate from the body. Because the data
in Table 9 suggest a difference in composition between the body of the statue (legs, front
of body) and its crown and base, on the bottom side of the base, a few square millimeters
of the corrosion layer were polished away, exposing the original metal. The results ob-
tained after polishing are also shown in Table 9 and indicate that there is, in fact, no large
difference in composition between base and body and that probably all parts of the statue
were made using a similar alloy composition and in the same workshop. The compositions
found after polishing are consistent with literature data on this type of material and
period. It is clear that the nondestructive analysis possibilities of m-XRF in this case can
contribute signicantly to a better understanding of ancient metallurgical techniques
employed in the period involved.
4. Document Forgeries
The ability to analyze materials in air and in a nondestructive manner makes m-XRF an
excellent technique for the analysis of valuable documents (e.g., to determine their au-
thenticity). Typically, dierent inks, while having the same visible appearance, will have
dierent chemical compositions. Although m-PIXE and m-SRXRF has been used to ex-
amine documents such as the Gutenberg Bible and other early printed works (Cahil, 1981,
1984; Mommsen, 1996), the relative simplicity and nondestructive nature of m-XRF make
it an attractive analytical alternative.
Larsson (1991) and Stocklassa and Nilsson (1993) described analysis by m-XRF of an
500-year-old Swedish possession letter (dated April 1, 1499). The document, a sales
contract for an estate, showed signs of alteration. It is suspected that the alteration was
made in the 1530s when the Swedish king gave land back to certain nobles, who had been
Copyright 2002 Marcel Dekker, Inc.
Table 9 Quantitative Results on Local Analysis of Different Locations on the Nefer-Hotep Statue Shown in Figure 37
Body Base (unpolished) Base (polished)
% (w=w)
Right
leg Front
Head
band
Crown
Front Back Bottom A Bottom A Bottom B Bottom C Lower Upper
Fe 0.12 0.14 0.25 0.09 0.09 0.11 0.11 0.15 0.09 0.07 0.08
Cu 81.2 82 76.5 72.8 75.8 72.6 70.7 71.4 80 84.9 84.8
Ag 0.05 0.04 0.12 0.18 0.13 0.17 0.19 0.13 0.07 0.06 0.06
Sn 13.7 12.9 14.8 19.8 17.5 22.3 21.6 21 15.3 12.2 12.2
Pb 4.8 4.5 7.6 6.5 6.2 4.3 7 6.7 4.6 2.5 2.5
Copyright 2002 Marcel Dekker, Inc.
Figure 37 Photograph of a bronze statue of the Egyptian god Nefer-Hotep (collection of the
Museum Vleeshuis, Antwerp, Belgium), dated to the XXII-XXIII dynasty (ca. 1000 BC).
Copyright 2002 Marcel Dekker, Inc.
stripped of their estates by an earlier ruler. Although alteration was suspected by visual
inspection (the original name of the estate was removed by scraping), the original text was
unreadable. By employing m-XRF-generated Zn maps (a trace constituent of the original
ink), the original name could be established however. In Figure 38, m-XRF scans of Ca
and Zn are shown across a selected area of the parchment. In the Ca map, the (falsied)
visible text can be seen, featuring the family name Ga smestad i Bo re; in the Zn map of
the same area, however, a completely dierent text reading Btinge i Asbo becomes
visible. Obviously, the forger used a dierent ink for the alteration, accounting for the
dierent chemical makeup.
In a forensic (see next subsection) rather than a historically oriented study (Stock-
lassa and Nillson, 1993), a comparison analysis of 10 blue ball-point inks of similar color
by m-XRF analysis revealed the presence in some inks of Fe and Rb, of W, of Fe and Cu,
or of Fe, Cu, and Zn patterns, indicating the possibility to distinguish between the various
inks used by dierent manufacturers. A text (the number 24, see Fig. 26) was written on
a piece of paper with one pen and completely overwritten with another ball-point pen of
Figure 38 The Ca (bottom panel), Mn, Fe, and Zn (top panel) maps of an area of a Swedish
possession letter. In the Ca map, the family name Ga smested i Bo re is visible, corresponding to the
visible (but falsified) writing; in the Zn map of the same area, however, a completely different name
reading Btinge i Asbo is visible.
Copyright 2002 Marcel Dekker, Inc.
similar color but dierent brand. By using the appropriate elemental image specic for the
ink of the rst pen, the original text could be easily visualized.
D. Forensic Applications
In view of the nondestructive nature of m-XRF and the minimal specimen preparation
required, it is an ideal tool for the analysis of material found at crime scenes and for the
investigation forgeries. The use of m-XRF permits to avoid the destructive of valuable
evidence, as is the case for many other methods commonly employed in forensic analytical
chemistry. In Stocklassa et al. (1992), the analysis of various forensic materials such as
single hair bers, paint chips, glass particles, and ball-point pen ink using m-XRF is de-
scribed.
1. Hair Analysis
With respect to the analysis of trace elements in hair, m-XRF has the advantage of being
nondestructive, quick, and able to map element distributions both along the length of the
ber and across its thickness.
Trace elements in hair originate from both internal (endogeneous) and external
(exogeneous) sources (Chatt and Katz, 1988). Exogeneous elements originate from en-
vironmental sources such as tap water or air pollution. The assumption is generally made
that exogeneous elements will be concentrated nearer the surface of hair strands and the
concentrations of those elements will increase with length, as long as the external de-
position continues; in contrast, endogeneous elements should be distributed relatively
uniformly throughout the thickness of the ber, the relative amounts of each element
determined by diet and individual metabolism.
In a few studies, the claim is made that within a persons hair, the compositional
ngerprint among individual hair is relatively constant and that this patterns is unique to a
single individual (Toribara et al., 1982; Seta et al., 1982, Valkovic et al., 1973); thus,
(m-XRF) analysis of a single hair strand could potentially provide the forensic scientist
with unique information to match hair bers found at crime scenes. Engstro m et al. (1989)
showed that detection limits of 1 ppm were possible in the analysis of trace elements in a
single hair ber using a m-XRF analyzer with a 200-mm beam. With the same instrument
and a small (19 mm) Cr-anode beam, Larsson et al. (1990) obtained self-absorption cor-
rected radial distributions of sulfur and calcium from single hair bers without cross
sectioning the ber. The Ca distribution was maximum at the surface of the ber and
showed a minimum in the center of the ber, consistent with other work indicating the
presence of signicant amounts of exogenous Ca in hair from washing in tap water and
with PIXE observations by Cookson and Pilling (1975).
Toribara (1987) constructed an XRF analyzer for the analysis of single hair bers in
12-mm steps along the length of the hair. The authors pointed out that the data obtained
from multiple hair samples from a single individual did not appear to vary signicantly
and that forensic applications were possible. Detection limits were found to be similar to
those for PIXE, or less than about 30 ppm. The longitudinal distribution of Hg in the hair
of an Iraqi woman who had consumed wheat produced by seed that had been treated
with methyl mercury insecticide was determined. From the fact that hair grows at an
average rate of about 1 cm per month, the period of ingestion of the Hg could be de-
termined. This is similar to PIXE data used to show slow poisoning of Napolean by As,
except that it was done faster and with higher spatial resolution (Forshufvud et al., 1961;
Copyright 2002 Marcel Dekker, Inc.
Smith et al., 1962). The same instrument was also used to determine trace elements in
brous materials: chemical dierences could be found in carpets made by dierent
manufacturers, pointing to the possibility to match=identify single textile bers found on
victims and=or criminals.
2. Individual Particle Analysis
The analysis of microscopic particles found on a suspect can potentially place the suspect
at the scene of a crime. Particles may originate from bullets passing through clothing, from
gun shot residue or from glass breakage.
a. Glass Fragment Analysis
The elements present in glass vary depending on the type of glass involved and the
manufacturing lot. A group of European forensic laboratories (Sto cklein et al., 1997)
compared various analytical techniques for discriminating between various types of oat
glass (as used, e.g., in shop windows). Float glass is one of the major types of evidence
encountered in crimes such as burglary, road accidents, and vandalism. A rapid and
widespread manner of distinguishing between the dierent oat glass manufacturers is to
determine the index of refraction of the glass (usually in the range 1.5171.520). However,
due to the globalization of trade and because narrow limits are set for the major com-
ponents used in the manufacturing process of oat glass (i.e., Na
2
O, K
2
O, CaO, and SiO
2
),
the ability of this method to discriminate between glass samples of dierent origin in the
future will be reduced.
Two varieties of m-XRF (employing a unltered, collimated Rh tube and capillary
focused Cu-anode beam, respectively), SEMEDX and ICPMS (inductively couple
plasmamass spectrometry) were used to analyze various standards glasses, some of which
with the same refractive index. Of these, only capillary m-XRF and ICPMS were able to
distinguish between glass samples made in the same furnace but produced in an interval of
9 months; these methods also detected minor inhomogeneities within individual pieces of
glass. Both varieties of m-XRF, with DL values in the range 1100 ppm, were able to
dierentiate between glass manufactured in dierent plants, by making use of the tran-
sition element ngerprints of the glass. Other quoted advantages of the use of m-XRF for
glass fragment ngerprint were its multielemental character, the simple sample prepara-
tion, and its nondestructive character (as compared to ICPMS). Disadvantages are its
limitations with respect to light-element determination and (for irregularly shaped frag-
ments) its semiquantitative character.
b. Gunshot Residue Analysis
When a gun is red, a pattern of residue is distributed on the person ring the weapon and
on the target. Analysis of particles (taken from the clothing or skin of the suspect=victim)
in this case is typically done using scanning electron microscopy and automated routines
may be used to search over large areas, identify particles, and carry out quantitative
analysis on them (Nilsson and Stocklassa, 1993), provided the particles are on the surface
of the material. m-XRF has been used to directly map the residue distribution on fabrics
where the particles have penetrated into the material. In a case where a man was shot
during hunting, the same authors describe the automated search and analysis of metallic
particles deposited into textile fabric by passage of a bullet; the bullet itself was not
available for analysis. All particles were found to contain the same concentrations of Ni,
Cu, and Zn; it could be concluded that the bullet fragments could only have come from
one of two weapons suspected to have red the bullet.
Copyright 2002 Marcel Dekker, Inc.
REFERENCES
Adler LA, Axelrod J, Branco JJR. Adv X-ray Anal 2:167, 1960.
Anderson TA, Evans KL, Carney F, Carney G, Berg H, Gregory R. Spectroscopy 28 (July=August
1991).
Asher RK, Ruiz ER. Plating Surf Finish 73:30, 1986.
Attaelmanan A. On resolution enhancement of capillary based scanning x-ray microuorescence
spectroscopy. PhD dissertation, Chalmers University of Technology, Go teborg, Sweden, 1995.
Attaelmanan A, Rindby A, Voglis P, Shermeat A. Nucl Instrum Methods Phys Res B 82:481, 1993.
Attaelmanan A, Larsson S, Rindby A, Voglis P, Kuczumow A. Rev Sci Instrum 65:7, 1994.
Attaelmanan A, Voglis P, Rindby A, Larsson S, Engstro m P. Rev Sci Instrum 66:24, 1995.
Azaro LV, Kaplow R, Kato N, Weiss RJ, Wilson ACJ, Young RA. X-ray Diraction. New York:
McGraw-Hill, 1974.
Bertsch PM, Hunter DB, Sutton SR, Bajt S, Rivers ML. Environ Sci Technol 28:980, 1994.
Bertucci G. Nucl Instrum Methods Phys Res A322:583, 1992.
Bethe A. Ann Phys 5:325, 1930.
Bichlmeier S, Janssens K, Heckel J, Gibson D, Homann P, Ortner H. X-ray Spectrom. 30:8, 2001.
Boehme DR. Adv X-ray Anal 30:39, 1987.
Bonnet N. Microchimi Acta, 120:195, 1995a.
Bonnet N. Ultramicroscopy 57:17, 1995b.
Bright D, Microbeam Anal 4:151, 1995.
Bright DS, Newbury DE, Marinenko R. In: Newbury DE, ed. Proceedings of the 17th Microbeam
Analysis Society Meeting. San Francisco: San Francisco Press, 1988, p 18.
Burton A. Biological monitoring of environmental contaminants. PhD dissertation, Kings College,
London, 1986.
Bush GT, Stebel MD. Connectors 85:129, 1985.
Cahil T, Kusko B, Schwab RN. Nucl Instrum Methods Phys Res B181:205, 1981.
Cahil T, Kusko B, Eldred RA, Schwab RN. Archaeometry 26:3, 1984.
Carpenter DA. X-ray Spectrom 18:253, 1989.
Carpenter D, Ed. X-ray Spectrom 26:300387, 1997 (entire issue).
Carpenter DA. In: Janssens K, Adams F, Rindby A, eds. Microscopic X-ray Fluorescence Analysis.
Chichester: Wiley, 2000, p 347.
Carpenter DA, Taylor MA. Adv X-ray Anal 34:217, 1991.
Carpenter DA, Gorin A, Shor JT. Adv X-Ray Anal 38:557, 1995a.
Carpenter DA, Taylor MA, Lawson RL. J Trace Microprobe Technol 13:141, 1995b.
Carpenter DA, Taylor MA, Holcombe CE. Adv X-ray Anal 32:115, 1989.
Carpenter D, Haney L, Lawson R, Poirier P, Paper presented at the 36th Annual Denver X-ray
Conference, Denver, 1987.
Carpenter D, Lawson R, Taylor M, Poirier P, Morgan R, Haney L. In: Newbury DE, ed.
Proceedings of the 17th Microbeam Analysis Society Meeting. San Francisco: San Francisco
Press, 1988, p 391.
Chatt A, Katz SA. Hair Analysis: Applications in the Biomedical and Environmental Sciences.
New York: VCH, 1988.
Chesley FG. Rev Sci Instrum 18:442, 1947.
Chung PS, Pantell RH. Electron Lett 13:527, 1977.
Compton AH. Philos Mag Series 6 45:270, 1923.
Cookson JA, Pilling FD. Phys Med Biol 20:1015, 1975.
Cosslett VE, Nixon WC. Proc Roy Soc B140:422, 1952.
Criss JW, Birks L. Anal Chem 40:1080, 1968.
Cross BJ, Wherry DC. Thin Solid Films 166:263, 1988.
Cross B, Lamb RD, Ma S, Paque JM. Adv X-ray Anal 35:1255, 1992.
Czechanshi M, Measurement of precious metal coating thickness using x-ray uorescence
instrumentation. Proceedings of the 9th International Precious Metals Conference, 1986, p 457.
Copyright 2002 Marcel Dekker, Inc.
Ding X, He J, Wei F, Xie J, Wong P, Li Y, Chen B, Chen J, Yan Y. Adv X-ray Anal 41:243, 1998.
Duncumb P. Ph.D dissertation, Cambridge University, Cambridge, 1957.
Ehrenberg W. Nature 160:330, 1947.
Ehrenberg W, Jentzsch F. Z Phys 54:227, 1929.
Engstro m P. Development of Capillary Optics for x-ray focussing. PhD dissertation, Chalmers
University of Technology, Go teborg, Sweden, 1991.
Engstro m P, Larsson S, Rindby A, Stocklassa B. Nucl Instrum Methods Phys Res B36:222, 1989.
Engstro m P, Larsson S, Rindby A, Buttkewitz A, Garbe S, Gaul G, Kno chel A, Lechtenberg F.
Nucl Instrum Methods Phys Res A302:547, 1991.
Forshufvud S, Smith H, Wassen A. Nature 192:103, 1961.
Fukumoto N, Kobayashi Y, Kurahashi M, Kojima I. Spectrochim Acta 54B:91, 1999.
Fukumoto N, Kobayashi Y, Uchiumi A, Kuruhashi M, Kawase A, Watanabe H. In: Shinohara K,
ed. X-ray Microscopy in Biology and Medicine. Tokyo: Japan Science Society Press, 1990,
p 227.
Furata K, Nakayama Y, Hosokawa Y. Rev Sci Instrum 54:135, 1993.
Furata K, Nakayama Y, Hoji M, Nakano H, Hosokawa Y. Rev Sci Instrum 62:828, 1991.
Gao N, Ponomarev I, Xiao QF, Gibson WM, Carpenter DA. Appl Phys Lett 71:3441, 1998.
Geladi P, Kowalski BR. Anal Chim Acta 185:1, 1986.
Geladi P. Microchim Acta 120:211, 1995.
Gilfrich JV, Gilfrich NL, Skelton EF, Kirkland JP, Quadri SB, Nagel DJ. X-Ray Spectrom 20:203,
1991.
Glockner S, Schreiber M. Ann Phys Lpz 85:1089, 1928.
Gordon B, Jones K. In: Morris MD, ed. Microscopic and Spectroscopic Imaging of the Chemical
State. New York: Marcel Dekker, 1993, p 303.
Go rgl R, Wobrauschek P, Kregsamer P, Streli C. X-ray Spectrom 21:37 1992.
Haschke M, Scholz W, Theis U. In: Fernandez JE, Tartari A, eds. Proceedings European Conference
on Energy Dispersive X-ray Spectrometry. Bologna: Editrice Compositori, 1998, p 157.
Hayakawa S, Gohshi Y, Iida A, Aoki S, Ishikawa M. Nucl Instrum Methods Phys Res B49:555,
1990.
He F, Van Espen P. Anal Chem 63:2237, 1991.
Herman GT. Image Reconstructions from Projections. London: Academic Press, 1980.
Herrnring G, Wiedner W. Quoted in: Hildebrand G. Fortschr Phys 4:11, 1956.
Hink W, Petzold W. Naturwissenschaften 45:107, 1958.
Hirsch PB. In: Peiser HS, Rooksby HP, Wilson AJC, eds. X-ray Diraction of Polycrystalline
Materials. London: 1955.
Hirsch PB, Keller JN. Proc Phys Soc 64:377B, 1951.
Homan S, Bilderback D, Thiel D. Nucl Instrum Methods Phys Res A347:384, 1994.
Holynska B, Ostachowicz B, Ostachowicz J, Ostrowski A, Ptasinski J, Wegrzynek D. J Trace
Microprobe Technol 13:163, 1995.
Horowitz P, Howell JA. Science 178:608, 1972.
Hosokawa Y, Ozawa S, Nakazawa H, Nakayama Y. X-ray Spectrom 26:380, 1997.
Ice G. X-ray Spectrom 26:315, 1997.
Iida A. In: Janssens K, Adams F, Rindby A, eds. Microscopic X-ray Fluorescence Analysis.
Chichester: Wiley, 2000, p 117.
Jaksic M, Grime, Henderson J, Watt F. Nucl Instrum Methods Res B54:491, 1991.
Janssens K, Van Langevelde F, Adams FC, Vis RD, Sutton, Rivers ML, Jones KW, Bowen DK.
Adv X-ray Anal 35:1265, 1992.
Janssens K, Adams F, Rindby A, eds. Microscopic X-ray Fluorescence Analysis. Chichester: Wiley,
2000a.
Janssens K, Adams FC, Rivers ML, Jones KW. Scan Microsc 7(Suppl):191, 1993a.
Janssens K, Vincze L, Van Espen P, Adams F. X-ray Spectrom 22:234, 1993b.
Janssens K, Vekemans B, Adams F, Van Espen P, Mutsaers P. Nucl Instrum Methods Phys Res
B109=110:179, 1996b.
Copyright 2002 Marcel Dekker, Inc.
Janssens K, Vekemans B, Vincze L, Adams F, Rindby A. Spectrochim Acta B51:1661, 1996a.
Janssens K, Vincze L, Vekemans B, Adams F, Haller M, Kno chel A. J Anal Atomic Spectrom
13:319, 1998.
Janssens K, Vincze L, Vekemans B, Aerts A, Adams F, Jones KW, Kno chel A. Microchim Acta
13:87, 1996c.
Janssens K, Vincze L, Wei F, Proost K, Vekemans B, Vittiglio G, He Y, Yan Y, Falkenberg G.
Feasibility of polycapillary-based trace-level m-XANES by means of bending magnet
radiation. Book of abstracts of European Conference on Energy Dispersive X-ray Spectro-
metry, Krakow, Poland, 2000c, p 18.
Janssens K, Aerts A, Vincze L, Adams F, Yang C, Utui R, Malmqvist K, Jones KW, Radtke M,
Garbe S, Lechtenberg F, Kno chel A, Wouters H. Nucl Instrum Methods Phys Res
B109=110:690, 1996d.
Janssens K, Vittiglio G, Deraedt I, Aerts A, Vekemans B, Vincze L, Wei F, Deryck I, Schalm O,
Adams F, Rindby A, Kno chel A, Simionovici A, Snigirev A. X-ray Spectrom 29:73, 2000b.
Jeanguillaume C. J Microsc Spectrosc Electron 10:409, 1985.
Jentzsch F. Z Physik 30:268, 1929.
Jentzsch F, Na hring N, Z Technol Phys 12:185, 1931.
Johan HH. Z Phys 69:185, 1931.
Johansson T. Z Phys 82:507, 1933.
Johansson S, Campbell J. PIXE, A Novel Technique for Elemental Analysis. Chichester: Wiley,
1988.
Johnson RA, Wickern DW. Applied Multivariate Statistical Analysis. 2nd ed. Englewood Clis, NJ:
Prentice-Hall International, 1988.
Jones KW, Gordon BM. Anal Chem 61:341A, 1989.
Jones PD, Bria KR, Schweingruber FH. Geophys Res Lett 22:1333, 1995.
Kardiawarman K, York BR, Qian XW, Xiao QF, MacDonald CA, Gibson WM. In: Hoover R,
Williams MB, eds. X-ray and Ultraviolet Sensors and Applications, Proc SPIE 2519. Society
of Photo-Optical Instrumentation and Engineering, 1995, p 197.
Kellerman K. Ann Phys Lpz 43:32, 1943.
Kempenaers L, Vincze L, Janssens K. Spectrochim Acta B 55:651, 2000.
Kirkpatrick R, Baez J. J Opt Soc Am 38:766, 1948.
Kuczumow A, Rindby A, Larsson S, X-ray Spectrom 24:19, 1995.
Kuczumow A, Vekemans B, Schalm O, Dorrine W, Chevallier P, Dillmann P, Ro C-U, Janssens K,
Van Grieken R. J. Anal Atomic Spectrom 14:435, 1999a.
Kuczumow A, Vekemans B, Schalm O, Vincze L, Dorrine W, Gysels K, Van Grieken R. X-ray
Spectrom 28:282, 1999b.
Kumakhov MA. Energoatomizdat Moscow, 1986 (in Russian).
Kumakhov MA. Nucl Instrum Methods Phys Res B 48:283, 1990.
Kumakhov MA. Adv X-ray Anal 41:214, 1998.
Kumakhov MA, Komarov FF. Phys Rev 191:289, 1990.
Lahanier Ch, Amsel G, Heitz Ch, Menu M, Anderson HH. Nucl Instrum Methods Phys Res B14:1,
1986.
Lankosz M. X-ray Spectrom 22:125, 1993.
Lankosz M, Pella PA. Adv X-ray Anal 37:685, 1994a.
Lankosz M, Pella PA. X-ray Spectrom 23:169, 1994b.
Lankosz M, Pella PA. X-Ray Spectrom 24:327, 1995.
Lankosz M, Pella PA. X-ray Spectrom 26:347, 1997.
Lankosz M, Holynska B, Pella PA. X-Ray Spectrom 22:54, 1993.
Larsson S. X-ray microbeam spectroscopy. PhD dissertation, Chalmers University of Technology,
Go teborg, Sweden, 1991.
Larsson S, Engstro m P, Rindby A, Stocklassa B. Adv X-ray Anal 33:623, 1990.
Larsson S, Engstro m P. Adv X-ray Anal 35:1019, 1992.
Larsson A, Siegbahn M, Waller I. Naturwissensch 52:1212, 1924.
Copyright 2002 Marcel Dekker, Inc.
Lechter P, Eckbauer S, Hartmann R, Richter R, Stru der L, Krisch S, Soltau H, Hau D, Fiorini C,
Gatti E, Longoni A, Sampietro M. Nucl Instrum Methods Phys Res A377:346, 1996.
Lechtenberg F. Ho chstortsauo sende ro ntgenuoreszenzanalyse mit weisser synchrotronstrahlung.
PhD dissertation, Fachbereich Physik der Westfa hlischen Universita t Munster, Munster,
1994.
Long JVP, Cosslett VE. In: Cosslett VE, Engstro m A, Pattee HH, eds. X-ray Microscopy and
Microradiography, New York: Academic Press, 1957.
Markert B. Plants as Biomonitors. Weinheim: VCH, 1993.
Metchnik V, Tomlin SG. Proc Phys Soc 81:956, 1963.
Miller DC. Adv X-ray Anal 4:513, 1961.
Mommsen H, Beier Th, Dittmann H, Heimermann D, Hein A, Rosenberg A, Boghardt M,
Manebutt-Benz EM, Halbey H. Archaeometry 38:347, 1996.
Mosher D, Stephanakis SJ. Appl Phys Lett 29:105, 1976.
Mott NF, Massey HSW. The Theory of Atomic Collisions. Oxford: Oxford University Press, 1949.
Nagj M, Injuk J, Valkovic V. Nucl Instrum Methods Phys Res B22:467, 1987.
Na hring E. Z Physik 31:401, 1930a.
Na hring E. Z Physik 31:799, 1930b.
Nakazawa H. J Appl Cryst 15:239, 1983.
Nellessen J. Erweiterung der micro-ro ntgenuoreszenzanalyse auf dreidimensionale proben.
Diplomarbeit, Fachbereich Chemie, Universita t Dortmund, 1996.
Newton RG, Paul A. Glass Technol (Berlin) 21:307, 1980.
Nichols MC, Ryan RW. Adv X-ray Anal 29:423, 1986.
Nichols MC, Boehme D, Ryan RW, Wherry D, Cross B, Aden G. Adv X-ray Anal 30:45, 1987.
Nilsson G, Stocklassa B. Bullet identication from microscopic fragments embedded in cloth. Paper
presented at the IAFS (International Association of Forensic Societies) Conference,
Du sseldorf, 1993.
Nozaki H, Nakazawa Y. J Appl Cryst 19:453, 1986.
Oost A, ed. Egypte Onomwonden, Catalogue of Egyptian Artefacts of the Museum Vleeshuis
Antwerp, 1995.
Osan J, To ro k Sz, Rindby A. In: Janssens K, Adams F, Rindby A, eds. Microscopic X-ray
Fluorescence Analysis. Chichester: Wiley, 2000, p 315.
Osan J, To ro k B, To ro k Sz, Jones KW. X-ray Spectrom 26:37, 1997.
Pantell RH, Chung PS. IEEE J Quant Electron 14:694, 1978.
Pantell RH, Chung PS. Appl Opt 18:897, 1979.
Paque JM, Browning R, King PL, Pianetta P. Proceedings 19th Microbeam Analysis Society
Meeting. San Francisco: San Francisco Press, 1990, p 202.
Pella PA, Feng L. Adv X-ray anal 35:1063, 1992.
Press WH, Teukolsky SA, Vetterline VT. Numerical Recipes in C: The Art of Scientic Computing.
2nd ed. Cambridge: Cambridge University Press, 1992.
Prutton M, El Gomati M, Kenny PG. J Electron Spectrosc Related Phenom 52:197, 1990.
Reed SJB. Electron Microprobe Analysis. Cambridge: Cambridge University Press, 1975.
Rindby A. Nucl Instrum Methods Phys Res A249:536, 1986.
Rindby A. X-ray Spectrom 22:187, 1993.
Rindby A. In: Janssens K, Adams F, Rindby A, eds. Microscopic X-ray Fluorescence Analysis.
Chichester: Wiley, 2000, p 95.
Rindby A, Voglis P, Attaelmanan A. X-ray Spectrom 25:39, 1996.
Rindby A, Engstro m P, Janssens K, J Synchrotr Radiat 4:228, 1997a.
Rindby A, Engstro m P, Janssens K. Nucl Instrum Methods Phys Res B124:228, 1997b.
Rindby A, Engstro m P, Larsson S, Stocklassa B. X-ray Spectrom 28:109, 1989.
Rindby A, Voglis P, Nilsson G, Stocklassa B. Adv X-ray Anal 35:1247, 1992.
Rivers ML, Sutton SR. Synchrotr Radiat News 4:23, 1991.
Robbiola L, Fiaud C. Ed Rev Me tall 6:156, 1993.
Robbiola L, Queixalos I, Hurtel LP, Pernot M, Volfovsky C. Studies Conserv 33:205, 1988.
Copyright 2002 Marcel Dekker, Inc.
Robinson JW. Handbook of Spectroscopy. Boca Raton, FL: CRC Press, 1974, Vol 1.
Russ JC. The Image Processing Handbook. 2nd ed. Boca Raton, FL: CRC Press, 1995.
Ryan CG, Clayton E, Grin WL, Sie SH, Cousens DR. Nucl Instrum Methods Phys Res B34:396,
1988.
Sakellariou A, Cholewa M, Saint A, Legge GFJ. Meas Sci Technol 8:746, 1997a.
Sakellariou A, Cholewa M, Saint A, Legge GFJ. Nucl Instrum Methods Phys Res B131:253, 1997b.
Seaman A. Finishing 9:18, 1985.
Selin E, Standzenieks P, Boman J, Teeyasoontranont V. X-Ray Spectrom 22:281, 1993.
Seta S, Sato H, Yoshino M, Miyasaka S. Scan Electron Microsc 1:127, 1982.
Shakir N, Larsson S, Engstro m P, Rindby A. Nucl Instrum Methods Phys Res B52:194, 1990.
Shiraiwa T, Fujino N. Jpn J Appl Phys 5:886, 1966.
Shiraiwa T, Fujino N, Adv X-ray Anal 11:63, 1968a.
Shiraiwa T, Fujino N, Adv X-ray Anal 11:95, 1968b.
Sievert RM, Acta Radiol Stockh 17:299, 1936.
Simionovici A, Chukalina M, Schroer Ch, Drakopoulos M, Snigireva I, Snigirev A, Schroer Ch,
Lengeler B, Janssens K, Adams F. IEEE Trans Nucl Sci 47:2736, 2000.
Smith H, Forshufvud D, Wassen A, Naute 194:725, 1962.
Somogyi A, Vincze L, Janssens K, Vekemans B, Rindby A, Adams F. Spectrochim Acta B 55:75,
2000a.
Somogyi A, Vincze L, Janssens K, Vekemans B, Rindby A, Adams F. Spectrochim Acta B 55:1037,
2000b.
Spiller E, Segmuller A. Appl Phys Lett 27:2105, 1974.
Stebel MD, Silverman WM. XRF programmable plating thickness measurement instrumentation
for automated testing of small parts and printed wiring boards. Proceedings of the ICWA 1984
Coil Winding Conference, 1984, p 75.
Stern EA, Kalman Z, Lewis A, Lieberman A. J Opt Soc Am 27:5135, 1988.
Stocklassa B, Nilsson G. Document examinations using the Intrax. Proceedings of the IAFS
(International Association of Forensic Societies), Du sseldorf, 1993.
Stocklassa B, Nilsson G, Paulsson N. Paper presented at the European Conference on EDXRS,
Myconos, 1992.
Sto cklein W, Gunaratnam L, Hicks T, Kidfelt G, Rindby A, Warman G, Widstedt I. The analysis of
oat glass using elemental analysis: Comparison of techniques. Proceedings of the EAFS
(European Association of Forensic Societies) 1997 Conference, Lyon, 1997.
Stru der L, Fiorini C, Gatti E, Hartmann R, Holl P, Krause N, Lechner P, Longoni A, Lutz G,
Kemmer J, Meidinger N, Popp M, Soltau H, Weber U, von Zanther C. J. Synchrotr Radiat
5:268, 1998.
Svietlicki E, Larsson NP, Yang C. Nucl Instrum Methods Phys Res B77:195, 1993.
Tissot RG Jr, Boehme DR. The novel use of XRMF spectroscopy for environmental toxicity
evaluation and recyclable potential in electronic components. Proceedings Denver X-ray
Conference, 1994.
Toribara TY. Adv X-Ray Anal 30:293, 1987.
Toribara TY, Jackson DA, French WR, Thompson AC, Jaklevic JM, Anal Chem 54:1844, 1982.
To ro k Sz, Faigel Gy, Jones KW, Rivers ML, Sutton SR, Bajt S. X-ray Spectrom 23:3, 1994.
Trebbia P, Wulveryk JM, Bonnet M. Microbeam Anal 4:85, 1995.
Ullrich JB, Huang KG, Owens SM, Aloisi DC, Homann FA, Gao N, Klotzko IL, Gibson WM.
Nucl Instrum Methods Phys Res A 364:362, 1995.
Valkovic V, Miljanic D, Wheeler RM, Liebert RB, Zabel T, Phillips GC. Nature 243:543, 1973.
Van Borm W, Wouters L, Van Grieken R, Adams F. Sci Total Environ 90:55, 1990.
Van Dyck PM, To ro k SzB, Van Grieken RE. Anal Chem 58:1761, 1986.
Van Espen P, Janssens K, Nobels J. Chemom Intell Lab Syst 1:109, 1986.
Van Espen P, Janssens G, Vanhoolst W, Geladi P. Analusis 20:81, 1992.
Vekemans B, Janssens K, Vincze L, Adams F, Van Espen P. X-ray Spectrom 23:278, 1994.
Vekemans B, Janssens K, Aerts A, Adams F, Hertogen J. X-ray Spectrom 26:333, 1997.
Copyright 2002 Marcel Dekker, Inc.
Vekemans B, Janssens K, Vincze L, Adams F, Van Espen P. Spectrochim Acta 50B:149, 1995.
Vekemans B, Janssens K, Vittiglio G, Adams F, Andong L, Yiming Y. Adv X-ray Anal 41:278,
1998.
Vekemans B, Vincze L, Szaloki I, Feng H, Riekel C, Van Grieken R, Adams F, Jones KW. Micro X-
ray uorescence tomography on single sediment particles from New York/New Yersey
Harbor. Book of Abstract of European Conference on Energy Dispersive X-ray Spectrometry,
Krakow, 2000, p106.
Vincze L. Monte Carlo Simulations of conventional and synchrotron x-ray uorescence spectro-
meters. PhD dissertation, University of Antwerp (UIA), Antwerp, 1995.
Vincze L, Janssens K, Adams F. Spectrochim. Acta 48B:553, 1993.
Vincze L, Janssens K, Adams F, Jones K. Spectrochim Acta B50:1481, 1995b.
Vincze L, Janssens K, Adams F, Rindby A. X-ray Spectrom 24:27, 1995c.
Vincze L, Janssens K, Vekemans B, Adams F. Spectrom Acta B54:1711, 1999a.
Vincze L, Janssens K, Vekemans B, Adams F. J Anal Atomic Spectrom 14:529, 1999b.
Vincze L, Janssens K, Vekemans B, Adams F. Monte Carlo simulaiton of x-ray uorescence and
scattering tomography experiments. Proc SPIE (Soc Photo-optical Inst Eng) Conference on
Developments in X-ray Tomography II, Denver, CO, 1999c, p 328.
Vincze L, Janssens K, Adams F, Rivers M, Jones K. Spectrochim. Acta B50:127, 1995a.
Vis R. In: Catlow C, Greaves R, eds. Applications of Synchrotron Radiation. London: Blackie,
1990, p 311.
Vittiglio G, Janssens K, Adams F, Oost A. Spectrochim Acta B54:1697, 1999.
Voglis P, Rindby A. J Trace Microprobe Technol 13:177, 1995.
Voglis P, Attaelmanan, Engstrom P, Larrson S, Rindby A, Bostro m K, Helander CG. X-ray
Spectrom 22:229, 1993.
von Ardenne M. Naturwissenschaften 27:485, 1939.
von Seeman H. Ann Phys 59:455, 1919.
Webster DL, Hansen WW, Duveneck FB. Phys Rev 44:258, 1933.
Wegrzynek D. A feasibility study: Computer micro tomography using a laboratory x-ray
uorescence micro-beam spectrometer. Book of Abstract of European Conference on Energy
Dispersive X-ray Spectrometry, Krakow, 2000, p 25.
Wegrzynek D, Holynska B, Ostachowicz J. X-ray Spectrom 28:209, 1999.
Welfringer J, Bennoit P, Guyon M, Adv X-ray Anal 35:1301, 1992.
Wherry D, Cross B. Kevex Analyst 12:8, 1986.
Wherry DC, Cross B, Briggs TH. Adv X-ray Anal 31:93, 1988.
Winston R. Sci Am 52:March, 1991.
Wolter H. Ann Phys Lpz 10:94, 1952a.
Wolter H. Ann Phys Lpz 10:286, 1952b.
Worley CG, Havrilla GJ. Anal Chem 70:2957, 1998.
Worley C, Havrilla G, Gao N, Xiao QF. Adv X-ray Anal 42:26, 1999.
Wuttmann M. In: Valloggia M, Balat I, Le Mastaba de Medoe-Nefer, Annexe III: Analyse et e tude
du me tal cuivreux de certains objets. FIFAO 31=1, Le Caire, 1986, p 215.
Wuttmann M. The beginnings of copper and copper alloys metallurgy in Egypt. The need for a
better characterisation of the artefacts. Proceedings First Int. Conference on Ancient Egyptian
Mining and Metallurgy and Conservation of Metallic Artefacts, Cairo, 1996.
Yamamoto N, Hosokawa Y. Jpn J Appl Phys 27L:2203, 1988.
Yap CT, Tang SM, Archaeometry 26:78, 1984.
Yap CT, Tang SM. Archaeometry 26:61, 1985.
Yu KN, Miao JM. Appl Radiat Isot 48:959, 1997.
Yu KN, Miao JM. Appl Radiat Isot 51:279, 1999.
Zahrt JD. Adv X-ray Anal 32:89, 1989.
Zietz L, Baez AV. In: Cosslett VE, Engstro m A, Pattee HH, eds. X-ray Microscopy and
Microradiography, New York: Academic Press, 1957.
Copyright 2002 Marcel Dekker, Inc.
12
Particle-Induced X-ray Emission Analysis
Willy Maenhaut
Ghent University, Ghent, Belgium
Klas G. Malmqvist
Lund University and Lund Institute of Technology, Lund, Sweden
I. INTRODUCTION
In 1970, Johansson et al. (1970) demonstrated that the bombardment of a specimen with
protons of a few megaelectron volts (MeV) gives rise to the emission of characteristic x-rays
and that this can form the basis for a highly sensitive elemental analysis. This landmark
article formed the starting point of the x-ray emission analysis technique which became
known as particle-induced x-ray emission analysis (PIXE). Although PIXE is sometimes
considered a variant of x-ray uorescence (XRF), such a classication is not correct in a
strict sense, because the technique does not rely on excitation of the sample by x-rays.
Instead, heavy charged particles (i.e., protons, a-particles, or heavy ions) are used in PIXE
to create inner-shell vacancies in the atoms of the specimen. As in XRF and electron-probe
microanalysis (EPMA), the characteristic x-rays produced by deexcitation of the vacancies
can be measured by either a wavelength-dispersive or an energy-dispersive detection sys-
tem. However, whereas the two detection systems are employed in both XRF and EPMA,
an energy-dispersive spectrometer with a Si(Li) detector is almost exclusively used in PIXE.
The incident charged-particle beams in PIXE are invariably generated by particle accel-
erators. For the great majority of PIXE work, protons of 14MeV, which can be produced by
small accelerators (e.g., Vande Graaaccelerators or compact cyclotrons), are employed. Such
small accelerators are also used in other analytical techniques that utilize ions of a few mega-
electron volts per mass unit, such as Rutherford backscattering spectrometry (RBS), nuclear
reaction analysis (NRA), charged-particle activation analysis (CPAA), and accelerator mass
spectrometry (AMS). Because of their common use of ion beams, these techniques and PIXE
are often jointly referred to as ion beam analysis (IBA) techniques. Furthermore, because
the same incident particle type and energy may be used in several of the techniques, the
simultaneous analysis of a sample by PIXE and some other IBA techniques (particularly RBS
and NRA) is often feasible, so that the elemental coverage may be increased down
to the very light elements and=or information on the depth distribution may be obtained.
Compared to x-rays, protons or other heavy charged particles have the advantage
that they can be focused by electrostatic or (electro)magnetic lenses and may be trans-
ported over large distances without loss in beam intensity. As a result, incident uence
Copyright 2002 Marcel Dekker, Inc.
densities (expressed as the number of impinging particles per square centimeter and per
second) are generally much higher in PIXE than in ordinary tube-excited XRF. Moreover,
focusing of particle beams down to micrometer sizes is possible, so that PIXE allows an
analysis with high spatial resolution. The microbeam variant of PIXE has become known
as micro-PIXE, whereas the common variant, which makes use of a millimeter-sized beam,
is now often referred to as macro-PIXE. Focusing to micrometer sizes is also possible with
electrons and has given rise to the powerful EPMA technique, but heavy charged particles
have the clear advantage that they give rise to much lower continuum background in-
tensity in the x-ray spectrum. As a result, the relative detection limits (micrograms per
gram) are typically two orders of magnitude better in micro-PIXE than in EPMA.
After the initial experiment by Johansson et al. (1970), the favorable characteristics
of PIXE were rapidly realized by many researchers, particularly within the nuclear physics
community. Its applicability and potential for solving numerous trace element analytical
problems were extensively examined and abundantly demonstrated. In addition to the
traditional bombardment in vacuum, external beam approaches (with the specimen either
in the laboratory air or in a nitrogen or He atmosphere) were also attempted and were
found to be useful, particularly in examining delicate and=or large objects. The progress of
PIXE over the years can be followed from the proceedings of the eight international
conferences exclusively dedicated to the PIXE technique and its applications (Johansson,
1977, 1981; Martin, 1984; Van Rinsvelt, 1987; Vis, 1990; Uda, 1993; Moschini and
Valkovic , 1996; Malmqvist, 1999) and of the ve conferences on nuclear microprobe
technology and applications (Grime and Watt, 1988; Legge and Jamieson, 1991; Lindh,
1993; Yang et al., 1995; Doyle et al., 1997), which contain numerous papers on micro-
PIXE. By now, PIXE has evolved into a rather well-used and mature technique, as is
demonstrated by the increasing numbers of research papers in which PIXE provided the
analytical results and by the publication of the two textbooks on the technique (Johansson
and Campbell, 1988; Johansson et al., 1995). For a comprehensive treatment of PIXE and
its applications, a reading of these books is highly recommended.
II. INTERACTIONS OF CHARGEDPARTICLES WITHMATTER,
CHARACTERISTICX-RAYPRODUCTION, AND CONTINUOUS PHOTON
BACKGROUNDPRODUCTION
A. Interaction of Charged Particles with Matter
1. Slowing Downof Charged Particlesin Matter: Stopping Power
When a beam of heavy charged particles of a few megaelectron volts per atomic mass unit
penetrates into matter, it loses its energy gradually with depth, until it is nally stopped.
The energy loss occurs mainly through inelastic Coulombic encounters with bound elec-
trons, and in contrast to the case of electron beams, the direction of travel of an ion beam
is scarcely altered during the slowing-down process.
The stopping power S(E) of an ion with energy E is dened as the energy loss per
unit mass thickness traversed:
SE
1
r
dE
dx
1
where r is the density of the stopping material and x is the distance. As dened here, S(E)
is expressed in units of kiloelectron volt per gram per square centimeter.
Copyright 2002 Marcel Dekker, Inc.
Numerous experimental measurements of stopping powers are available. They
formed the database to t the parameters of semiempirical equations (Anderson and
Ziegler, 1977; Ziegler et al., 1985), which are now commonly used to obtain the stopping
powers for all elements of the periodic table. For the energy range of 14 MeV, which is
most important in PIXE, the uncertainty of the values calculated with the semiempirical
equations is estimated at less than 3%.
The stopping power for compounds or more complex matrices is obtained from
those of the constituent elements through the BraggKleemann additivity rule:
S
matr
E
X
n
i1
w
i
S
i
E 2
where w
i
and S
i
E are the mass fraction and stopping power of constituent element i,
respectively.
The total path length R of an ion may easily be obtained by integration of the
stopping powers:
R
Z
0
E
0
dE
SE
3
where E
0
is the incident ion energy. Although the total path length is larger than the
projected range, the dierence between the two is smaller than 1% for incident protons of
a few megaelectron volts.
2. Inner-Shell Vacancy Creation: Ionizationand X-ray Production Cross Sections
Many of the Coulombic interactions between protons or heavier ions and matter result in
the ejection of inner-shell electrons. It is those interactions and their cross sections that are
of importance in PIXE. Three basic theoretical approaches have been used to calculate the
cross sections for inner-shell vacancy creation: the binary encounter approximation (BEA),
the semiclassical approximation (SCA), and the plane-wave Born approximation (PWBA).
The PWBA model, which applies perturbation theory to a transition from an initial
state (plane-wave projectile and bound atomic electron) to a nal state (plane-wave pro-
jectile and ejected continuum electron), has been most elaborated, particularly by Brandt
and co-workers. These researchers incorporated a series of modications in the model to
correct for its inherent approximations, and this resulted in the so-called ECPSSR treat-
ment (Brandt and Lapicki, 1979, 1981) of K and L shell ionization cross sections. The
ECPSSR treatment deals with the deection of the projectile due to the nuclear Coulomb
eld (C), perturbation of the atomic stationary states (PSS) by the projectile, relativistic
eects (R), and energy loss (E) during the collision.
Cohen and Harrigan (1985) used the ECPSSR model to produce an extensive ta-
bulation of K and L subshell ionization cross sections for most target elements and for
protons and helium ions between 100 keV and 10 MeV. As in the original ECPSSR version
of Brandt and Lapicki (1981), Cohen and Harrigan (1985) employed nonrelativistic hy-
drogenic wave functions to describe the atomic electrons. More elaborate relativistic
DiracHartreeSlater (DHS) wave functions within a Brandt and Lapicki (1979) form-
alism were used by Chen and Crasemann (1985, 1989) to produce K, L, and M shell io-
nization cross sections for protons of a few selected energies from 100 keV to 5 MeV and
for a narrow range of selected target elements.
Copyright 2002 Marcel Dekker, Inc.
For K shell ionization with protons, the cross sections as predicted by the ECPSSR
and other theories were thoroughly compared with experimental data by Paul and co-
workers (Paul and Muhr, 1986; Paul, 1987, 1989; Paul and Sacher, 1989) and a set of
reference cross sections was derived (Paul and Sacher, 1989). Although no theory emerges
that will predict the experimental data within a few percent for all target elements and
energies (Paul, 1989), it is generally agreed (Johansson and Campbell, 1988; Cohen, 1990;
Campbell et al., 1990) that both the tables of Cohen and Harrigan (1985) and of Chen and
Crasemann (1985, 1989) are adequate for most K shell proton PIXE work. For the case of
the L ionization, the situation is less satisfactory, however, as is discussed in detail by
Johansson and Campbell (1988), Campbell (1988), Cohen (1990), and Johansson et al.
(1995). Orlic and co-workers compared a large number of experimental L shell ionization
cross section data for proton impact with the ECPSSR theoretical predictions (Orlic et al.,
1994) and derived reference cross sections (Smit and Orlic , 1994) in a similar manner as
was done for the K shell by Paul and Sacher (1989).
For practical purposes, one requires x-ray production cross sections for individual
x-ray lines (i.e., s
X
p
with p the x-ray line used for analysis). The x-ray production cross
sections are related to the ionization cross sections through the following equations. For K
x-rays,
s
X
Kp
s
I
K
o
K
G
Kp
G
K
4
where the index K refers to the K shell, s
I
K
is the K shell ionization cross section, o
K
is the
uorescence yield and G
Kp
=G
K
is the fractional radiative width for line p.
For L x-rays,
s
X
Li;p
s
X
Li
G
Li;p
G
Li
5
where Li (with i 1; 2; 3) refers to the ith L subshell, s
X
Li
is the x-ray production cross
section for the subshell Li whose vacancy is lled up by the radiative transition, and
G
Li;p
=G
Li
is the fractional radiative width. The x-ray production cross sections s
X
Li
are
related to the L subshell ionization cross sections s
I
Li
through
s
X
L1
s
I
L1
o
1
6
s
X
L2
s
I
L2
s
I
L1
f
12
o
2
7
s
X
L3
s
I
L3
s
I
L2
f
23
s
I
L1
f
13
f
12
f
23
f
0
13
o
3
8
where o
i
denotes the uorescence yield, f
ij
denotes the CosterKronig probability, and f
0
13
is the small radiative intrashell vacancy transfer probability.
Calculation of the x-ray production cross section from the ionization cross sections
thus involves additional atomic parameters; that is, uorescence yields and fractional
radiative widths in the case of the K shell and uorescence yields, CosterKronig yields,
and fractional radiative widths for the L shell. As to the K fractional radiative widths,
either the experimental data of Salem et al. (1974) or the theoretical values of Scoeld
(1974a) that are derived from DiracHartreeFock (DHF) calculations are considered
accurate within 1%. There is, however, a small failure of the theory in the atomic number
region 22 < Z < 30, where experimenters (Perujo et al., 1987) agree on a deviation of a
few percent. The K uorescence yields are usually taken from Krause (1979). These data
Copyright 2002 Marcel Dekker, Inc.
are quite accurate for atomic numbers above Z 20, but the situation for the lighter
elements is less clear (Maenhaut and Raemdonck, 1984; Paul, 1989). For these lighter
elements, one currently prefers the data from Hubbell (1989). For the case of the L x-rays,
Cohen and Clayton (1987) and Cohen (1990) suggest employing the fractional radiative
widths of Salem et al. (1974) and the Krause (1979) uorescence and CosterKronig yields,
and Cohen and Harrigan (1986) used this approach to convert their table of ECPSSR L
ionization cross sections (Cohen and Harrigan, 1985) into a very useful table of produc-
tion cross sections for up to 16 individual L x-ray lines. Campbell (1988), however, ad-
vocates the use of the DHF radiative widths of Scoeld (1974b) and of the DHS
uorescence and CosterKronig yields of Chen et al. (1981) in combination with the
tabulated DHS L ionization cross sections of Chen and Crasemann (1985, 1989).
A very practical alternative to tabulated theoretical ionization and x-ray production
cross sections, particularly with computer calculations in mind, are parameterized or
analytical formulas which are obtained by tting polynomial expressions to theoretical or
empirical cross-sectional data. The equations derived by Johansson and Johansson (1976)
have been widely used by PIXE analysts in the past, but it is now known that they pro-
gressively underpredict cross sections with increasing Z of the target element. More ac-
curate formulas, in which similar functions were often used as in the Johansson and
Johansson equations, have been presented by Paul (1984), Cohen and Clayton (1987),
Miyagawa et al. (1988), Johansson and Campbell (1988), and Sow et al. (1993).
For the proton energy range of 14 MeV typically used in PIXE, the ionization (and
x-ray production) cross sections increase with increasing proton energy and decrease with
Figure 1 K and L shell ionization cross sections in barns (1 barn 10
724
cm
2
) as a function of
proton energy and target atom. The values are theoretical ECPSSR predictions (Cohen and
Harrigan, 1985). (From Johansson and Campbell, 1988. Copyright John Wiley & Sons, Ltd.
Reproduced with permission.)
Copyright 2002 Marcel Dekker, Inc.
increasing atomic number of the target atom. This is illustrated in Figure 1, which displays
theoretical K and total L shell ionization cross sections (Cohen and Harrigan, 1985) for
selected target elements. The very steep fall in the K ionization cross section with Z is
particularly notable.
3. Elastic Encounters
Protons or other heavy charged particles that pass near atomic nuclei can be scattered
elastically (i.e., without causing nuclear or atomic excitation). The cross section for the
elastic scattering is strongly dependent on the scattering angle and on the atomic number
(Z) of the scattering nuclide; it decreases with increasing angle and increases with increasing
Z. The energy of the scattered particles also increases with Z. By measuring the elastically
scattered particles, information can be obtained on the elemental composition of the sample
and on the distribution of the elements with depth. This has given rise to the widely used
technique of Rutherford backscattering spectrometry (Chu et al., 1978), which can often be
elegantly employed in combination with PIXE to provide complementary information. On
the other hand, the energetic scattered particles can also cause problems in PIXE. Indeed,
when they enter the Si(Li) detector, their large energy is deposited there, and this gives rise
to serious resolution deterioration and other electronic problems in many x-ray pulse
processing systems (Mingay, 1983). A solution to this problem is to interpose a light-ele-
ment absorber (preferably Be) between specimen and Si(Li) detector (e.g., Maenhaut and
Raemdonck, 1984), but this invariably results in loss of sensitivity for the light elements.
4. Nuclear Reactions
In addition to elastic encounters between incident particles and target nuclei, various in-
elastic interactions or nuclear reactions are possible. They include ( p, g), ( p, p
0
g), and
( p, ag) reactions when protons are used as incident particles. The cross sections for those
reactions vary in a rather irregular way with target nuclide and with incident particle
energy. They generally increase with increasing energy but may exhibit intense resonance
peaks at particular energies. Also, because of the Coulomb barrier, the cross sections are
smaller for the heavier target elements than for lighter ones. By detecting the promptly
emitted g-rays or charged particles for these nuclear reactions, elemental analysis and
depth proling of certain elements is possible. The analytical technique that employs these
possibilities is referred to as nuclear reaction analysis. A comprehensive treatment of this
technique can be found in the Handbook of Modern Ion Beam Materials Analysis (Tesmer
et al., 1995). The technique is also briey presented in Sec. VII.B.
5. Other Interactions
As indicated earlier, the interactions of charged particles and matter occur mainly through
inelastic Coulombic encounters with bound electrons. This results in electron excitation
and ionization but also in secondary phenomena that contribute to the continuous photon
background in the PIXE spectrum. Another interaction which contributes to that back-
ground is projectile bremsstrahlung. All these interactions are discussed in some detail here.
B. Continuous Photon Background Production
1. Electron Bremsstrahlung
The characteristic x-ray lines in a PIXE spectrum are superimposed on a continuum
background that has a strong resemblance to that observed in EPMA. In both cases,
Copyright 2002 Marcel Dekker, Inc.
electron bremsstrahlung is the major background component. Although this originates
from primary electron interactions in EPMA, secondary phenomena are at its origin in
PIXE. In the mid-1970s, Folkmann and various co-workers made very signicant con-
tributions to our understanding of the continuum background in PIXE [e.g., Folkmann
et al. (1974a, 1974b), Folkmann (1976)]. This work was later complemented and extended
by Ishii and Morita (1984, 1987, 1988, 1990). It is by now clear that the electron brems-
strahlung in PIXE originates essentially from three processes: quasi-free electron brems-
strahlung (QFEB), secondary electron bremsstrahlung (SEB), and atomic bremsstrahlung
(AB). SEB is formed by a two-step process: the incident particle rst ejects an electron
from a target atom, and the secondary electron is subsequently scattered in the Coulomb
eld of a target nucleus, thus producing the bremsstrahlung. The photon spectrum of SEB
is characterized by an end-point energy T
m
4m
e
E
p
=M
p
, with m
e
and M
p
the electron
and projectile masses and E
p
the projectile energy. Above T
m
, the intensity of SEB de-
creases rapidly. QFEB is emitted when an electron of a target atom is scattered by the
Coulomb eld of the projectile (this is a process in the projectile frame). The QSEB end-
point energy T
r
is equal to T
m
=4. The process AB occurs when a bound target electron is
excited to a continuum state by the projectile and, returning to its original state, emits a
photon. The relative contributions of QFEB, SEB, and AB to the electron bremsstrahlung
background are schematically shown in Figure 2. As can be seen from Figure 2, AB
predominates in the high-energy part of the spectrum (i.e., for photon energies above T
m
),
whereas QFEB becomes the prevailing component at low photon energies (below T
r
). As
both T
r
and T
m
increase linearly with increasing projectile energy E
p
, the PIXE spectrum
(which typically extends from 0 at about 2030 keV) has a quite dierent appearance
depending on the value of E
p
. In addition to E
p
, the matrix composition of the target also
plays a critical role in both the shape and intensity of the electron bremsstrahlung back-
ground. The intensity of this background per unit mass thickness is roughly proportional
to the average Z of the matrix.
The electron bremsstrahlung is emitted anisotropically and is lower at forward and
backward angles than at 90
, 67.5
, or 90
. When a
much smaller angle (035 mrad) is used, the MeV proton (or a-particle) beam can be
totally reected from the surface (Vis and Van Langevelde, 1991). This phenomenon
forms the basis of total-reection PIXE (TPIXE). In TPIXE, characteristic x-rays and
bremsstrahlung (and g-ray) background are only produced from a very shallow surface
layer. This feature can advantageously be used to increase the ratio of the characteristic
X-ray signal intensity to the background when analyzing thin layers on a thick substrate,
such as top layers on Si wafers or dried solutions on a backing lm. In normal PIXE,
the bulk of the Si wafer or the backing lm will also provide a substantial contribution
to the background intensity, whereas this contribution is absent or at least much reduced
in TPIXE. The principle and applicability of TPIXE are similar to those of total-
reection XRF (TXRF). Pioneering work on TPIXE was done by the Amsterdam PIXE
group and further details on the technique can be found in several publications
from that group (Vis and Van Langevelde, 1991; van Kan and Vis, 1995, 1996, 1997;
van Kan, 1996).
III. INSTRUMENTATION
A. Accelerators
The ion beams used for PIXE analysis are produced in an accelerator. The energy range
required, 14 MeV=u, means that relatively small accelerators are sucient. Electrostatic
accelerators or cyclotrons are suitable sources (Scharf, 1989). Electrostatic accelerators are
most commonly used. They may be of a regular or modied Van de Graa type, in which
a high-voltage terminal is charged by a belt or a metal chain. A second type of electrostatic
accelerator uses a high-frequency voltage multiplication stage to charge the terminal, the
ions being accelerated in the electrostatic eld created by the terminal. Cyclotrons have
been employed at some laboratories. These are particularly suitable for PIXE at higher ion
energies (Durocher et al., 1988; McKee et al., 1989; Peisach and Pineda, 1990) and in
certain combinations of PIXE with complementary techniques, such as particle elastic
scattering analysis (PESA) or forward alpha scattering (FAST) (Cahill et al., 1984).
In most analytical situations in which PIXE is applied, the choice of accelerator is
governed by access to a particular machine rather than by free choice. When a dedicated
accelerator can be selected, however, usually a single-ended or tandem electrostatic ma-
chine of 23 MV is chosen. Such accelerators may also favorably be employed for a
combined use of ion implantation and ion-beam analysis, including PIXE.
Certain types of analysis, such as analyses by microbeam techniques, impose special
requirements on ion source brightness, energy stability, beam emittance, and minimum
scattering of the ion beam. Such requirements may play an important role in selecting an
accelerator.
It is often stated that the need for an accelerator in PIXE is a major drawback, for
reasons or both technical complexity and economy. Although such objections contain
some truth, it should be pointed out that the prices of small electrostatic accelerators are of
the same order as those for complex traditional analytical equipment, such as mass
spectrometers or electron microscopes, and that a comparable technical skill is required
of the operator. It is therefore unjustied to rule out the PIXE method for such reasons.
Copyright 2002 Marcel Dekker, Inc.
B. BeamHandling and Target Chambers
1. Macro-PIXEinVacuum
Commercially available vacuum chambers used as scattering chambers in nuclear physics
experiments are generally not well suited for PIXE. Their shapes and dimensions hamper
obtaining large solid angles for detection and introducing multispecimen holders and
sample changers. As a consequence, few existing PIXE chambers are of commercial origin.
There are exceptions to this rule, however, and certain chambers have been built entirely
from commercially available components.
Figure 3 shows the design and principal components of a hypothetical typical
setup for macro-PIXE analysis. In the following, the details of such a system are outlined.
The ion beam emerging from the accelerator rst passes through an analyzing
magnet, which sorts out the ions of the correct mass and velocity, and is then normally
focused by electrostatic or magnetic quadrupole lenses onto the specimen.
A typical experimental facility for macro-PIXE analysis employs a system for pro-
ducing a homogeneous beam so that the specimen is evenly irradiated. This is required for
quantitative analysis of heterogeneous samples. There are various methods of achieving
this:
1. Focusing the beam onto a scattering foil, which distributes the ions homo-
geneously over the cross section selected by a diaphragm. The foil material
should be able to withstand the ion beam for long times. It is often made of thin
metal, such as gold, which is an excellent scatterer with low stopping power for
the ions.
2. Rastering the beam over the sample by electrostatic deection, for example.
This does not give a genuinely homogeneous distribution and may, therefore,
produce high instantaneous count rates in heterogeneous samples. On the other
hand, this method has the advantage of relatively low loss of beam intensity,
whereas such a loss may be more than 90% in the case of method 1.
3. After the initial alignment of the beam, straightforward symmetric defocusing by
the quadrupole magnets.
After homogenization, the beam is dened by a pair of collimators (diaphragms), which
are normally circular with diameters between 1 and 10 mm and placed approximately
Figure 3 Typical experimental arrangement for routine PIXE, with detectors for x-rays and
g-rays. Normally, the collimators can be changed between 1 and 10 mm diameter, and the x-ray
absorber holder would contain 510 different filters.
Copyright 2002 Marcel Dekker, Inc.
100200 mm apart, just before the entrance to the irradiation chamber. The collimator
material should preferably withstand bombardment by the ions over long periods and
produce low background in the PIXE spectra. This can be achieved by using a material
containing nuclei with low cross sections for producing g-rays at the ion energy interval in
question. Common choices of collimator materials are tantalum and graphite (Johansson
et al., 1995). To measure the beam uence on the target accurately, the collimators should
be grounded and electrically shielded from the irradiation chamber in which the target is
placed, and secondary electrons generated at these collimators should be prevented from
entering the target chamber. A target holder denes the proper position of the sample
during analysis. In a typical setup for PIXE analysis, in vacuo, several samples are si-
multaneously inserted into the vacuum, so that a high sample throughput is facilitated.
The normal slide frame format (565 cm
2
) is a standard target frame in many laboratories,
and often 3680 target frames are contained in vacuum. The samples are irradiated at an
angle suitable to detect the x-rays, and often the detector views the sample from a
backward angle to 135
), the CrKa intensity is raised by about 50% as a result of the secondary uorescence
from the FeK x-ray lines (Campbell et al., 1989).
As for thin specimens, several quantitation approaches are possible for thick spe-
cimens. If one relies on the fundamental parameter approach and thus solves Eqs. (11) and
(12) for C
Z
, other parameters are needed in addition to those already required for the thin-
specimen case. These additional parameters are m
p
and SE, the mass attenuation co-
ecient for line p in the sample matrix and the matrix stopping power, respectively. The
values of those parameters for the sample matrix can be obtained from those for the
matrix constituents by employing Braggs additivity rule, as already indicated in Sec.
II.A.1 for the matrix stopping power. Such calculations require knowledge of the matrix
composition and databases for the mass-attenuation coecients in the various elements
and for the elemental stopping powers. The stopping power database and its accuracy
were dealt with in Sec. II.A.1. The problem of selecting an accurate database for the mass-
attenuation coecients is the same as in all other x-ray emission techniques and will not be
discussed here. The matrix elemental composition places the major burden on the calcu-
lations and usually contributes most to the uncertainty in the calculated m
p
and SE
Copyright 2002 Marcel Dekker, Inc.
values for the sample matrix. For heavier-element matrices, in which all matrix elements
are detected in the PIXE spectrum, iterative procedures can be applied to obtain the
matrix composition, but for light-element matrices, one must resort to a priori information
(e.g., obtained by other techniques) or certain assumptions have to be made (e.g., that the
elements are present as oxides). In any case, this problem of the matrix composition
is common to all x-ray emission analysis techniques. When specimens of intermediate
thickness are analyzed, the transmitted proton energy E
f
(or rather the energy loss
E
0
E
f
) is also needed for evaluating the integral in Eq. (11). This implies knowledge (or
determination) of the specimen mass thickness, because the energy loss is related to the
latter through the matrix stopping power. Alternatively, the energy loss can be measured
experimentally. It is evident that any uncertainty in the specimen mass thickness (or in the
experimental energy loss) will also be transmitted to the value of C
Z
. The magnitude of
this uncertainty transmission increases with decreasing specimen thickness and, ultimately,
a given relative uncertainty in the specimen mass thickness produces an identical relative
uncertainty in the value of C
Z
, as is, in fact, also the situation for the innitely thin
specimens just discussed. The uncertainty transmission from the matrix composition, from
the databases, and for intermediately thick specimens also from the specimen mass
thickness can be much reduced by the use of an internal standard element. Indeed, the
uncertainty transmitted in the integral of Eq. (11) is, to a large extent, in the same sense for
the analyte and the spike, so that a signicant uncertainty reduction occurs when dividing
the two integrals.
An alternative to the pure fundamental parameter quantitation approach is to make
use of experimental thin-target sensitivities k
p
Z, as dened by Eq. (10), so that Eq. (11)
can be written as
Y
p
Z
k
p
ZNC
Z
s
X
pZ
E
0
= sin y
Z
E
f
E
0
s
X
pZ
ET
p
E
SE
dE 13
By solving this equation for C
Z
, one basically uses a relative method (relative to thin-lm
standards), but the correction for matrix eects is made by a fundamental parameter ap-
proach. As in the relative quantitation method for thin specimens, this mixed approach
requires no knowledge of the absolute detection eciency or of the radiative transition
probabilities and uorescence yields [the latter two parameters cancel out in the ratio of the
x-ray production cross sections in Eq. (13)], but ionization cross sections and Coster
Kronig yields are still required. However, the division of the x-ray production cross sections
also has the eect that the impact of the CosterKronig yields is marginal and that for
ionization cross sections, essentially only their dependence on proton energy is needed,
which has a much smaller uncertainty than the absolute value of the ionization cross section.
In the analysis of innitely thick specimens, one can also utilize experimental thick-
target calibration factors instead of relying on the fundamental parameter approach or on
experimental thin-target sensitivities. The thick-target calibration factors incorporate the
integral of Eq. (11) and are usually expressed in x-ray counts/mC and per mg=g. They are
commonly derived from PIXE measurements on samples with known trace element
composition (standards). In a strict sense, the thick-target factors are only valid for the
analysis of unknown samples with identical (matrix) composition to the standards, but in
practice, some variability in composition can be tolerated or corrected for. The necessary
correction factor is, in this case, the ratio of the integral of Eq. (11) for the standard to the
corresponding integral for the unknown.
Copyright 2002 Marcel Dekker, Inc.
As discussed by Johansson and Campbell (1988) and Johansson et al. (1995), still
other quantitation approaches are possible, for example, making use of thick single-
element standards.
Before closing this section on quantitation for semithick and innitely thick speci-
mens, it should be warned that Eqs. (11)(13) are, in a strict sense, only valid for perfectly
at homogeneous samples and that, for specimens made up of particulate material, the
particle size should be as small as possible (ideally below 1 mm when quantifying light
elements, such as Na through Si). The surface roughness eects in PIXE were dealt with by
Cookson and Campbell (1983). With regard to particle size eects in PIXE, these are
especially of concern when analyzing deposits of atmospheric particulate material (aero-
sols) on lter or cascade impactor samples. Model calculations for proton-induced x-ray
generation in a monolayer of spherical particles having realistic matrix compositions were
presented by Jex et al. (1990). In aluminosilicate particles of 2 mm diameter, the attenua-
tion is over 20% for the elements Na to P.
D. Detection Limits inThin- and Thick-Target PIXE
As in other spectrometric techniques, the detection limits in PIXE are determined by the
sensitivity (calibration) factors, on the one hand, and by the spectral background intensity
where the analyte signal (x-ray line) is expected, on the other hand. Various denitions for
the limit of detection (x
L
) were proposed in the early years of PIXE (Johansson and
Campbell, 1988), but it is now general practice to dene x
L
as that amount (or con-
centration) of analyte element that gives rise to a peak area which is equal to three times
the standard deviation (square root) of the background intensity N
B
in the spectral in-
terval of the principal x-ray line. The spectral interval for integration of N
B
is usually
dened in terms of the full width at half-maximum (FWHM) of the principal x-ray line,
but regions of one, two, and three FWHMs have been used in the PIXE literature; this
range of choice introduces a variation of 1.7 in detection limits deduced from the same
dataset.
For thin specimens, the relationship between line intensity Y
p
Z and analyte con-
centration C
Z
was given by Eq. (9). If we represent the probability for the production of
continuum background radiation per unit of x-ray energy by s
B
and if we further assume
(to keep the formulation simple) that the background originates from a single matrix
element with atomic mass A
B
, the following relation can be written for the background
intensity N
B
:
N
B
N
0
s
B
E
0
nFWHME
p
Nrt
A
B
sin y
14
where nFWHM indicates the spectral interval used for summation of the background, and
all other symbols have the same meaning as in Eq. (9). By setting Y
p
Z equal to 3N
1=2
B
and
solving Eqs. (9) and (14) for the x
L
value of C
Z
, one obtains
x
L
Z
3A
Z
s
X
pZ
E
0
s
B
E
0
nFWHM sin y
N
0
A
B
E
p
Nrt
s
15
It thus appears that x
L
is proportional (or inversely proportional) to the square root of the
experimental parameters FWHM, E
p
; N, and the specimen mass thickness. Hence, to op-
timize x
L
, a detector with very good resolution has to be used, and, rather obviously, the
solid angle of detection should be made as large as possible, but improvement in this
Copyright 2002 Marcel Dekker, Inc.
parameter is limited by the area of the detector and by the fact that the detector can only
get to about 2 cm from the specimen. Much more exibility is provided by the number of
incident protons N (preset charge), which can be increased either by a longer measurement
time or by an increase in beam current.
Both theoretical calculations and experimental measurements have been performed
to obtain x
L
values for thin specimens [e.g., Folkmann (1976), Johansson and Johansson
(1976), Ishii and Morita (1988)]. In most of this work, one adopted (or employed) a light-
element matrix (typically carbon or an organic polymer) and a specimen mass thickness of
1 or 0.1 mg=cm
2
. Johansson and Johansson (1976) produced a very useful contour plot of
x
L
values as a function of incident proton energy and atomic number of the analyte ele-
ment for the case of 0.1-mg=cm
2
-thick carbon matrix. Their plot, reproduced here in
Figure 9, was based on experimental measurements of the continuum background, and it
was further assumed that elements with atomic number up to about 50 are determined
through their Ka x-ray line and the heavier elements through their La line. As can be seen
in Figure 9, there is a valley of optimum detection limits for both the K and L cases, with
the best Kx
L
values (less than 0.5 mg=g) obtained at lower proton energy than the best Lx
L
values (0.51 mg=g). Furthermore, within either the K or L case, the bombarding energy for
optimum detection limits depends on the atomic number of the analyte elements of
interest, with higher bombarding energies favoring the heavier elements. Selection of the
energy should thus be made with the objective of the analysis in mind, but, in practice,
some compromise is necessary. Johansson and Johansson (1976) concluded from their
contour plot that the optimum proton energy is about 2 MeV for the analysis of trace
elements in biological and environmental samples. Such bombarding energy also has the
advantage that the x
L
values show rather little variation (about one decade only) for the
analyte elements with Z between 15 and 90. More recently, Ishii and Morita (1988)
produced a contour plot similar to that of Johansson and Johansson (1976), but they
based it solely on theoretical calculations and adopted a pure oxygen matrix (which was
considered representative for biological samples). The conclusion of this study was that the
Figure 9 Contour plot of the limit of detection (x
L
) as a function of incident proton energy and
atomic number of the analyte element for the case of a 0.1-mg=cm
2
-thick carbon matrix.
Experimental conditions: detector FWHM 165 eV, solid angle of detection 38 msr, collected charge
10 mC. The background interval selected for calculating x
L
was equal to one FWHM. (From
Johansson and Johansson, 1976, with permission from Elsevier Science.)
Copyright 2002 Marcel Dekker, Inc.
best detection limits were obtained with about 3 MeV protons. Considering these in-
vestigations on x
L
values, it is thus no surprise that most PIXE laboratories employ
proton energies in the range of 23 MeV. Several laboratories actually consistently use the
same bombarding energy (e.g., 2.5 MeV) for their typical applications.
Detection limits have also been examined for other projectiles than protons, but the
results were generally not very encouraging. A noteworthy exception is given in the de-
tailed study by Johansson (1992) on the use of 5-MeV helium ions (a-particles). It was
found that these oer signicantly improved detection limits (about three times better than
protons) for thin organic specimens.
The detection limits in Figure 9 are expressed in relative units (concentrations), but
they are easily converted into absolute units (masses) by multiplying by the specimen mass
probed by the beam. Because this mass is invariably small (e.g., for a 0.1-mg=cm
2
-thick
specimen as in Figure 9 and a typical beam size of 0.2 cm
2
, the probed mass is 0.02 mg),
it follows that the absolute detection limits are in the range of 10100 pg for the case of
2-MeV protons and the experimental parameters of Figure 9.
As indicated by Eq. (15), the concentration limit of detection improves with the
square root of the mass thickness. However, this relation remains only valid as long as the
thin-specimen criterion holds. Because of the existence of matrix eects (i.e., the decrease
in s
X
pZ
and in the cross sections for bremsstrahlung production with decreasing proton
energy and the attenuation of the characteristic x-rays and continuum radiation by the
sample matrix), x
L
rapidly approaches its optimum value. In general, the improvement in
x
L
is quite limited above a few mg=cm
2
. Depending on the origin of the continuum
background and the energy dependence of its production cross section, x
L
may actually
deteriorate somewhat beyond a certain specimen thickness. Such situations may occur for
analyte lines in the spectral region where prompt g-radiation forms the major background
source (typically above about 10 keV).
Whereas x
L
studies for thin specimens have concentrated on light-element matrices,
in similar investigations for innitely thick targets, heavy-element matrices were also
considered. Teesdale et al. (1988) conducted a comprehensive experimental study of x
L
values for 15-MeV proton bombardment of pure single-element matrices of carbon,
aluminum, silicon, titanium, iron, germanium, molybdenum, silver, tin, ytterbium, and
lead. They paid particular attention to the choice of appropriate x-ray lters for sup-
pression of the matrix characteristic x-ray lines and their pileup peaks and to the choice of
the optimum proton energy. It was found that increase of energy up to 3 MeV is protable,
but that a further increase confers only small benets. Increasing absorber thickness were
suggested for increasing atomic number of the matrix up to Z 40. Under these condi-
tions and using a beam charge of 1 mC, the x
L
values were a few micrograms per gram for
the light-element matrices (C and Al) and 1050 mg=g for the intermediate matrices Ti
through Ge.
By using an optimized PIXE setup and a preset charge of 100 mC (which corresponds
to a 15-min bombardment with 100 nA current), x
L
values down to a few tenths of mi-
crogram per gram may be obtained for light-element matrices. The detection limit, ex-
pressed as concentration in the original sample, can be further improved by a physical or
chemical separation of the material of interest or by a chemical separation of the analyte
element(s) from the bulk of the sample. Drying or freeze-drying is an obvious pre-
concentration step for natural water samples, but also for biological tissues (which typically
contain 80% water). For natural waters with high mineral content, such as seawater, che-
mical preseparation schemes are advantageous, and detection limits of 1 ng=L (or 1 pg=g)
have been obtained by such an approach (Johansson and Johansson, 1984). For dried
Copyright 2002 Marcel Dekker, Inc.
biological tissues, further preconcentration may be realized by resorting to high- or low-
temperature ashing. As demonstrated by Pallon and Malmqvist (1981) and Maenhaut et al.
(1984), however, the gain in detection limits remains limited to a factor of about 2.
In this entire discussion on x
L
values, it was assumed that the PIXE analyses were
carried out in a vacuum chamber. In external beam or nonvacuum PIXE (in air or in a
helium or nitrogen atmosphere), poorer x
L
values are expected because of the background
contribution from interactions in the beam exit foil material and in the air or chamber gas
and, for the light elements, also because of the substantial attenuation of their soft x-rays
by the same gases. However, practical x
L
values in nonvacuum PIXE appear to be rather
comparable to the x
L
values in vacuum PIXE, at least for analyte elements with atomic
number above 25 (Ra isa nen, 1986).
E. Precision and Accuracy inThin- and Thick-Target PIXE
As in any other analytical technique, high precision and accuracy should be aimed for in
PIXE. It is therefore essential that careful attention be given to all stages of the analysis.
These include sample and specimen preparation, specimen bombardment, spectral data
processing, quantication, and correction for matrix eects. For a discussion of the critical
facets in the various stages, the reader is referred to the specic section dealing with each
stage. As far as the specimen bombardment stage is concerned, it should be added here
that one should be aware of the danger of radiation- or heat-induced losses during PIXE
bombardment. Such losses are particularly feared for volatile analyte elements (e.g., the
halogens, S, As, Se, and Hg) and, in the case of organic or biological specimens, also for
certain matrix elements (mainly H and O). The current density applied during analysis
plays a major role, and the danger for losses is therefore more severe in micro-PIXE than
in macro-PIXE. In any case, it should be determined which irradiation conditions are safe
for a particular application. More information on this subject can be found in a tutorial
paper by Maenhaut (1990a) and in research papers of Legge and co-workers (Legge and
Mazzolini, 1980; Cholewa and Legge, 1989; Kirby and Legge, 1991), Themner and co-
workers (Themner et al., 1990; Themner, 1991), and Van Lierde et al. (1995, 1997).
The reproducibility (precision) of an entire PIXE analytical procedure (including the
contribution from sample processing and specimen preparation) can be examined by
preparing several specimens from the same material, subjecting these to PIXE, and cal-
culating a standard deviation (s) from the spread in the results obtained. Under optimum
conditions, this standard deviation should be the same as that expected on the basis from
counting statistics alone (s
c
). However, when the percentage standard deviation from
counting statistics (%s
c
) approaches values smaller than about 12%, dierences between
s and s
c
are often unavoidable because of the limitations in sample and specimen homo-
geneity. It should indeed be realized that, even in macro-PIXE of innitely thick speci-
mens, the probed sample mass is at most a few milligrams, so that only nanogram amounts
of analyte elements are actually examined for concentration levels of a few micrograms per
gram. Particularly in the analysis of biological, geological, and atmospheric aerosol
samples, a %s of 12% is often the ultimate practical limit of precision. Such precisions
were obtained by Maenhaut et al. (1987), for example, in PIXE analysis of biological
reference materials.
The accuracy of a PIXE procedure should be evaluated by analyzing (certied) re-
ference materials or through comparisons with other analytical techniques. Ultimately, the
accuracy will depend on the extent of spectral interferences and matrix eects and on how
well these can be controlled or corrected for. Several PIXE analyses of reference materials
Copyright 2002 Marcel Dekker, Inc.
and other accuracy investigations have been reported in the literature, and selected studies
dealing with the analysis of trace elements in biological, environmental, and geological
samples were reviewed by Maenhaut (1987). The book of Johansson et al. (1995) also
reports on several accuracy studies and presents some ne examples of investigations in
which micro-PIXE was used. As an example of a macro-PIXE accuracy investigation on
biological materials, the study by Maenhaut et al. (1987) can be cited. A total of 18 ele-
ments were measured in up to 14 (certied) reference materials, and from a comparison of
the PIXE results with the reference values (when available), it was concluded that the
accuracy was better than 5%. For assessing the accuracy in macro-PIXE of atmospheric
aerosol samples, one has to rely on comparisons with other techniques. Good examples
of such intercomparisons are the work by Wa tjen and co-workers (Wa tjen et al., 1983;
Bombelka et al., 1984) and more recently by Nejedly et al. (1995).
V. SAMPLE COLLECTIONANDSAMPLE ANDSPECIMENPREPARATION
FOR PIXE ANALYSIS
A. General
In this section, general aspects of sample collection (sampling) and sample and specimen
preparation are discussed. Methods or procedures which only apply to samples of a
specic type (e.g., biological and environmental) are touched upon in Sec. VI, which deals
with the applications of PIXE. Furthermore, the present section discusses aspects that are
of general importance in both macro- and micro-PIXE. Points that are relevant for micro-
PIXE only are dealt with in Sec. V.B.
Particle-induced x-ray emission can, in principle, be applied to any type of sample.
Considering that the bombardments are normally done in vacuum, however, it is evident
that the technique is more suitable for analyzing solids than liquids. PIXE analysis of
liquids normally involves some preconcentration by drying (which can be as simple as
drying a drop of the liquid on a suitable backing lm) or some other physical or chemical
separation of the analyte elements from the liquid phase. As far as the analysis of solids is
concerned, it should be kept in mind that even in macro-PIXE of innitely thick samples,
the mass actually probed by the beam is at most a few milligram. Determination of the
bulk composition of a solid sample without any preliminary sample preparation is
therefore only possible for samples that are homogeneous in all three dimensions. Hence,
before deciding on analyzing a sample by PIXE, it should be carefully examined whether
the analytical problem is not much better solved by some other technique. The analysis of
liquid samples and the bulk analysis of large heterogeneous samples are clearly problems
for which PIXE may not be most appropriate. However, as concluded in a paper by
Maenhaut (1990b) in which PIXE and various nuclear and atomic spectrometric techni-
ques were compared, there are numerous analytical problems for which PIXE is the most
suitable technique or, at least, among the more suitable. Examples of such are the mul-
tielemental analysis of milligram-sized samples consisting of a light-element matrix (e.g.,
biomedical and atmospheric aerosol samples), the nondestructive analysis of millimeter-
sized areas on a large sample or of thin supercial layers on a bulk sample, and various
problems that require sensitive analysis with high spatial resolution.
When it has been decided to tackle an analytical problem by PIXE, full use should be
made of the inherent characteristics of the technique, particularly of its nondestructive and
instrumental character. Therefore, if possible, sampling should be done in such a way that
subsequent sample preparation can be avoided or kept to a strict minimum. The collection
Copyright 2002 Marcel Dekker, Inc.
of suspended atmospheric particulate material (atmospheric aerosols) is a good example
where such strategy should be adopted. For unique samples or samples of high commercial
or historical value, sample preparation or subsampling may even not be allowed because
the sample generally must be returned unaltered after the analysis. Examples of such
samples are historical documents, various objects of art, and extraterrestrial dust particles.
In many situations, however, some sample preparation is required. This may vary
from simple cleaning of the sample (to remove surface contamination), polishing (to
eliminate surface roughness eects), and powdering (to homogenize the sample and to re-
duce the particle size), to digestion or physical or chemical preconcentration or separation.
Furthermore, the last step in the sample preparation usually consists of preparing specimens
that are suitable for PIXE bombardment. Such specimen preparation may involve depos-
iting a drop of a liquid (e.g., of an acid digest) or a few milligrams of powdered material on a
clean, strong substrate lm (for thin and intermediate specimens) or pressing a certain
amount of sample into a pellet (for innitely thick specimens). The prepared specimen is
often mounted or held in a target frame (e.g., a 25-mm-diameter plastic ring or a square
target frame that ts in a standard 35-mm slide tray holder). Overall, the sample and spe-
cimen preparation procedures in PIXE are quite similar to those in the other x-ray emission
analysis techniques (see Chapter 14). As far as the backing lms for thin and semithick
specimens are concerned, an additional requirement in PIXE is that such lms should be
able to withstand the irradiation by the particle beam. More detailed information on
specimen backing lms for PIXE is given by Johansson and Johansson (1976) and Russell
et al. (1981).
Considering that PIXE analyses often aim at measuring microgram per gram levels
of trace elements and that the absolute amounts of analyte elements actually examined are
then in the nanogram region or below, contamination control is very important. Hence,
acid-cleaned plastic (e.g., polyethylene or Teon) or quartz containers and tools should be
employed during sampling and sample processing. The chemicals, acids, and water used in
sample preparation (e.g., for digestion or dissolution) should be of high purity. Also, all
critical manipulations should be done in a clean bench with laminar airow. When ap-
plying thin-specimen procedures, realistic blank specimens should always be prepared.
This should be done by applying the same procedures and using the same substrate lms
as for the actual sample specimens. Another point of concern is potential losses of analyte
elements during sample storage, sample processing, and specimen preparation. During
storage of aqueous samples, analyte elements may be deposited on the container walls.
Perhaps more important is that some analyte elements (e.g., the halogens, S, As, Se, Hg,
and Pb) may be volatilized by drying of the sample at elevated temperature and especially
in sample preparation methods which involve high- or low-temperature ashing or acid
digestion in an open vessel.
For more information about general aspects of sample and specimen preparation for
PIXE, the tutorial paper by Mangelson and Hill (1990) can be recommended. This paper
also provides a ne overview of the various physical and chemical methods of sample
preparation.
B. Specimen Preparation for Micro-PIXE
Because of the similarity between micro-PIXE and EPMA, the specimen preparation
techniques developed for EPMA are generally also applicable in micro-PIXE. However,
the dierence in ionizing particles (typically 1020-keV electrons in EPMA versus MeV
protons in PIXE) has the eect that the depth probed in the analysis is signicantly greater
Copyright 2002 Marcel Dekker, Inc.
in micro-PIXE. For example, this depth amounts to several tens of micrometers for 3-
MeV protons. In order to obtain meaningful results, the specimen should be homogeneous
throughout the depth analyzed, and for optimum use of the lateral capability of the nu-
clear microprobe, the specimen thickness should preferentially be of the same order as the
size of the focus of the microbeam. On the other hand, the specimen has to be suciently
thick (0.11.5 mg=cm
2
) in order to obtain a high x-ray yield. It should also be realized that
the support material (backing lm or other support material) may cause interference in the
x-ray spectrum, and it should therefore be selected with care.
The actual specimen preparation will, to a large extent, be determined by the ma-
terial to be studied. For biological materials, cryosectioning of frozen samples followed by
freeze-drying and mounting of the material on a thin clean plastic foil is often the method
of choice. For minerals, the specimens may consist of thin, polished, or ion-milled disks
mounted on a glass plate or an electron microscope grid. Finally, the preparation of thin
or semithick specimens should be done in such a way that the sample mass examined in
each pixel can be determined. This mass thickness is required in order to allow expressing
the results as elemental concentrations (see Secs. IV.B and IV.C).
VI. APPLICATIONS
The applicability of PIXE to various analytical problems has been amply demonstrated in
many publications. Furthermore, numerous studies have been carried out in which PIXE
provided part or all of the requested trace element concentration data. This section pre-
sents a brief selection of the applications of PIXE and is mainly based on publications
from the period 1992 through mid-1998. Many more examples can be found in the pro-
ceedings of the international conferences on PIXE (Johansson, 1977, 1981; Martin, 1984;
Van Rinsvelt, 1987; Vis, 1990; Uda, 1993; Moschini and Valkovic , 1996; Malmqvist, 1999)
and nuclear microprobe technology (Grime and Watt, 1988; Legge and Jamieson, 1991;
Lindh, 1993; Yang et al., 1995; Doyle et al., 1997), in the proceedings of the international
symposia on Bio-PIXE (Ishii et al., 1992; Zheng et al., 1996), in two textbooks on the
PIXE technique (Johansson and Campbell, 1988; Johansson et al., 1995), and in the
chapter on PIXE in the rst edition of this Handbook of X-Ray Spectrometry (Maenhaut
and Malmqvist, 1992).
A. Biological and Medical Samples
Most samples of biological origin are composed of essentially organic material, so that its
matrix elements are light and do not give rise to characteristic x-ray lines in the PIXE
spectrum. Furthermore, the electron bremsstrahlung background is lower for an organic
matrix than for a matrix of heavy elements. Biomedical samples are therefore well suited
for trace element determinations by PIXE. The disadvantage is that the organic matter is
sensitive to radiation damage and to heating eects. As discussed in Sec. IV.E, the light
matrix elements, especially H and O, but also some analyte elements may be lost to some
extent, in particular in micro-PIXE.
The elements of interest in biological materials are either essential minor or trace
elements (e.g., K, Ca, Mn, Fe, Cu, Zn, and Se) or toxic trace elements (e.g., Cd and Pb).
A comparison of the detection limits in PIXE (see, e.g., Fig. 9) with the levels of the minor
or trace elements in biological tissues [e.g., Iyengar et al., (1978), Iyengar (1989)] reveals
that most elements of interest in physiology or pathology, with a few exceptions, such as
Copyright 2002 Marcel Dekker, Inc.
Cd, can suitably be determined by PIXE in most tissue types. Consequently, ever since the
development of PIXE started, biological and medical applications have been prominent.
In the last decade, however, competition from new ultratrace techniques and the need for
high-sensitivity measurements in bioscience have resulted in growing diculties to pro-
mote PIXE and other x-ray spectrometric techniques. Furthermore, in recent years, the
emphasis has shifted from total element determination toward elemental speciation (i.e.,
measuring of the chemical compound or biological molecule to which the element is bound
or associated). PIXE and the other x-ray techniques lend themselves much less to this
speciation work than some of the new atomic spectrometric techniques. In applying PIXE
to biomedical problems, one should therefore look for cases where one can fully utilize the
special advantages of PIXE, such as high spatial resolution, accurate quantitative analysis,
and small samples. Particularly micro-PIXE, as used in the nuclear microprobe, is in-
valuable in biomedical trace element research.
1. Sampleand Specimen Preparation
The preparation of biomedical samples for PIXE analysis depends on the type of sample,
its composition, the information looked for (bulk concentrations or spatially resolved
data), the elements of interest, and the mode of irradiation (vacuum or nonvacuum). For
nuclear microprobe analyses, special requirements apply, as indicated in Sec. V.B.
Many elements exhibit suciently high concentrations in biological material that a
simple physical sample preparation method may suce in order to obtain the requested
concentration data. Several of the purely physical sample preparation methods are dis-
cussed in detail by Mangelson and Hill (1990). They include drying or freeze-drying,
homogenizing and pulverizing, and cutting of thin sections. In some cases, however,
particularly for natural levels of toxic elements and for levels of some essential elements in
certain tissue types, preconcentration by destruction of the organic matrix or some other
chemical preconcentration or preseparation is required. This may be performed in various
ways (Mangelson and Hill, 1990). The most common methods are (1) wet digestion in
acids, either in open or closed vessels, (2) dry ashing in an oven, (3) low-temperature
ashing in a plasma asher, and (4) biochemical separation techniques.
In wet digestion, the organic sample matrix is decomposed by concentrated strong
acids, normally nitric and=or hydrochloric acid. As the acid has to be removed by eva-
poration prior to PIXE analysis, the use of hygroscopic acids (i.e., sulfuric, perchloric, and
phosphoric acid) should denitely be avoided. Moderate heating in an oven or microwave-
assisted digestion signicantly increases the decomposition rate. The digestion rate may be
further increased by performing the digestion in a closed pressurized vessel [e.g., in a
Teon bomb in a regular oven (Duou et al., 1987) or in a closed vessel in a microwave
oven (Pinheiro et al., 1990)]. Such a procedure has the additional advantage that losses of
volatile analyte elements are greatly reduced. Unfortunately, losses of certain elements
(particularly of the halogens) may still occur while opening the vessel and during the
drying of the digest on the backing lm while preparing the PIXE specimen.
Dry ashing provides a greater mass reduction factor than wet ashing, with high-
temperature ashing even being better in this respect than low-temperature ashing. By dry
ashing at 550
C. This indicated that MeV Au ion implantation inhibits the As release from GaAs.
For a given annealing temperature and duration, the amount of As loss was dependent on
the dose of implanted Au. Wendler et al. (1996) measured displacements in 2-MeV Se
implanted GaAs by RBS and PIXE channeling experiments. It is also possible to combine
micro-PIXE and micro-RBS with channeling. Under channeling conditions (i.e., when the
projectiles are guided along crystal axes or crystal planes), the interaction probability is
decreased and this leads to a reduced yield of induced x-rays or scattered particles.
By comparing the yields obtained for channeling and for random orientation, information
on the crystal structure and on interstitial atoms may be obtained (King et al., 1993;
Tesmer et al., 1995). To visualize the results from micro-PIXE and micro-RBS channeling
experiments, channeling contrast microscopy (CCM) can be used. The Microanalytical
Research Centre (MARC) in Melbourne is very active in studies on frontier materials,
mainly semiconductors (Jamieson, 1997). It applied micro-PIXE, micro-RBS, and CCM
in a study on diamond ion implantation and annealing. A diamond specimen was im-
planted with 4-MeV P
particles
were used as incident ions for both types of measurements. The study indicated that a
solution annealing process is absolutely necessary for producing a good single crystal. It
was further found that the P atoms were mostly on substitutional sites of the face-centered
cubic (fcc) structure and that the MeV He
and 59
(for H).
Whereas elastic scattering spectrometry is a very useful tool in aerosol analysis, its
applicability as a complement to macro-PIXE for analyzing biological and geological
samples is more limited. As far as biological samples are concerned, scattering analysis is
generally only employed to obtain information on the mass thickness of the specimen
and=or to determine the beam uence. These types of applications are especially useful
in nuclear microprobe investigations, in which the local mass thickness of, for instance,
a microtome slice must be measured to convert the x-ray intensities from the PIXE
spectrum into concentration values (Heck and Rokita, 1984; Themner and Malmqvist,
1986).
As indicated, the depth-proling capability of RBS is an important asset for mate-
rials research. This characteristic is also very useful in investigations related to art and
archaeology. For example, RBS can be employed to measure depth proles in samples
where surface uniformities may be of high signicance, as in patina, corrosion and surface
segregation processes, glass aging, and polishing procedures (Amsel et al., 1986).
For very light elements, in particular hydrogen, it is advantageous to detect the
elastically recoiled target nuclei instead of the scattered projectiles. The technique using
this approach is termed elastic recoil detection (ERD). Because of the kinematics of the
scattering process, the particle detector has to be placed in the forward direction in ERD,
and it therefore implies that thick specimens must be bombarded at a glancing angle.
Time-of-ight techniques are very suitable for energy determination of the recoiling par-
ticle (Rijken et al., 1992). ERD is particularly useful in materials research, but it has also
been employed in other applications, such as for depth proling of hydrogen in obsidians
(volcanic glasses) (Pretorius et al., 1988), with the aim of dating them.
Copyright 2002 Marcel Dekker, Inc.
In addition to the elastically scattered particles, the transmitted beam may also be
employed to extract valuable information on thin specimens. Indeed, by placing a surface
barrier downstream from the sample and measuring the energy loss of the incident par-
ticles, the local mass thickness may be derived. Because of the small beam currents re-
quired, this approach is particularly useful in nuclear microprobe work, where it has given
rise to the technique of scanning transmission ion microscopy (STIM). This technique is
employed for high-resolution imaging of a specimen, and because an extremely low beam
dose is sucient, eects of beam induced damage can also be studied in the virgin and
damage state (Saint and Legge, 1997). By judicious optimization, it is possible to obtain
images with a spatial resolution of better than 100 nm.
B. Nuclear Reaction Analysis
Nuclear reaction analysis is based on the detection of the prompt g-rays or prompt par-
ticles emitted as a result of nuclear reactions between the incident beam and the target
nuclei (Tesmer et al., 1995). As indicated in Sec. II.A.4, the cross sections for such reac-
tions vary in a rather irregular way with target nuclide and with incident particle energy.
When using light ion beams of only a few megaelectron volts, nuclear reaction cross
sections are only important for light- and medium-weight target nuclei. Of the two forms
of NRA, the one in which the prompt g-rays are detected is by far the most common. It is
usually referred to as particle-induced g-ray emission analysis (PIGE). The prompt g-ray
measurement has the advantage over the detection of the promptly emitted charged
particles that it allows for greater exibility in the experimental setup, and for thick
samples, the PIGE spectra are much less complicated and far more easily analyzed than
the prompt particle spectra.
Proton-induced g-ray emission analysis, which employs ( p, g), ( p, p
0
g), or ( p, ag)
reactions, lends itself easily to concurrent use with PIXE for virtually all specimen types.
Moreover, it is able to provide good detection limits for several elements that are not
accessible by PIXE, so that it is a truly complementary technique. Ra isa nen (1987) carried
out extensive investigations on the applicability of PIGE for analyzing thick biomedical
specimens and reported that, under favorable conditions, the detection limits are down to
the submicrogram per gram level for Li, B, F, and Na, about 1025 mg=g for N, Mg, and P,
and 100300 mg=g for C, O, and Cl. Unfortunately, however, the optimum proton energies
for all these elements are not the same. Moreover, the detection limits in PIGE depend
strongly on the sample composition and may be much worse when some of the elements
for which the sensitivity of the method is highest (e.g., Li, B, F, and Na) are present in
elevated concentrations.
The most popular application of PIGE on biomedical samples is unquestionably
the determination of uorine. Either the reaction
19
F p; p
0
g
19
F or
19
F p; ag
16
O may be
used for this purpose, but the former oers about ve times better detection limits
(Ra isa nen, 1987). Particularly favorite study objects for PIGE uorine analyses are
teeth. For example, Coote et al. (1997) used a nuclear microprobe to measure Ca and F
in teeth from sheep by a combination of PIGE and PIXE. The enhanced F con-
centrations due to exposure in conjunction with a volcanic eruption were investigated,
and in Figure 19, results from these measurements are shown. Nuclear microprobes were
also used in several studies on human teeth. It was possible to determine the uorine
prole across the entire thickness of the enamel layer (Svalbe et al., 1984a) and across
precarious and articially induced lesions (Svalbe et al., 1984b; Coote and Vickridge,
1988).
Copyright 2002 Marcel Dekker, Inc.
Particle-induced g-ray emission is also useful as a complementary technique to PIXE
in the analysis of atmospheric aerosol samples. Several groups have examined the opti-
mum bombarding energy and=or presented procedures or systems for measuring several
light elements in such samples by PIGE (Robaye et al., 1985; Asking et al., 1987; Boni
et al., 1989). Asking et al. (1987) conclude that two proton energies, 2.64 and 2.96 MeV,
are most suited for measuring Na in thin (<0.25 mg=cm
2
) aerosol samples, and they report
Figure19 Two-dimensional distributions of fluorine, as determined by the
19
F( p, ag)
16
O reaction,
for sections of sheep incisor, with the enamel to the left: (a) nonexposed sheep; (b) sheep with teeth in
developing phase during fluorine exposure; (c) sheep with teeth in the mature state during exposure.
Dark areas indicate high concentration and light areas indicate low concentration. (From Coote
et al., 1997, with permission from Elsevier Science.)
Copyright 2002 Marcel Dekker, Inc.
a detection limit of 100 ng=cm
2
. To smooth out the variations in the cross-section curves,
Boni et al. (1989) advocate spreading out the beam energy, and they propose an incident
proton beam of 3.23.5 MeV, with rectangular energy distribution. The detection limits
obtained with their setup are 1 ng=cm
2
for Li, 3 ng=cm
2
for F, 25 ng=cm
2
for Na, and of
the order of 100 ng=cm
2
for B, Mg, Al, Si, and P.
Simultaneous PIGE=PIXE analyses of thick geological samples have been done
by Carlsson and Akselsson (1981) and by Carlsson (1984). Carlsson (1984) used
PIGE for measuring Li, F, Na, Mg, and Al, whereas heavier elements were determined by
PIXE.
Archaeological specimens are often examined by a combination of PIGE and PIXE.
Such combination was, for example, used for provenancing obsidian artifacts (Duerden
et al., 1986). The determination of 11 elements by PIXE and 3 by PIGE provided a very
good characterization of the samples, as concluded from applying principal-components
analysis on the dataset obtained. Tuurnala et al. (1986) applied external milliprobe
PIGE=PIXE to check the authenticity of oil paintings. A relatively high beam energy of
4 MeV was used to raise the penetration depth, and Na, Mg, and Al were measured by
PIGE. Still other examples of applications of PIGE in art and archaeology are given by
Demortier (1989). For analyses in this eld, the depth-proling capacity of PIGE is also
quite useful.
Although PIGE is almost exclusively performed with proton beams, deuterons are
more commonly used as incident particles when the nuclear reaction analysis is based on
the detection of the promptly emitted charged particles. Consequently, this form of NRA
is less suitable for concurrent use with PIXE. Furthermore, as indicated earlier, prompt
particle spectra are often much more complex than PIGE spectra, thus making the
technique less attractive for routine use.
As far as biomedical materials are concerned, NRA with particle detection can be
used for measuring nitrogen (an indicator of protein content). For example, Go nczi et al.
(1982) applied the
14
Nd; p
15
N nuclear reaction, with 6-MeV deuterons, to measure the
nitrogen depth prole in 1000 individual wheat grains. The nitrogen distributions showed
striking correlations with parameters describing the nitrogen content of the fertilizer, the
time of harvesting, the grain position in a head, and the analyzed variety. The (d, p) re-
actions on
12
C and
14
N were used in combination with deuteron-induced x-ray emission
(DIXE) measurements to determine the N=C and S=N ratios along a single hair (Varga
et al., 1984). Some general aspects of NRA with particle detection using protons as in-
cident particles have been discussed by Ra isa nen (1987).
As an example of the application of (d, p) reactions for analyzing atmospheric
aerosol samples, the work by Braga Marcazzan et al. (1987) can be cited. In this work,
oxygen, nitrogen, and carbon were measured in particulate matter collected on a silver
lter. In nuclear microprobe investigations on meteorites and cosmic dust particles, a (d, p)
reaction was used to measure the carbon content (Vis et al., 1987). The applicability of
particle detection NRA in art and archaeology was addressed by Amsel et al. (1986) and
Demortier (1989).
Similarly to RBS, NRA (including PIGE) is particularly useful in materials analysis.
In such applications, the depth-proling capacity of NRA (e.g., by making use of strong
resonance peaks in the cross-sectional curves) is often quite valuable. As indicated in
Sec. VI.E, however, PIXE, if applied at all, generally serves as the complementary tech-
nique, whereas the essential information is obtained by RBS or NRA. For examples of
applications of NRA (including PIGE) to materials research, the reader is referred to the
textbooks and proceedings mentioned at the beginning of this section.
Copyright 2002 Marcel Dekker, Inc.
To improve the detection limits in NRA, Martinsson and Kristiansson (1993) and
Kristiansson and Martinsson (1997) have developed a new technique, photon-tagged
NRA (pRNA). By using fast coincidence electronics and applying stringent timing criteria
between photons and charged particles, the background in the particle energy spectra is
much reduced and better detection limits are obtained for several elements. In subsequent
work, the principles of the pRNA technique were implemented in a nuclear microprobe
setup (Sjo land et al., 1997d; Kristiansson et al., 1998). Figure 20 shows the schematic
design of such a system. The technique was further developed for the microanalysis of
special light elements (e.g., H, Li, and F) (Sjo land et al., 1997a, 1997b, 1997c).
C. Chemical and Structural Information
In biomedical and environmental research, considerably more insight may often be gained
by determining the chemical species or chemical association of minor or trace elements
than merely their elemental concentrations. For example, to understand the source and
transformation processes that are responsible for particulate nitrogen in the atmosphere, it
is of interest to nd out what fraction of total nitrogen is present as ammonium, nitrate, or
some other N-containing species, and whether the ammonium is associated with the
nitrate or sulfate, for example. Similarly, chemical and=or structural data on crystalline
specimens in mineralogy are of great importance to geoscientists. The close-encounter
reactions used in PIXE and most other IBA techniques do not provide any chemical or
structural information. Such information can, however, be obtained by making use of
other interactions or by employing special experimental conditions.
Figure 20 Schematics of experimental arrangements for photon-tagged nuclear reaction analysis
in a nuclear microprobe. P: the incoming ion beam; SBD: surface barrier detector; PMT: photo-
multiplier tube. (From Sjo land et al., 1997d, with permission from Elsevier Science.)
Copyright 2002 Marcel Dekker, Inc.
1. Ion-BeamThermography
To obtain speciation information with IBA techniques and IBA-compatible aerosol
samplers, Martinsson and Hansson (1988) developed the technique of ion-beam ther-
mography (IBT). This technique consists of an ingenious use of controlled heating of the
aerosol sample, whereby its temperature is gradually raised, with concurrent analysis by
PIXE and PESA. The technique has been further developed and integrated with a data
acquisition and specimen control computer system (Mentes et al., 1996). In principle, also
simultaneous PIGE measurements are possible. To reduce or eliminate vacuum- or beam-
induced losses of compounds or analyte elements from the sample, the measurements are
performed at a prevacuum pressure ( p < 10
2
mbar) or in a 10-mbar He atmosphere. The
sample is heated by passing an electrical current through the sample substrate, which
consists of a 0.8-mm-thick aluminum foil. The temperature of the sample is monitored by
determining the change in relative resistivity (dened as the ratio of the resistance of the
heated sample to that of the unheated sample). To achieve time resolution in the data
acquisition so that the course of sample deterioration during the thermographic treatment
can be followed, a special data acquisition program has been developed. The IBT results
are presented as a set of thermograms, one for each element, which show the amount of
element remaining in the sample as the temperature increases. From these thermograms,
the chemical compounds and their concentrations are inferred. Figure 21 shows the set of
thermograms obtained for a mixture of equal amounts of NH
4
NO
3
and (NH
4
)
2
SO
4
. The
rst 10 data points, at relative resistivities between 0.9 and 1, were taken without heating
Figure 21 Thermograms from a mixture of NH
4
NO
3
and (NH
4
)
2
SO
4
. Amounts inferred:
m(N) 4.1 mg=cm
2
, m(O) 4.3 3.6 mg=cm
2
, m(S) 1.8 mg=cm
2
, m(NH
4
NO
3
) 7.2 mg=cm
2
,
m((NH
4
)
2
SO
4
) 7.4 mg=cm
2
. (From Martinsson and Hansson, 1988, with permission from Elsevier
Science.)
Copyright 2002 Marcel Dekker, Inc.
the sample and are listed in chronological order. The elemental amounts inferred from the
thermograms are given in the legend to Figure 21 and are in good agreement with those
expected on the basis of stoichiometry.
The especially interesting features of IBT are that chemical compounds (not ions) are
determined, that it is a multicompound technique, that it requires no pretreatment of the
sample, and that some compounds may be measured down to trace levels under favour-
able conditions.
2. Ionoluminescence
When a signicant amount of energy is transferred to a crystalline material, be it by elec-
trons, photons, or charged particles, near-visible light is emitted due to luminescence in the
material. The well-known techniques of photoluminescence (PL) and cathodoluminescence
(CL) are widely used in material science and geology in order to study the chemistry,
structure, or electrical properties of the material. The exact character of the impinging
radiation does not matter very much for the basic interaction mechanisms involved in lu-
minescence. As a consequence, much of the physical understanding from the PL and CL
techniques can be directly incorporated in the ionoluminescence technique (IL) that was
introduced and developed in a collaboration between the nuclear microprobe groups of
Lund and Melbourne (Yang et al., 1993). The intensity and wavelength of the ionolumi-
nescent light provide information concerning the nature of luminescence centers, such as
trace substituents and structural defects, present in the matrix. This makes IL a useful
complement to other IBA techniques, such as PIXE and, to a some extent, also RBS.
Luminescence can be divided into two subgroups (intrinsic and extrinsic) according
to its origin. The intrinsic luminescence, which usually contributes in emission through
crystal structural defects, is not related to impurities but to crystal-lattice properties. The
extrinsic luminescence depends on the impurities in a crystal through the processes acti-
vation, sensitization, and quenching. Luminescence phenomena are common in many
solids, yet the use of the light produced is restricted because of the complex physical
mechanisms behind the emission. For insulators, to which many minerals belong, crystal-
eld theory can be used to explain luminescence (Henderson and Imbusch, 1989), whereas
for semiconductor crystals, band theory is often employed (Yacobi and Holt, 1990). As
long as no secondary eects such as lattice damage or crystal modication occur, the
general luminescence properties of a material are not dependent on the nature of the
excitation source. For excitation by energetic particles, the processes leading to lumines-
cence take place in three steps. First, the energy of the MeV particles is dissipated into the
excitation volume mainly through electronic stopping and partially through nuclear
stopping processes. Ionization takes place in the excitation volume. Recombination of the
electrons and the excited ions allows the crystal lattice to absorb the energy released and
the optical system becomes highly excited. The second stage involves the de-excitation of
the states of high excitation through radiationless transitions. The third stage, lumines-
cence emission, occurs when the atoms de-excite from a low-excitation state to the ground
level. The term activation is used when trace impurities cause a material to produce ex-
trinsic luminescence. Transition metal ions, with an electron conguration of 3d
(19)
, can
interact strongly with the crystal eld. This results in changes in the energy level structure
of the free ion. Usually, the luminescence activated by transition metal ions is char-
acterized by a broad peak width and a peak position strongly dependent on the host
matrix. The centroid and width of the peak are sensitive to the chemical surroundings in
the crystal and may be used to distinguish the host matrices involved. For example, with
Copyright 2002 Marcel Dekker, Inc.
Mn
2
(3d
5
) ions located in a crystal with a strong crystal eld, the luminescence produced
tends to be of longer wavelength. The luminescence of Mn
2
-activated calcite (Mn
2
ions
in octahedral coordination), for example, is orange, whereas the luminescence of Mn
2
-
activated willemite (Mn
2
ions in tetrahedral coordination) is greenish.
Trivalent rare-earth element ions and most divalent rare-earth element ions, except
for those of Ce, Gd and Lu, have the electron conguration 4f
(k)
[5s
2
5p
6
]. These special
electron congurations give rise to characteristic narrow luminescence emission bands for
rare earth element (REE) ions with the conguration 4f
(k1)
f
1
* through a f
! f tran-
sition. The narrow luminescence emission band of the REE ion (trivalent or divalent)
originates in the 4f subshell, which is partially shielded by [5s
2
5p
6
] electrons. Therefore, for
REE ions with a 4f
k1
f
1
conguration, the structure of the energy levels in the free REE
ions is basically the same in dierent host matrices. Figure 22 shows an IL spectrum from
a natural zircon grain, where three peaks from REE are superimposed on a broad intrinsic
luminescence band.
Experimentally, the IL light is detected by using mirrors or lenses and either directly
counting the photons in a PM tube or analyzing them in a spectrometer. The detailed
spectrum (see Fig. 22) provides information on the crystal structure and can be used to
determine the REE. A scanning grating spectrometer is a very slow system. Variation in
beam current, beam-induced lattice damage, and so forth can adversely aect the acquired
spectrum. The situation can be improved by implementing a photodiode-array detector
with, for instance, 1024 detector elements, each with a size of a few tens of micrometers.
Such a system allows the simultaneous recording of all wavelengths and can signicantly
improve the analytical capacity, speed up the analysis, and reduce beam damage eects.
Furthermore, because ion-beam bombardment of a solid material modies its properties
and changes its luminescence characteristics, the rapid spectrum acquisition makes it
possible to perform a detailed study of these transformations.
The most common experimental facility for IL is the nuclear microprobe, where the
imaging capability of IL is an invaluable asset, especially in studies on geological samples.
Figure 22 Ionoluminescence spectrum from a natural zircon with peaks, emanating from various
rare earth elements, superimposed on a broad yellow intrinsic luminescence band.
Copyright 2002 Marcel Dekker, Inc.
The very high cross sections for IL processes make it possible to employ extremely low
beam doses. Synergistic eects of the simultaneous use of PIXE and IL enhance the
capability of both techniques (Malmqvist et al., 1996). Because of L-line interferences,
PIXE is not very suitable for trace determination of the REE which are of great im-
portance in geoscience. The elemental information obtained by IL (on the REE) and by
PIXE (on many other elements) complement each other and the PIXE data can facilitate
the interpretation of the IL spectrum, so that both techniques enhance each other.
VIII. CONCLUSIONS
Particle-induced x-ray emission is undoubtedly an invaluable and very powerful x-ray
emission spectrometry technique. It has truly multielemental capability, covering a large
part of the periodic system (from Na to U), with detection limits that vary smoothly as a
function of atomic number. The use of x-ray absorbers can improve this Z dependence
further. The sensitivity of PIXE is high and detection limits are low, although they do
depend on the particular material being analyzed. The most favorable situation is the
determination of trace elements in a light-element matrix (Z<11), where detection limits
are of the order of 0.11 mg=g. For samples with appreciable amounts of heavy and
mediumheavy elements, the detection limits are somewhat higher; for example, typically
a few micrograms per gram for elements between Z20 and Z50 in geological mate-
rials. In absolute terms, the detection limits are even more pronounced, down to 10
12
g in
macro-PIXE and down to 10
17
g in micro-PIXE. Other favorable features of PIXE are
the ability to analyze tiny samples (1 mg or less in macro-PIXE, and much less in micro-
PIXE), the speed of the analysis (110 min bombardment time per specimen), and the
possibility for automation. The degree of specimen preparation needed in PIXE varies
from zero at one extreme (e.g., aerosol samples and archaeological materials) through
modest (e.g., mineralogical samples and materials science) to signicant at the other ex-
treme (e.g., some biomedical specimens). The specimens are normally in solid form, and
the irradiated target material is, in general, not aected by the particle beam at the current
densities needed for conventional macro-PIXE, so that the technique is nondestructive.
Even delicate materials such as paper or parchment are unaected by so-called millibeams,
although in these cases, special precautions such as low beam intensity and=or external
beams are necessary. In micro-PIXE, the larger current densities incident on the specimen
can cause damage and alteration, but, in practice, analyses are often conducted at current
densities that do not cause such problems. However, as beam diameters become smaller
and current densities higher, damage will become more important.
Particle-induced x-ray emission is extremely versatile in terms of the size of sample
that it can accommodate. Depending on the size, dierent experimental arrangements may
be used: a conventional macrobeam in vacuum (with beam size of the order of several
millimeters), an external macrobeam or millibeam, or a genuine microbeam. The variety of
specimens analyzed to date is enormous, ranging from a large painting to a single blood
cell. They have in common, though, that they are normally in solid form. Micro-PIXE as
used in the nuclear microprobe makes it possible to perform analyses with excellent spa-
tial resolution (down to 1 mm or better), high sensitivity (detection limits of the order of
1 mg=g), and good accuracy. The importance of this can hardly be overestimated. For
example, microscopically small grains in minerals or single cells in biological tissue can be
analyzed. If the microbeam is used in the scanning mode, as is often done, elemental maps
can be produced, giving a much more detailed picture of trace element distributions than
measurements made at single points.
Copyright 2002 Marcel Dekker, Inc.
Another favorable feature of PIXE is that it can be complemented with other IBA
techniques, such as RBS and other forms of elastic scattering spectrometry, NRA, and
PIGE. As a result, a simultaneous measurement of the light elements (H, Li, B, C, N, O,
and F), which cannot be measured by PIXE, is feasible. This is an important asset when
analyzing atmospheric aerosol samples, as discussed in detail by Cohen (1993, 1998),
Cahill (1995), and Swietlicki et al. (1996a), but also for various other sample types. Some
of the other IBA techniques, such as RBS and NRA, predate PIXE, and in certain elds of
application (e.g., materials analysis), RBS is the major IBA technique, and PIXE generally
serves only as a complementary technique. Some of the ancillary IBA techniques (e.g., o-
axis STIM and ionoluminescence) have been developed in the last decade as adjuncts to
micro-PIXE, demonstrating the continued development of the IBA eld overall.
Particle-induced x-ray emission combines a particular and unique set of advanta-
geous features, but as any technique, it has also it limitations. Limitations which are also
shared by XRF are that it suers from spectral interferences and matrix eects and that it
does not allow the direct measurement of ultratrace elements that are present at nano-
grams per gram levels. Unlike in some other techniques, however, the matrix eects are
well understood and can be corrected for. Consequently, PIXE can provide results with an
accuracy better than 5% relative. The most serious drawbacks of the technique are that it
requires a MeV particle accelerator and that commercial PIXE apparatus are not readily
available (the great majority of PIXE laboratories have built their own setup, and com-
mercial systems are usually custom-built). Most PIXE laboratories have been set up in
nuclear physics institutes, and although there have been interesting developments during
the past decade, with the installation of dedicated PIXE laboratories and nuclear micro-
probes, the growth in PIXE has remained slow. There are currently more than 100 PIXE
groups worldwide.
Part of the reason for the slow growth of PIXE is, of course, that the technique has
to compete with a wide plethora of other analytical techniques and that there have been
substantial advances in several of these techniques over the past 1015 years, for example,
in ICPMS. It is therefore worthwhile to compare the characteristics and capabilities
of PIXE with those of the other techniques and to assess for what sample types and=or
problems PIXE is most appropriate. Maenhaut (1990b) made such an evaluation for
macro-PIXE 10 years ago, which was retained in the Conclusions section in the chapter on
PIXE in the rst edition of Handbook of X-Ray Spectrometry (Maenhaut and Malmqvist,
1992). Although many of the conclusions still stand, that evaluation is now somewhat
outdated. A more recent evaluation is that of Johansson and Campbell (1995) in the
chapter Comparison with Other Methods: Future Prospects in the textbook Particle-
Induced X-Ray Emission Spectrometry (PIXE). In the present Conclusions section, we have
drawn heavily on that excellent and insightful chapter of Johansson and Campbell (1995).
As indicated by them, any comparison with other techniques is fraught with risk. The
range of techniques available means that it is not an easy matter to select the most ap-
propriate for a particular task. It is often the case that proximity, familiarity, and ease of
access play just as large a role as the matching between task and technique. Furthermore,
because of the rapid progress in analytical techniques, any comparison becomes dated
after a few years. The recent progress in x-ray emission spectrometry techniques can be
judged from the various chapters in this handbook. For continued updates on this, the
biennial fundamental review on x-ray spectrometry in the June issue (of every even year) of
the journal Analytical Chemistry can be recommended [see To ro k et al. (1998) for the most
recent one in the series]. That same June issue of Analytical Chemistry also contains
fundamental reviews on many other analytical techniques. Furthermore, some other
Copyright 2002 Marcel Dekker, Inc.
analytical chemistry journals, such as Journal of Analytical Atomic Spectrometry, also
publish technique-oriented reviews on a regular basis.
A natural starting point in a comparison of PIXE with other analytical techniques is
conventional tube-excited XRF. In actual numbers, the relative detection limits of PIXE
for thin samples and a light-element matrix vary within the range 0.11 mg=g for the ele-
ments of greatest interest (the transition metals and the heavy metals around Pb). With
careful optimization (a-particle excitation, tight geometry, and heavy shielding), detection
limits as low as 0.02 mg=g can even be reached in PIXE (Johansson, 1992). In EDXRF, the
corresponding detection limits are of the order of 110 mg=g. Even more pronounced is the
dierence in detection limits in absolute terms. To obtain optimum relative detection limits
in XRF, the sample weight has to be at least 10100 mg, which means that the absolute
detection limits are of the order of some hundred nanograms. In macro-PIXE, the amount
probed by the proton beam can be 0.1 mg or even lower, and the corresponding absolute
detection limits are of the order of a few picograms. However, as discussed by Johansson
and Campbell (1995), besides sensitivity and detection limits, several other factors have to
be considered. There are certainly many sample types and=or problems where EDXRF is
to be preferred over macro-PIXE as the technique of choice.
One substantial dierence between PIXE and XRF is that with micro-PIXE, the
spatial distribution of trace elements can be studied down to the submicrometer level,
whereas conventional XRF does not oer this possibility. In the mid-1980s, the use of
glass capillaries for the concentration of the primary radiation from an x-ray tube was
introduced into XRF (see Chapter 11). This and subsequent developments have led to
micro-XRF and desktop x-ray microprobes, whereby a spatial resolution down to 10 mm
with reasonable incident x-ray intensity may be obtained. These are important technical
advances, conferring on tabletop XRF some, although not all, of the abilities of micro-
PIXE. Another signicant step forward in XRF was the introduction of TXRF (see
Chapter 9). However, this technique is much more suitable for the analysis of dilute
aqueous solutions than of solid samples. It tends to compete more directly with other
techniques (e.g., optical atomic spectrometry) than with PIXE. There appears, for ex-
ample, to be little potential for high-throughput nondestructive analysis of aerosol par-
ticulate samples. A third far-reaching change in the XRF technique has occurred through
the introduction of synchrotron radiation (SR) as the primary source, which has led to
SRXRF (see Chapter 8). The main limitation of this technique is the same as for PIXE,
but to a much greater extent, namely the availability. In the case of PIXE, it is quite
feasible to set up an accelerator laboratory to be used solely for PIXE and related IBA
work, whereas it is clearly out of the question to use a SR source mainly for elemental
analysis. One is therefore limited to using existing facilities, whose number is small. The
potential of SRXRF undoubtedly lies in its microbeam capability. However, it seems
unlikely, for example, that the continued development and success of SRXRF will aect
the rapid growth of routine micro-PIXE analysis of mineral grains or the growing use of
micro-PIXE and associated IBA techniques in art and archaeometry. Perhaps the main
competition oered to micro-PIXE will be for in situ problems where the energy de-
position in the irradiated specimen is an important issue. Thus, microbeam SRXRF can
complement micro-PIXE by extending in situ microbeam analysis to specimens that
cannot withstand micro-PIXE.
Beyond doubt, the greatest competition for the nuclear microprobe and micro-PIXE
comes from electron probe microanalysis (EPMA) and other electron microscopic tech-
niques (see Chapter 13). Micro-PIXE has the advantage over EPMA that it oers relative
detection limits that are typically two orders of magnitude better, that it is easier to
Copyright 2002 Marcel Dekker, Inc.
quantify, and that it can be complemented with other IBA techniques. EPMA and other
electron microscopic techniques, on the other hand, have as advantages that the in-
strumentation is more compact and is commercially available and that they oer a much
higher speed of analysis, better spatial resolution, and better imaging capabilities. It is
clear that EPMA will be preferred over the nuclear microprobe in many studies. However,
rather than being competitors of each other, EPMA and the nuclear microprobe can very
well complement each other, whereby micro-PIXE is used for problems where its excellent
detection limits and other specic advantages are required or at least highly desirable.
The majority of bulk element analyses are carried out by optical atomic spectrometry
or atomic mass spectrometry, in particular ICPMS. In most of these techniques, the
sample material has to be introduced into the instrument as a liquid, and in ICPMS, a
dilute solution is even preferred. Solid samples must therefore be dissolved, which means
an extra preparation stage with the risk of incomplete dissolution, losses, or contamina-
tion. The dissolution also implies that the analysis is destructive. Although some techni-
ques allow one to analyze powdered materials [e.g., electrothermal atomization atomic
absorption spectrometry (ETAAAS)] or can handle various types of solid samples (e.g.,
laser ablation mass spectrometry), and continued research is done on the introduction of
powders in other techniques, there are overall serious limitations with the analysis of solid
samples by optical atomic spectrometry or atomic mass spectrometry. Such analyses suer
from severe matrix eects and diculties in obtaining accurate quantitative results for a
wide range of elements, and they are invariably destructive. Also, certain techniques, such
as ETAAAS, can only measure one element at a time and require the addition of an
appropriate matrix modier. It is clear that the optical atomic spectrometry and atomic
mass spectrometry techniques dier very much from PIXE. Also, the optical atomic
spectrometry techniques do not oer any equivalent of microbeam capability. Such cap-
ability does exist in some atomic mass spectrometry techniques [e.g., in secondary ion mass
spectrometry (SIMS)] and laser ablation inductively coupled plasmamass spectrometry
[LAICPMS], but their spatial resolution is clearly worse than in micro-PIXE and
accurate quantication is dicult.
One way to answer the question of which sample types or analytical problems are
better handled by PIXE and micro-PIXE than by other techniques is to survey the various
applications. Section VI provides such a survey, from which one can identify certain
applications where it appears dicult to nd alternatives. To a large extent, these are
applications involving solid and particulate sample material and having minimum or
zero sample preparation.
Because of its inherent characteristics, macro-PIXE has been very much applied for
measuring trace elements in various types of biomedical samples. However, for several of
such samples and for ultratrace determinations, the optical atomic spectrometric and ICP
MS techniques are now better suited. Also, in recent years, the emphasis has shifted from
total element determination toward elemental speciation. PIXE and the other x-ray
techniques lend themselves much less to this speciation work than some of the other
techniques. In applying PIXE to biomedical problems, one should therefore look for cases
in which one can fully utilize the special advantages of PIXE, such as spatial resolution,
accurate quantitative analysis, and small samples. Particularly micro-PIXE, as used in the
nuclear microprobe and complemented with other IBA techniques such as STIM, is in-
valuable in biomedical trace element research.
One area of application where macro-PIXE has been and continues to be highly
successful is in the analysis of airborne particulate material (atmospheric aerosols). For the
analysis of very large numbers of small aerosol deposits, as collect by compact samplers
Copyright 2002 Marcel Dekker, Inc.
which provide good time or size resolution, or both, there is virtually no competition to
PIXE. Conventional tube-excited EDXRF cannot be used because its absolute sensitivity
is too low. TXRF has the absolute sensitivity but is not physically appropriate and
SRXRF is much more expensive and not really applicable for routine analysis of large
numbers of samples. The optical atomic spectrometric techniques and ICPMS require
dissolution, which is time-consuming and may be incomplete for certain important matrix
elements. Conventional ICPMS also suers from spectral interferences, causing the
problem that some important elements cannot be measured. PIXE provides at the same
time data for the major elements (S, Na, Cl, Al, Si, Fe), from which the concentrations of
important aerosol types (sulfate, sea salt, crustal material) can be estimated, and for
several anthropogenic and natural minor and trace elements (P, V, Mn, Ni, Cu, Zn, As,
Se, Br, Rb, Pb) that can be used in source type identication and apportionment. Fur-
thermore, by complementing PIXE with other IBA techniques, concentrations can be
obtained for the light elements (H, C, N, O) that make up most of the aerosol mass, and
the hydrogen concentration can then be used to estimate the concentration of the im-
portant organic aerosol type (Cahill, 1995). Considering the current interest in the eects
of ne particles on human health and in the role of aerosols in climate, it is expected that
the analysis of atmospheric aerosol samples will remain a successful application for PIXE
for many years to come. However, as indicated in Sec. VI.B, one should realize that PIXE
provides only part of the desired information. With regard to the chemical composition, it
is highly desirable to also perform measurements for important ionic species (e.g., am-
monium, nitrate) and for organic carbon (OC) and elemental carbon (EC). Also, in order
to arrive at a complete aerosol characterization, the chemical measurements have to be
complemented with physical and optical (radiative) aerosol measurements. PIXE re-
searchers should try to complement their work with these various other measurements or
otherwise cooperate with other groups who are involved in such research, and they should
try to become integrated in larger atmospheric aerosol and atmospheric chemistry research
projects. Fortunately, this is increasingly realized within the PIXE community.
Earth science is another eld in which the use of PIXE or specically micro-PIXE
has increased dramatically during the past decade. Here, however, there are some powerful
competing techniques (e.g., the x-ray spectrometric techniques of micro-XRF and
SRXRF). Furthermore, other physically based analytical techniques continue to develop,
such as SIMS and LAICPMS. SIMS provides spatial resolution of a few micrometers
and relative detection limits that are frequently below 1 mg=g; it is clearly superior to
micro-PIXE for the rare earth elements. LAICPMS routinely oers a spatial resolution
of 2040 mm and detection limits of 0.5 mg=g. These two techniques also oer isotopic
discrimination, which is not an option with PIXE. However, in each of these two tech-
niques, the matrix eects are much more complex and quantication more dicult than in
PIXE. Micro-PIXE still maintains its advantage as a truly multielemental, in situ, non-
destructive technique with detection limits of a few micrograms per gram, 1 mm spatial
resolution, and straighforward matrix corrections based on simple, well-understood
physics. It can be used for surveys involving a large number of samples for ore pro-
specting, and for the detailed study of minerals. In the latter case, complementary use of
EPMA and micro-PIXE has turned out to be very fruitful. EPMA is a standard technique
in mineralogy for the determination of major and minor elements, and micro-PIXE allows
these studies to be extended in seamless fashion to trace elements. In studies of extra-
terrestrial materials, such as micrometeoritics and interstellar particles, the high sensitivity
and nondestructiveness of micro-PIXE are a prerequisite. Similarly, the study of ore body
emanations gives samples of extremely small mass and any other technique than PIXE
Copyright 2002 Marcel Dekker, Inc.
seems to be excluded. Micro-PIXE is ideal for investigations of mineral and melt inclu-
sions in many materials (e.g., diamonds). Meanwhile, in the quite dierent context of
multielement bulk rock analysis with sub-microgram per gram detection limits, INAA,
ICPAES, and ICPMS remain the techniques of choice.
Because PIXE and related IBA techniques allow the analysis of delicate samples
without giving any visible damage and without introducing any radioactivity, they can
also very advantageously be used in studies in art and archaeology or for analyzing other
unique samples of which subsampling is impossible. XRF is often the natural choice for
the nondestructive analysis of for example, archaeological artifacts, but PIXE has other
advantages due to its somewhat better sensitivity and imaging capability. A situation in
which the greater sensitivity is desired is the determination of the trace element prole in
various items, which can help in the characterization and identication of material for
addressing questions such as provenance, manufacturing procedure, and trade routes.
Perhaps, the greatest asset of PIXE in these elds is its versatility. Bulk analysis can be
carried out with macro-PIXE, and microscopic details can be studied with a microbeam.
A very useful arrangement is an external beam with a cross section of 0.11 mm, called a
millibeam. This can be used for studies of details in paintings or the ink of single letters in
books and other documents. However, the developments in micro-XRF and microbeam
SRXRF will give rise to strong competition for PIXE in this type of application.
Finally, PIXE will certainly nd further application in the eld of materials research.
However, as indicated earlier, RBS is the major IBA technique in this eld, and PIXE
generally serves only as a complementary technique. The complementarity of PIXE resides
in the fact that it is able to detect elements that are present at levels that are too low for
RBS or for which RBS does not oer sucient Z discrimination.
REFERENCES
Ahlberg M, Akselsson KR, Brune D, Lorenzen J. Nucl Instrum Methods 123:385, 1975a.
Ahlberg M, Johansson G, Malmqvist K. Nucl Instrum Methods 131:377, 1975b.
Ahmed M, Faiz M, Al-Ohali MA, Fageeha O. Nucl Instrum Methods B132:507, 1997.
Aldape F, Flores J, Garcia R, Nelson JW. Nucl Instrum Methods B109=110:502, 1996.
Ali A, Borbe ly-Kiss I, Szabo G, Koltay E, Medve F, Bacso J. Int J PIXE 4:9, 1994.
Allan GL, Camakaris J, Legge GJF. Biol Trace Element Res 40:103, 1994.
Amemiya S, Masuda T, Popa-Simil L, Mateescu L. Nucl Instrum Methods B118:403, 1996.
Amsel G, Heitz Ch, Menu M. Nucl Instrum Methods B14:30, 1986.
Andersen HH, Ziegler JF. The Stopping and Ranges of Ions in Matter, Vol. 3, Hydrogen: Stopping
Powers and Ranges in All Elements. New York: Pergamon Press, 1977.
Andrade F, Orsini C, Maenhaut W. Atmos Environ 28:2307, 1994.
Andreae MO, Crutzen PJ. Science 276:1052, 1997.
Annegarn HJ, Flanz M, Kenntner T, Kneen MA, Helas G, Piketh SJ. Nucl Instrum Methods
B109=110:548, 1996.
Aoki T, Katayama Y, Kagawa A, Koh S, Yoshida K. Nucl Instrum Methods B136138:919, 1998.
Arai N, Sakamoto W. Int J PIXE 3:275, 1993.
Arai N, Sakamoto W, Maeda K. Int J PIXE 4:123, 1994.
Arai N, Takai N, Sakamoto W, Yoshida K, Maeda K. Int J PIXE 5:153, 1995.
Araujo F, Pinheiro T, Alves LC, Valerio P, Gaspar F, Alves J. Nucl Instrum Methods B136
138:1005, 1998.
Artaxo P, Hansson H-C. Atmos Environ 29:393, 1995.
Artaxo P, Gerab F, Yamasoe MA, Martins JV. J Geophys Res 99:22857, 1994.
Artaxo P, Rabello MLC, Maenhaut W, Van Grieken R. Tellus 44B:318, 1992.
Copyright 2002 Marcel Dekker, Inc.
Asking L, Swietlicki E, Garg LM. Nucl Instrum Methods B22:368, 1987.
Bao WL, Orlic I, Tang SM. Int J PIXE 5:235, 1995.
Barrie LA, Bottenheim JW, Schnell RC, Crutzen PJ, Rasmussen RA. Nature 334:138, 1988.
Baumann S, Houmere PD, Nelson JW. Nucl Instrum Methods 181:499, 1981.
Bench G, Saint S, Legge GJF, Cholewa M. Nucl Instrum Methods B77:175, 1992.
Benyaich F, Makhtari A, Torrisi L, Foti G. Nucl Instrum Methods B132:481, 1997.
Bergin MH, Jarezo J-L, Davidson CI, Dibb JE, Pandis SN, Hillamo R, Maenhaut W, Kuhns HD,
Ma kela T. J Geophys Res 100:16,275, 1995.
Berner A. In: Liu BYH, Pui DYH, Fissan HJ, eds. Aerosols. Science, Technology, and Industrial
Applications of Airborne Particles. New York: Elsevier, 1984, p 139.
Bird JR, Williams JS, eds. Ion Beams for Materials Analysis. Sydney: Academic Press, 1989.
Bogdanovic I, Fazinic S, Jaks ic M, Grime G, Valkovic V. Nucl Instrum Methods B95:732, 1994.
Bogdanovic I, Fazinic S, Itskos S, Jaks ic M, Karydas E, Katselis V, Paradellis T, Tadic T,
Valkovic O, Valkovic V. Nucl Instrum Methods B99:402, 1995.
Bombelka E, Richter F-W, Ries H, Wa tjen U. Nucl Instrum Methods B3:296, 1984.
Boni C, Caruso E, Cereda E, Braga Marcazzan GM, Redaelli P. Nucl Instrum Methods B40=41:620,
1989.
Borbe ly-Kiss I, Koltay E, Szabo G. Nucl Instrum Methods B109=110:445, 1996.
Borbe ly-Kiss I, Koltay E, Laszlo S, Szabo Gy, Zolnai L. Nucl Instrum Methods B12:496, 1985.
Boscher-Barre N, Trocellier P. Nucl Instrum Methods B73:413, 1993.
Boubel RW, Fox DL, Turner DB, Stern AC. Fundamentals of Air Pollution. 3rd ed. San Diego:
Academic Press, 1994.
Braga Marcazzan GM. Nucl Instrum Methods B109=110:429, 1996.
Braga Marcazzan GM, Caruso E, Cereda E, Redaelli P, Bacci P, Ventura A, Lombardo G. Nucl
Instrum Methods B22:305, 1987.
Brandt W, Lapicki G. Phys Rev A20:465, 1979.
Brandt W, Lapicki G. Phys Rev A23:1717, 1981.
Breese MBH, Jamieson DN, King PJC. Materials Analysis Using a Nuclear Microprobe. New York:
Wiley, 1996.
Breese MBH, Cookson JA, Cole JA, Ledbury AE. Nucl Instrum Methods B54:12, 1991.
Breese MBH, Grime GW, Linford W, Harold M. Nucl Instrum Methods B158:48, 1999.
Breese MBH, Romano LT, Salter CJ, Grime GW, Watt F. J Mater Res 7(9):2373, 1992.
Brune D, Hellborg R, Whitlow HJ, Hunderi O, eds. Surface Characterization. Weinheim: Wiley
VCH, 1997.
Buoso MC, Fazinic S, Haque AMI, Moschini G, Volpe A, Caravello GU. Nucl Instrum Methods
B68:269, 1992.
Cabri LJ, Campbell JL, Laamme JHG, Leigh RG, Maxwell JA, Scott JD. Can Mineral 23:133,
1985.
Cahill TA. In: Johansson SAE, Campbell JL, Malmqvist KG, eds. Particle-Induced X-Ray Emission
Spectrometry (PIXE). Chichester: Wiley, 1995, p 237.
Cahill TA, Morales R, Miranda J. Atmos Environ 30:747, 1996.
Cahill TA, Wilkinson K, Schnell R. J Geophys Res 97:14,513, 1992.
Cahill TA, Matsuda Y, Shadoan D, Eldred RA, Kusko BH. Nucl Instrum Methods B3:263, 1984.
Cahill TA, Schwab RN, Kusko BH, Eldred RA, Mo ller G, Deutschke D, Wick DL, Pooley AS.
Anal Chem 59:829, 1987.
Cahill TA, Asbaugh LL, Barone JB, Eldred R, Feeney PJ, Flocchini RG, Goodart C, Shadoan DJ,
Wolfe G. J Air Pollut Control Assoc 27:675, 1977.
Cambria R, Del Carmine P, Grange M, Lucarelli F, Mando` PA. Nucl Instrum Methods B75:488,
1993.
Campbell JL. Nucl Instrum Methods B31:518, 1988.
Campbell JL, McGhee PL. Nucl Instrum Methods 248:393, 1986.
Campbell JL, Teesdale WJ, Leigh RG. Nucl Instrum Methods B6:551, 1985.
Campbell JL, Teesdale WJ, Wang J-X. Nucl Instrum Methods B50:189, 1990.
Copyright 2002 Marcel Dekker, Inc.
Campbell JL, Czamanske GK, MacDonald L, Teesdale WJ. Nucl Instrum Methods B130:608, 1997.
Campbell JL, Higuchi D, Maxwell JA, Teesdale WJ. Nucl Instrum Methods B77:95, 1993.
Campbell JL, Wang J-X, Maxwell JA, Teesdale WJ. Nucl Instrum Methods B43:539, 1989.
Campbell JL, Maenhaut W, Bombelka E, Clayton E, Malmqvist K, Maxwell JA, Pallon J,
Vandenhaute J. Nucl Instrum Methods B14:204, 1986.
Caridi A, Cereda E, Grime GW, Jaks ic M, Braga Marcazzan GM, Valkovic V, Watt F. Nucl
Instrum Methods B77:524, 1993.
Carlsson L-E. Nucl Instrum Methods B3:206, 1984.
Carlsson L-E, Akselsson KR. Nucl Instrum Methods 181:531, 1981.
Cecchi R, Ghermandi G, Zonta R. Nucl Instrum Methods B109=110:407, 1996.
Cereda E, Braga Marcazzan GM, Grime GW, Mattachini F. Nucl Instrum Methods B109=110:696,
1996.
Cereda E, Braga Marcazzan GM, Pedretti M, Grime GW, Baldacci A. Nucl Instrum Methods
B99:414, 1995a.
Cereda E, Braga Marcazzan GM, Pedretti M, Grime GW, Baldacci A. Atmos Environ 29:2323,
1995b.
Chan YC, Simpson RW, McTainsh GH, Vowles PD, Cohen D, Bailey G. Atmos Environ 31:3773,
1997.
Charlson RJ, Heintzenberg J, eds. Aerosol Forcing of Climate. Chichester: Wiley, 1995.
Chaudri MA, Crawford A. Nucl Instrum Methods 181:31, 1981.
Chen MH, Crasemann B. Atomic Data Nucl Data Tables 33:217, 1985.
Chen MH, Crasemann B. Atomic Data Nucl Data Tables 41:257, 1989.
Chen MH, Crasemann B, Mark H. Phys Rev A24:177, 1981.
Cheng HS, He WQ, Tang JY, Yang FJ, Wang JH. Nucl Instrum Methods B118:377, 1996.
Cholewa M, Legge GJF. Nucl Instrum Methods B40=41:651, 1989.
Chu W-K, Mayer JW, Nicolet M-A. Backscattering Spectrometry. New York: Academic Press,
1978.
Climent-Font A, Swietlicki E, Revuelta A. Nucl Instrum Methods B85:830, 1994.
Climent-Font A, Demortier G, Palacio C, Montero I, Ruvalcaba-Sil JL, Diaz D. Nucl Instrum
Methods B134:229, 1998.
Cohen DD. Nucl Instrum Methods B49:1, 1990.
Cohen DD. Nucl Instrum Methods B79:385, 1993.
Cohen DD. Nucl Instrum Methods B136138:14, 1998.
Cohen DD, Clayton E. Nucl Instrum Methods B22:59, 1987.
Cohen DD, Harrigan M. Atomic Data Nucl Data Tables 33:255, 1985.
Cohen DD, Harrigan M. Atomic Data Nucl Data Tables 34:393, 1986.
Cookson JA, Campbell JL. Nucl Instrum Methods 216:489, 1983.
Cookson JA, Ferguson ATG, Pilling FD. J Radioanal Chem 12:39, 1972.
Coote GE, Vickridge IC. Nucl Instrum Methods B30:393, 1988.
Coote GE, Cutress TW, Suckling GW. Nucl Instrum Methods B130:571, 1997.
Correia A, Artaxo P, Maenhaut W. Ann Glaciol 27:560, 1998.
Crutzen PJ, Andreae MO. Science 250:1669, 1990.
Cruvinel PE, Flocchini RG. Nucl Instrum Methods B75:415, 1993.
Culbertson RJ, ed. Proc. 12th Int. Conf. on Ion Beam Analysis. Nucl Instrum Methods B118 (nos.
14), 1996.
Damman AH, Kars SM, Touret JLR, Riee WC, Kramer JALM, Vis RD, Pintea I. Eur J Mineral
8:1081, 1996.
da Silva MF, Soares JC, Breese M, eds. Proc. 13th Int. Conf. on Ion Beam Analysis. Nucl Instrum
Methods B136138, 1998.
De Bock LA, Van Grieken RE, Camuo D, Grime GW. Environ Sci Technol 30:3314, 1996.
Demortier G. Spectroscopy 4(6):35, 1989.
Demortier G, Morciaux Y. Nucl Instrum Methods B85:112, 1994.
Demortier G, Decroupet D, Mathot S. Nucl Instrum Methods B54:346, 1991.
Copyright 2002 Marcel Dekker, Inc.
Denecke B, Bambynek W, Grosse G, Wa tjen U, Ballaux C. Nucl Instrum Methods B49:152, 1990.
De Reuck J, Hebbrecht G, Maenhaut W. Cerebrovasc Dis 3:94, 1993.
Doyle BL, Walsh DS, Lee SR. Nucl Instrum Methods B54:244, 1991.
Doyle BL, Maggiore CJ, Bench G, eds. Proc. 5th Int. Conf. On Nuclear Microprobe Technology
and Applications. Nucl Instrum Methods B130 (nos. 14), 1997.
Duerden P, Clayton E, Bird JR, Cohen DD. Nucl Instrum Methods B14:50, 1986.
Duou H, Maenhaut W, De Reuck J. Biol Trace Element Res 13:1, 1987.
Duoy H, Maenhaut W, De Reuck J. Neurochem Res 14:1099, 1989.
Durocher JJG, Halden NM, Hawthorne FC, McKee JSC. Nucl Instrum Methods B30:470, 1988.
Dutta RK, Sundarshan M, Bhattacharyya SN, Chakravortty V, Chintalapudi SN. Nucl Instrum
Methods B143:403, 1998.
Dzubay TG, Stevens RK. Environ Sci Technol 9:663, 1975.
Echalar F, Gaudichet A, Cachier H, Artaxo P. Geophys Res Lett 22:3039, 1995.
Ecker KH, Quan Z, Schurig T, Weise HP. Nucl Instrum Methods B118:382, 1996.
Ektessabi AM, Mouhyi J, Louvette P, Sennerby L. Int J PIXE 7:179, 1997.
Eldred RA, Cahill TA. Atmos Environ 28:1009, 1994.
Eldred RA, Cahill TA, Wilkinson LK, Feeney PJ, Chow JC, Malm WC. In: Mathai CV, ed.
Transactions, Visibility and Fine Particles. Pittsburgh, PA: Air and Waste Management
Association, 1990, p 187.
Elfman M, Kristiansson P, Malmqvist KG, Pallon J, Sjo land A, Utui R, Yang C, Nucl Instrum
Methods B130:188, 1997.
Elliot TA, Grime GW, Nucl Instrum Methods B77:537, 1993.
Elsom DM. Atmospheric Pollution. A Global Problem. Oxford: Blackwell, 1992.
Eltayeb MAH, Van Espen PJ, Cafmeyer J, Van Grieken RE, Maenhaut W. Sci Total Environ
120:281, 1992.
Eltayeb MAH, Xhoer CF, Van Espen PJ, Van Grieken RE, Maenhaut W. Atmos Environ 27B:67,
1993.
Feng SL, Ren MQ, Zhong M. Nucl Instrum Methods B109:584, 1996.
Fischer B. Nucl Instrum Methods B30:284, 1988.
Folkmann F. In: Meyer O, Linker G, Ka ppeler F, eds. Ion Beam Surface Layer Analysis. New York:
Plenum, 1976, Vol 2, p 239.
Folkmann F, Borggreen J, Kjeldgaard A. Nucl Instrum Methods 119:117, 1974a.
Folkmann F, Gaarde C, Huus T, Kemp K. Nucl Instrum Methods 116:487, 1974b.
Formenti P, Annegarn HJ, Piketh SJ. Nucl Instrum Methods B136138:948, 1998a.
Formenti P, Van Den Heever DJ, Annegarn HJ. Nucl Instrum Methods B136-138:961, 1998b.
Franc ois F, Cafmeyer J, Gilot C, Maenhaut W. In: Borrell PM, Borrell P, Cvitas T, Seiler W, eds.
Proc. of EUROTRAC Symposium 92. The Hague: SPB Academic Publishing, 1993, p 788.
Gatz DF, Prospero JM. Atmos Environ 30:3789, 1996.
Gaudichet A, Echalar F, Chatenet B, Quiset JP, Malingre G, Cachier H, Buat-Me nard P, Artaxo P,
Maenhaut W. J. Atmos Chem 22:19, 1995.
Gerab F, Artaxo P, Gillett R, Ayers G. Nucl Instrum Methods B136138:955, 1998a.
Gerab F, Artaxo P, Swietlicki E, Pallon J. Nucl Instrum Methods B136138:318, 1998b.
Ghorai SK, Tekyi-Mensah O, Sims JF, Williams JR, Alford WL. Nucl Instrum Methods B79:389,
1993.
Giuntini L, Lucarelli F, Mando` PA, Hopper W, Baker PH. Nucl Instrum Methods B95:389, 1995.
Glass GA, Hasenstein KH, Chang H-T. Nucl Instrum Methods B79:393, 1993.
Go nczi L, Didriksson R, Sundqvist B, Awal MA. Nucl Instrum Methods 203:577, 1982.
Gosser DC, Ohnersorgen MA, Simon AW, Mayer JW. Nucl Instrum Methods B136138:880, 1998.
Grin WL, Slack JF, Ramsden AR, Win TT, Ryan CG. Econ Geol 91:657, 1996.
Grime GW, Watt F. Beam Optics of Quadrupole Probe-Forming Systems. Bristol: Adam Hilger,
1984.
Grime GW, Watt F, eds. Proc. 1st Int. Conf. on Nuclear Microprobe Technology and Applications.
Nucl Instrum Methods B30 (no. 3), 1988.
Copyright 2002 Marcel Dekker, Inc.
Grime GW, Watt F, Duval AR, Menu M. Nucl Instrum Methods B54:353, 1991.
Guerra MF, Gondonneau A, Barrandon JN. Nucl Instrum Methods B136138:875, 1998.
Gyulai J, Pa szti F, Lohner T, Battistig G, eds. Proc. 11th Int. Conf. On Ion Beam Analysis. Nucl
Instrum Methods B85 (nos. 14), 1994.
Halden NM, Babaluk JA, Kristoerson AH, Campbell JL, Teesdale WJ, Maxwell JA, Reist JD.
Nucl Instrum Methods B109=110:592, 1996.
Hamanaka H, Ito G, Ogura J, Yamamoto Y. Int J PIXE 4:29, 1994.
Hansson H-C, Ekholm A-K, Ross HB. Environ Sci Technol 22:527, 1988.
Hansson H-C, Swietlicki E, Larsson NP-O, Johnsen SJ. Nucl Instrum Methods B75:428, 1993.
Harju L, Lill JO, Saarela KE, Heselius SJ, Hernberg FJ, Lindroos A. Nucl Instrum Methods
B109=110:536, 1996.
Harrison RM, Smith DJT, Pio CA, Castro LM. Atmos Environ 31:3309, 1997.
Havranek V, Maenhaut W, Ducastel G, Hanssen JE. Nucl Instrum Methods B109=110:465, 1996.
Hebbrecht G, Maenhaut W, De Reuck J. In: Bra tter P, Ribas-Ozonas B, Schramel P, eds. Trace
Element Analytical Chemistry in Medicine and Biology. Madrid: Consejo Superior de
Investigaciones Cienticas, 1994a, Vol 6, p 191.
Hebbrecht G, Maenhaut W, De Reuck J. Cerebrovasc Dis 4:142, 1994b.
Heck D, Rokita E. Nucl Instrum Methods B3:259, 1984.
Henderson B, Imbusch GF. Optical Spectroscopy of Inorganic Solids. Oxford: Clarendon Press,
1989.
Hickmott DD, Herrin JM, Abell R, George M, Stimac J, Gauerke ER, Denniston RF. Nucl Instrum
Methods B130:582, 1997.
Hillamo RE, Kerminen V-M, Maenhaut W, Jarezo J-L, Balachandran S, Davidson CI. Atmos
Environ 27A:2787, 1993.
Hinds WC. Aerosol Technology. Properties, Behavior and Measurement of Airborne Particles. New
York: WileyInterscience, 1982.
Hirokawa T, Hu J-Y, E-Guchi S, Nishiyama F, Kiso Y. J Chromatogr 538:413, 1991.
Hopke PK, Xie Y, Raunemaa T, Biegalski S, Landsberger S, Maenhaut W, Artaxo P, Cohen D.
Aerosol Sci Technol 27:726, 1997.
Horowitz P, Grodzins L. Science 189:795, 1975.
Hu J-Y, Hirokawa T, Nishiyama F, Kiso Y. J Chromatogr 589:339, 1992.
Hubbell JH. National Institute of Standards and Technology Report NISTIR 89-4144. US
Department of Commerce, Gaithersburg, MD, 1989.
Hyvonen-Dabek M, Ra isa nen J, Dabek JT. J Radioanal Chem 63:163, 1982.
Ikeda Y, Arai N, Sakamoto W, Murayama T, Maeda K, Yoshida K. Int J PIXE 5:159, 1995.
Inamura K, Martins E, Themner K, Malmqvist KG, Terashi A, Sieso BK. In: Hartmann A,
Kuschinsky W, eds. Cerebral Ischemia and Calcium. Berlin: Springer-Verlag, 1989, p 440.
Injuk J, Otten P, Laane R, Maenhaut W, Van Grieken R. Atmos Environ 26:2499, 1992.
Ishii A, Nakamura K. Nucl Instrum Methods B75:388, 1993.
Ishii K, Morita S. Phys Rev A30:2278, 1984.
Ishii K, Morita S. Nucl Instrum Methods B22:68, 1987.
Ishii K, Morita S. Nucl Instrum Methods B34:209, 1988.
Ishii K, Morita S. Int J PIXE 1:1, 1990.
Ishii K, Orihara H, Morita S, eds. Proc. Int. Symp. on Bio-PIXE. Int J PIXE 2 (nos. 34), 1992.
Iyengar GV. Elemental Analysis of Biological Systems, Volume I, Biomedical, Environmental,
Compositional and Methodological Aspects of Trace Elements. Boca Raton, FL: CRC Press,
1989.
Iyengar GV, Kollmer WE, Bowen HJM. The Elemental Composition of Human Tissues and Body
Fluids. Weinheim: Verlag Chemie, 1978.
Jacobs RW, Duong T, Jones RE, Trapp GA, Scheibel AB. Can J Neurol Sci 16:498, 1989.
Jarezo J-L, Hillamo RE, Davidson CI, Maenhaut W. Atmos Environ 27A:2803, 1993.
Jagielski J, Viguier H, Frontier JP, Trouslard P, Kopcewicz B, Kopcewicz M, Thome L. Nucl
Instrum Methods B118:388, 1996.
Copyright 2002 Marcel Dekker, Inc.
Jaks ic M. Bogdanovic I, Cereda E, Fazinic S, Valkovic V. Nucl Instrum Methods B77:505, 1993.
Jallot E, Oudadesse H, Kilian L, Benhayoume H, Balossier G, Bonhomme P. Electron Microscopy
1998, Proceedings 14th International Congress, 1998, Vol 3, p 215.
Jamieson DN. Nucl Instrum Methods B130:706, 1997.
Jamieson DN, Legge GJF. Nucl Instrum Methods B34:411, 1988.
Jex DG, Hill MW, Mangelson NF. Nucl Instrum Methods B49:141, 1990.
Johansson E-M, Johansson SAE. Nucl Instrum Methods B3:154, 1984.
Johansson GI. X-ray Spectrom 11:194, 1982.
Johansson GI, Pallon J, Malmqvist KG, Akselsson KR. Nucl Instrum Methods 181:81, 1981.
Johansson SAE, ed. Proc. Int. Conf. on Particle Induced X-ray Emission and its Analytical
Applications. Nucl Instrum Methods 142 (nos. 12), 1977.
Johansson SAE, ed. Proc. 2nd Int. Conf. on Particle Induced X-ray Emission and its Analytical
Applications. Nucl Instrum Methods 181 (nos. 13), 1981.
Johansson SAE. Int J PIXE 2:33, 1992.
Johansson SAE, Campbell JL. PIXE: A Novel Technique for Elemental Analysis. Chichester: Wiley,
1988.
Johansson SAE, Campbell JL. In: Johansson SAE, Campbell JL, Malmqvist KG, eds. Particle-
Induced X-Ray Emission Spectrometry (PIXE). Chichester: Wiley, 1995, p 419.
Johansson SAE, Johansson TB. Nucl Instrum Methods 37:473, 1976.
Johansson SAE, Campbell JL, Malmqvist KG, eds. Particle-Induced X-Ray Emission Spectrometry
(PIXE). Chichester: Wiley, 1995.
Johansson TB, Akselsson KR, Johansson SAE. Nucl Instrum Methods 84:141, 1970.
Kakuta I, Takasaki M, Sato T, Ishii K. Int J PIXE 5:221, 1995.
Kallithrakas-Kontos N, Katsanos AA, Aravantinos A, Oeconomidou M, Touratsoglou J.
Archaeometry 35:265, 1993.
Kallithrakas-Kontos N, Katsanos AA, Potiriadis C, Oeconomidou M, Touratsoglou J. Nucl
Instrum Methods B109=110:662, 1996.
Kasahara M, Gotoh N. Int J PIXE 5:123, 1995.
Kasahara M, Ogiwara H, Yamamoto K. Nucl Instrum Methods B118:400, 1996.
Kawatsura K, Nakae T, Takahashi R, Nakai Y, Arai S, Aoki Y, Goppelt-Langer P, Yamamoto S,
Takeshita H, Naramoto H, Horino Y, Mokuno Y, Fujii K, Mitamura T, Terasawa M, Uchida
H, Koterazawa K. Nucl Instrum Methods B118:363, 1996.
Kemp K. In: Dzubay TG, ed. X-ray Fluorescence Analysis of Environmental Samples. Ann Arbor,
MI: Ann Arbor Science, 1977, p 203.
Kemp K. Atmos Environ 27A:823, 1993.
Kemp K. Nucl Instrum Methods B109=110:530, 1996.
Kennedy VJ, Augusthy A, Varier KM, Magudapathy P, Panchapakesan S, Ramesh C, Nair KGM,
Vijayan V. Nucl Instrum Methods B134:224, 1998.
Kerminen V-M, Hillamo RE, Ma kela T, Jarezo J-L, Maenhaut W. J Geophys Res 103:5661, 1998.
King PJC, Breese MBH, Booker GR, Whitehurst J, Wilshaw PR, Grime GW, Watt F, Goringe MJ.
Nucl Instrum Methods B77:320, 1993.
Kirby BJ, Legge GJF. Nucl Instrum Methods B54:98, 1991.
Kirby BJ, Dankas DM, Legge GJF, Mercer JFB. J Inorg Biochem 71:189, 1998.
Kneis H, Martin B, Nobiling R, Povh B, Traxel K. Nucl Instrum Methods 197:79, 1982.
Kra mer U, Grime GW, Smith JAC, Hawes CR, Baker AJM. Nucl Instrum Methods B130:346,
1997.
Krause MO. J Phys Chem Ref Data 8:307, 1979.
Kristiansson P, Martinsson BG. Nucl Instrum Methods B132:159, 1997.
Kristiansson P, Al-Suhaili S, Elfman M, Malmqvist KG, Pallon J, Sjo land KA. Nucl Instrum
Methods B136138:362, 1998.
Kristiansson P, Nilsson CM, Busk H, Malmqvist L, Elfman M, Malmqvist KG, Pallon J, Sjo land
KA, Utui RJ, Yang C. Nucl Instrum Methods B130:303, 1997.
Kupila-Rantala T, Hyvo nen-Dabek M, Dabek JT. Nucl Instrum Methods B117:320, 1996.
Copyright 2002 Marcel Dekker, Inc.
Kuri G, Rout B, Sundaravel B, Mahapatra DP, Dev BN. Nucl Instrum Methods B119:403, 1996.
Kusko BH, Cahill TA, Eldred RA, Schwab RN. Nucl Instrum Methods B3:689, 1984.
Laj P, Ghermandi G, Cecchi R, Ceccato D. Nucl Instrum Methods B109=110:252, 1996.
Landsberger JP, McDonald B, Watt F. Nature 360:65, 1992.
Lane DW, Duy CA. Nucl Instrum Methods B118:392, 1996.
Legge GJF. Nucl Instrum Methods B118:9, 1997.
Legge GJF, Jamieson DN, eds. Proc. 2nd Int. Conf. on Nuclear Microprobe Technology and
Applications. Nucl Instrum Methods B54 (nos. 13), 1991.
Legge GJF, Mazzolini AP. Nucl Instrum Methods 168:563, 1980.
Legge GJF, Jamieson DN, OBrien PMJ, Mazzolini AP. Nucl Instrum Methods 197:85, 1982.
Lepy MC, De Sanoit J, Alvarez R. Appl Radiat Isotopes 43:847, 1992.
Limic N, Valkovic V. Nucl Instrum Methods B109=110:415, 1996.
Lin EK, Wang CW, Teng PK, Huang YM, Chen CY. Nucl Instrum Methods B68:281, 1992.
Lind T, Kauppinen EI, Jokiniemi JK, Maenhaut W. In: Twenty-fth Symposium (International) on
Combustion. Pittsburgh, PA: The Combustion Institute, 1994, p 201.
Lind T, Kauppinen EI, Maenhaut W, Shah A, Huggins F. Aerosol Sci Technol 24:135, 1996.
Lind T, Valmari T, Kauppinen EI, Sris G, Nilsson K, Maenhaut W. Environ Sci Technol 33:496,
1999.
Lindh U, ed. Proc. 3rd Int. Conf. on Nuclear Microprobe Technology and Applications. Nucl
Instrum Methods B77 (nos. 14), 1993.
Lipinski MR, Przybylowicz WJ, Durholtz MD, Mesjasz-Przybylowicz J. Nucl Instrum Methods
B130:374, 1997.
Liu X, Gao N, Hopke PK, Cohen D, Bailey G, Crisp P. Atmos Environ 30:9, 1996.
Lodge JP Jr, ed. Methods of Air Sampling and Analysis. 3rd ed. Chelsea, MI: Lewis Publishers,
1991.
Loo BW, Jaklevic JM, Goulding FS. In: Liu BYH, ed. Fine Particles. New York: Academic Press,
1976, p 311.
Lo vestam NEG, Swietlicki E. Nucl Instrum Methods B43:104, 1989.
Lo vestam NEG, Swietlicki E. Nucl Instrum Methods B45:307, 1999.
Lucarelli F, Mando PA. Nucl Instrum Methods B109=110:644, 1996.
Luis-Simon FJ, Gomez-Camacho J, Respaldiza MA. Int J PIXE 5:43, 1995.
Ma X-P, MacArthur JD, Roeder PL, Mariano AN. Nucl Instrum Methods B75:423, 1993.
Maenhaut W. Anal Chim Acta 195:125, 1987.
Maenhaut W. In: Pacyna JM, Ottar B, eds. Control and Fate of Atmospheric Trace Metals.
Dordrecht: Kluwer Academic Publishers, 1989, p 259.
Maenhaut W. Scanning Microsc 4:43, 1990a.
Maenhaut W. Nucl Instrum Methods B49:518, 1990b.
Maenhaut W. Int J PIXE 2:609, 1992.
Maenhaut W, Cafmeyer J. X-ray Spectrom 27:236, 1998.
Maenhaut W, Malmqvist KG. In: Van Grieken RE, Markowicz AA, eds. Handbook of X-Ray
Spectrometry. Methods and Techniques. New York: Marcel Dekker, 1992, p 517.
Maenhaut W, Raemdonck H. Nucl Instrum Methods B1:123, 1984.
Maenhaut W, De Reu L, Vandenhaute J. Nucl Instrum Methods B3:135, 1984.
Maenhaut W, Franc ois F, Cafmeyer J. In: Applied Research on Air Pollution Using Nuclear-
Related Analytical Techniques. NAHRES-19. Vienna: IAEA, 1994a, p 249.
Maenhaut W, Hebbrecht G, De Reuck J. Nucl Instrum Methods B75:180, 1993a.
Maenhaut W, Koppen G, Artaxo P. In: Levine JS, ed. Biomass Burning and Global Change.
Cambridge, MA: MIT Press, 1996a, Vol 2, p 637.
Maenhaut W, Vandenhaute J, Duou H. Fresenius Z Anal Chem 326:736, 1987.
Maenhaut W, Ducastel G, Beyaert K, Hanssen JE. In: Borrell PM, Borrell P, Cvitas T, Seiler W, eds.
Proc. of EUROTRAC Symposium 94. The Hague: SPB Academic Publishing, 1994b, p 467.
Maenhaut W, Ducastel G, Hillamo RE, Pakkanen TA. Nucl Instrum Methods B75:249, 1993b.
Maenhaut W, Salomonovic R, Ptasinski J, Grime GW. Nucl Instrum Methods B130:576, 1997a.
Copyright 2002 Marcel Dekker, Inc.
Maenhaut W, Ducastel G, Leck C, Nilsson ED, Heintzenberg J. Tellus 48B:300, 1996b.
Maenhaut W, Franc ois F, Cafmeyer J, Gilot C, Hanssen JE. In: Borrell PM, Borrell P, Kelly K,
Cvitas T, Seiler W, eds. Proc. of EUROTRAC Symposium 96. Southampton UK: Computa-
tional Mechanics Publications, 1997b, Vol 1, p 277.
Maenhaut W, Ryset O, Vadset M, Kauppinen EI, Lind TM. Nucl Instrum Methods B75:266,
1993c.
Maenhaut W, Salma I, Cafmeyer J, Annegarn HJ, Andreae MO. J Geophys Res 101:23,631, 1996c.
Maenhaut W, Beyaert K, Ducastel G, Havra nek V, Salomonovic R, Hanssen JE. In: Borrell PM,
Borrell P, Kelly K, Cvitas T, Seiler W, eds. Proc. of EUROTRAC Symposium 96.
Southampton, UK: Computational Mechanics Publications, 1997c, Vol 1, p 273.
Maenhaut W, Hillamo R, Ma kela T, Jarezo JL, Bergin MH, Davidson CI, Nucl Instrum Methods
B109=110:482, 1996d.
Maetz M, Orberger B, Traxel K, Wathankul P, Pitragool S. Nucl Instrum Methods B130:666, 1997.
Makjanic J, McDonald B, Watt F. Nucl Instrum Methods B130:439, 1997.
Malm WC, Pitchford ML. Atmos Environ 31:1315, 1997.
Malm WC, Molenar JV, Eldred RA, Sisler JF. J Geophys Res 101:19,251, 1996.
Malm WC, Sisler JF, Human D, Eldred RA, Cahill TA. J Geophys Res 99:1347, 1994a.
Malm WC, Gebhart KA, Molenar J, Cahill T, Eldred R, Human D. Atmos Environ 28:347,
1994b.
Malmqvist KG. Nucl Instrum Methods B104:138, 1995a.
Malmqvist KG. In: Johansson SAE, Campbell JL, Malmqvist KG, eds. Particle-Induced X-Ray
Emission Spectrometry (PIXE). Chichester: Wiley, 1995b, p 167.
Malmqvist KG. Nucl Instrum Methods B113:336, 1996.
Malmqvist KG, ed. Proc. 8th Int. Conf. on Particle Induced X-ray Emission and its Analytical
Applications. Nucl Instrum Methods B150 (nos. 14), 1999.
Malmqvist KG, Karlsson E, Akselsson KR. Nucl Instrum Methods 192:523, 1982.
Malmqvist KG, Elfman M, Remond G, Yang C. Nucl Instrum Methods B109=110:227, 1996.
Mangelson NF, Hill MW. Scanning Microsc 4:63, 1990.
Mangelson NF, Eatough DJ, Eatough NL, Hansen LD, Hill MW, Lee ML, Phillips LR, Post ME,
Richter BE, Ryder JF. IEEE Trans Nucl Sci NS-28:1378, 1981.
Marple VA, Rubow KL, Behm SM. Aerosol Sci Technol 14:434, 1991.
Martin B, ed. Proc. 3rd Int. Conf. on Particle Induced X-ray Emission and its Analytical
Applications. Nucl Instrum Methods B3 (nos. 13), 1984.
Mart n JE, Garc a-Tenorio R, Respaldiza MA, Ager FJ, da Silva MF. Nucl Instrum Methods
B139:175, 1998a.
Mart n JE, Garc a-Tenorio R, Respaldiza MA, Bol var JP, da Silva MF. Nucl Instrum Methods
B136138:1000, 1998b.
Mart n JE, Respaldiza MA, Garc a-Tenorio R, Bol var JP, Go mez-Camacho J, da Silva MF. Nucl
Instrum Methods B109=110:506, 1996.
Martinsson BG. Nucl Instrum Methods B15:636, 1986.
Martinson BG. Nucl Instrum Methods B22:356, 1987.
Martinson BG, Hansson H-C. Nucl Instrum Methods B34:203, 1988.
Martinsson BG, Kristiansson P, Nucl Instrum Methods B82:589, 1993.
Mauro T, Bench G, Sidderas-Haddad E, Feingold K, Elias P, Cullander C. J Invest Dermatol
111:1198, 1998.
Maxwell JA, Campbell JL, Teesdale WJ. Nucl Instrum Methods B43:218, 1989.
Maxwell JA, Teesdale WJ, Campbell JL. Nucl Instrum Methods B95:407, 1995.
McKee JSC, Smith GR, Yeo YH, Abdul-Retha K, Gallop D, Durocher JJG, Mulholland W, Smith
CA. Nucl Instrum Methods B40=B41:680, 1989.
Mentes B, Elfman M, Martinsson BG. Nucl Instrum Methods B109=110:511, 1996.
Menu M, Calligaro T, Salomon J, Amsel G, Moulin J. Nucl Instrum Methods B45:610, 1990.
Mingay DW. X-ray Spectrom 12:52, 1983.
Mingay DW, Barnard E. Nucl Instrum Methods 157:537, 1978.
Copyright 2002 Marcel Dekker, Inc.
Miono S, Nakayama Y, Shoji M, Tsuji H, Nakanishi A. Nucl Instrum Methods B75:435, 1993.
Miono S, Zheng CZ, Nakayama Y, Manabe S, Ohmichi K. Nucl Instrum Methods B109=110:677,
1996.
Miranda J. Nucl Instrum Methods B109=110:439, 1996.
Miranda J, Cahill TA, Morales JR, Aldape F, Flores M. J, Di
andrik R, Kobzev AP, Shirokov DM, Kliment V. Nucl Instrum Methods B75:392, 1993.
S
andrik R, Jergel M, S
CG, Ramboz C, Ai
all j
C
j
C
i
since
all j
C
j
1 2
where C refers to the elemental concentrations, expressed as weight fractions, and the
subscript j refers to all of the elements in the sample.
Equation (2) only applies to a system in which the sample and the standard are
identical and have been measured under identical experimental conditions. In practice,
as the similarities between the sample and the standard decrease, Eq. (2), even as an ap-
proximation, fails and a series of corrections must be applied to the k ratio to obtain an
Figure 7 Relative dimensions of the primary electron beam, the electron excitation and emission
volumes, and the x-ray emission volumes in a Cu target. (From Goldstein et al., 1981a. Reproduced
with permission of Plenum Press.)
Copyright 2002 Marcel Dekker, Inc.
accurate quantitative analysis. The corrections that must be applied to the k ratio include
the following:
1. The atomic number correction for the dierence between the electron scattering
and penetration in the sample and the standard
2. The absorption correction for the dierence in the absorption of the x-rays as
they pass through the sample or standard before reaching the detector
3. The uorescence correction for the uorescence of x-rays induced by
the absorption of characteristic and continuum x-rays that are generated in
the sample by the primary electron beam and which propagate out from this
source
These corrections are applied to the various k ratios as part of theoretical or empirical
correction procedures in order to obtain quantitative results.
A. Analytical Corrections for Quantitative Electron-Probe Microanalysis
1. ZAFCorrections
In the rst correction procedures that were used to obtain quantitative elemental analysis
from electron-probe data, the various corrections were applied to the k ratios as separate
multiplicative terms:
C
i
k
Z
k
A
k
F
k
i
3
where the terms k
Z
, k
A
, and k
F
refer to the atomic number, absorption, and uorescence
corrections, respectively. The methods based on Eq. (3) were appropriately referred to as
ZAF methods. In the remainder of this section, each of the corrections will be discussed
separately.
a. Atomic Number Correction, k
Z
The atomic-number correction in electron microprobe analysis is applied to the k ratio to
compensate for the dierence in the electron retardation and electron backscattering be-
tween the sample and standard. Both the electron retardation and backscattering are
dependent on the average atomic number of the sample
"
Z, Eq. (4). Therefore, any dif-
ference between the average atomic number of the sample and the standard should be
addressed by this correction. As a general rule, ignoring the eects of the atomic-number
correction will result in an underestimation of the concentration of high-Z elements in
low-Z matrices and the overestimation of concentrations of low-Z elements in high-Z
matrices (Goldstein et al., 1981b). The average atomic number
"
Z for the sample is given by
"
Z
i
C
i
Z
i
4
The general formulation of k
Z
for element i is given in Eq. (5):
k
Z
i
R
sam
i
R
std
i
_
E
0
E
c
Q
i
S
sam
i
_ _
dE
_
E
0
E
c
Q
i
S
std
i
_ _
dE
5
where the R and the S terms refer to the electron backscattering and the electron stopping
power, respectively, and Q is the ionization cross section. The limits on the integral are
from the incident electron energy, E
0
, to the critical excitation energy, E
c
for the x-ray line
of interest.
Copyright 2002 Marcel Dekker, Inc.
Electron stopping power. The electron stopping power S is dened in Eq. (6) as the
energy lost per unit electron path length in material of density r (Widdington, 1912):
S
1
r
_ _
dE
dX
6
One of the most commonly used terms for S is the approximation by Bethe (1930) and
Bethe and Ashkin (1953), which assumes a continuous function for the electron energy
loss:
S 78;500
Z
i
A
i
_ _
1
E
ln
1:166E
J
i
_ _
7
where S is in keVcm
2
=g and E is in keV.
The value for the mean ionization potential, J, in Eq. (7) is not directly measured and
several dierent expressions have been used in the literature for the calculation of J.
Various literature values for J are listed in Table 1 and are plotted as a function of Z in
Figure 8. The various expressions for J all yield similar results for elements above Z10
with relatively large discrepancies among expressions for Z<10. The BergerSeltzer ex-
pression for J is one of the most widely used in quantitative analysis procedures (Berger
and Seltzer 1964).
There are also several models in the literature that have been used to calculate Q,
all of which have the general form described by Bethe (1930):
Q C
ln U
UE
2
c
8
where U is the overvoltage dened as E
0
=E
c
.
Heinrich and Yakowitz (1970), however, have shown that the dierence in the
various models resulted, in negligible changes in the nal elemental concentrations for
elements with Z > 10.
Duncumb and Reed (1968) simplied the integration in Eq. (5), eliminating the need
for the numerical integration Eq. (5) and the evaluation of Q. In their procedure, they
assumed that the values for (1=S) dE were constant for the sample and standard over the
electron range used in electron-probe analysis and, therefore, could be removed from the
integral. As a result of this assumption, the integration of Q is unnecessary because it
appears in both the numerator and denominator and is the same for the unknown and the
standard. The stopping power factor can then be expressed as
Table1 Different Values for J
Equation (in eV) Reference
J=Z 13 Bloch (1993)
J=Z 11:5 Wilson (1941)
J=Z 9:76 58:82Z
1:19
Berger and Seltzer (1964)
J=Z 9:01 Z
0:67
0:03Z Springer (1967)
J=Z 12:4 0:02Z Heinrich and Yakowitz (1970)
J=Z 14:01 e
0:1Z
75:5=Z
0:13Z
Z=100 Z Duncumb et al. (1969)
J=Z 10:04 8:25e
Z=11:22
Zeller (1975)
Source: From Heinrich, 1981b.
Copyright 2002 Marcel Dekker, Inc.
S
Z
A
_ _
ln
1:166E
J
_ _
9
where A is atomic weight, and J, from Duncumb et al. (1969), equals
J
Z
141 exp0:1Z
75:5
Z
Z=7:5
Z100 Z 10
Equation (9) is one of the most commonly used formulations for classical ZAF micro-
analysis. Duncumb and Reed (1968) have shown from experimental work that the stop-
ping power for a multielement specimen can be expressed as a weighted sum of the
stopping power factor for each element:
S
i
j
0
C
j
S
ij
11
Figure 8 Different formulations for the mean ionization potential, J, plotted as a function of the
atomic number of the target. (From Goldstein, 1981c. Reproduced with permission of Plenum Press.)
Copyright 2002 Marcel Dekker, Inc.
where S
i
is the stopping power for element i; S
ij
is the stopping power for element i in the
presence of element j, and C
j
is the weight fraction of element j.
In addition to Eq. (9), alternate formulations for S have been proposed by Philibert
and Tixier (1968), Pouchou and Pichoir (1991a), and Love et al. (1978). Philibert and
Tixier derived an exact analytical solution to Eq. (6) that denes S in terms of a loga-
rithmic integral:
1
S
1
m
U
0
1
ln w
w
hLiwU
0
Liwi
_ _
12
where
w
1:1666E
c
"
J
m
alli
C
i
Z
i
A
i
ln
"
J
alli
C
i
Z
i
=A
i
ln J
i
m
U
0
E
0
E
c
The logarithmic integral Li of a variable Y expressed as an innite series is
LiY ln j ln Yj
1
F1
ln Y
F
FF!
D 13
where D is Eulers constant, which equals 0.577.
Love et al. (1976) noted that the Bethe expression (Bethe, 1930) for S is valid only if
E ) J and, as a result, modied the Bethe expression . . . to give better limiting behavior
as E approaches J. Their work results in the following formulation for S:
1
S
1 16:05
"
J
E
c
_ _
1=2
U
1=2
0
1
U
0
1
_ _
1:07
_
_
_
_
alli
C
i
Z
i
A
i
_
_
_
_
1
14
Similarly, Pouchou and Pichoir (1991), to avoid problems with the limiting behavior as E
approaches J in the Bethe expression, dened S as
S
1
"
J
_ _
alli
C
i
Z
i
A
i
_ _
_
_
_
_
1
f V
_ _
15
where V E=
"
J,
"
J is same as Philibert and Tixier (1968), and J
i
is from Zeller (Table 1).
f V
k3
k1
D
k
V
p
k
with D
1
6:6 10
6
, D
2
1:12 10
5
1:35--0:45
"
J
2
, D
3
2:2
10
6
=
"
J, p
1
0:78, p
2
0:1, and p
3
0:50:25
"
J .
Electron backscatter factor. The electron backscatter factor Rin Eq. (5) is dened as
R 1
I
b
I
t
_ _
16
where I
b
is the x-ray intensity lost due to backscattered electrons and I
t
is the x-ray in-
tensity if there are no electrons backscattered.
Copyright 2002 Marcel Dekker, Inc.
The fraction of electrons that are backscattered from a sample is known as the
electron backscatter coecient Z, and is given by
Z
_
1
0
dZ
dw
dw 17
where w is the ratio of the energy of the backscattered electron, E
b
, to the beam energy
E
0
i:e:; E
b
=E
0
. An empirical expression for Z, Eq. (18), was obtained by Reuter (1972)
from a t to Heinrichs data (Heinrich 1966):
Z 0:0254 0:016Z 1:86 10
4
Z
2
8:3 10
7
Z
3
18
The number of ionizations generated in a sample by an electron with energy E is given,
for a characteristic X-ray line with critical excitation energy E
c
, by
_
E
c
E
Q
dE=drs
dE 19
I
b
can then be obtained by multiplying Eqs. (17) and (19):
I
b
_
1
w
0
dZ
dw
_
E
c
E
Q
dE
_
drs
dEdw 20
where the integration limit for Z is w
0
E
c
=E
0
because electrons with energies less than E
c
cannot excite the x-rays of interest.
Similarly, I
t
can be calculated from Eq. (21) with integration limits of E
0
to E
c
:
I
t
_
E
c
E
0
Q
dE=drs
dE 21
Finally, substituting Eqs. (20) and (21) into Eq. (16) results in the following formulation
for R:
R 1
_
1
w
0
dZ
dw
_
E
c
E
Q
dE=drs
dEdw
_
E
c
E
0
Q
dE=drs
dE
_
_
_
_
_
_
1
22
Several tabulations of R have been made for pure elements as a function of Z and U
(Duncumb and Reed, 1968); Green, 1963; Springer, 1966. Duncumb and Reed produced a
table of R values for various elements and several dierent overvoltage values. Their
values were determined indirectly from Bishops (1966) measurements of the energy dis-
tributions of backscattered electrons and were in agreement with the direct measurements
of R made by Derian and Castaing (1966). Figure 9 shows a plot of the DuncumbReed R
values versus Z at dierent overvoltages.
Duncumb derived an algebraic expression for R in terms of w
q
and Z from his
calculated values; Eq. (23) is still used in many analytical procedures (Heinrich, 1981c):
R 1 0:581 2:162w
q
5:137w
2
q
9:213w
3
q
8:619w
4
q
2:962w
5
q
10
2
Z
1:609 8:298w
q
28:791w
2
q
47:744w
3
q
46:540w
4
q
17:676w
5
q
10
4
Z
2
5:400 19:184w
q
75:733w
2
q
120:050w
3
q
110:700w
4
q
41:792w
5
q
10
6
Z
3
Copyright 2002 Marcel Dekker, Inc.
5:725 21:645w
q
88:128w
2
q
136:060w
3
q
117:750w
4
q
42:445w
5
q
10
8
Z
4
2:095 8:947w
q
36:510w
2
q
55:694w
3
q
46:079w
4
q
15:851w
5
q
10
10
Z
5
23
where w
q
E
c
=E
0
.
Yakowitz et al. (1973) obtained a simplied expression for R from a t of the
Duncumb and Reed values:
R
ij
R
0
1
R
0
2
lnR
0
3
Z
j
25 24
where
R
0
1
8:73 10
3
U
3
0:1669U
2
0:9662U 0:4523
R
0
2
2:703 10
3
U
3
5:182 10
2
U
2
0:302U 0:1836
R
0
3
0:887U
3
3:44U
2
9:33U 6:43=U
3
In this Eq. (24) i represents the element being measured and j represents the elements in the
specimen, including i. R
ij
is therefore the backscatter correction for element i in the pre-
sence of element j.
Figure 9 Duncumb backscatter correction factor R versus Z and overvoltage (From Heinrich,
1981c. Reproduced with permission of Van Nostrand Reinhold.)
Copyright 2002 Marcel Dekker, Inc.
Myklebust (1984) further simplied the expression for R:
R 1 0:0081512Z 3:613 10
5
Z
2
0:009582Z expU 0:00114E
0
25
This expression represents a t of the R values obtained from the NBS Monte Carlo
program (Myklebust et al., 1975). Figure 10 shows the behavior of various R values from
dierent authors as a function of Z for selected x-ray lines. Myklebust and Newbury
(1991) suggest that the dierent formulations of R produce only small dierences in the
quantitative results. However, they recommend the formulation of R in Eq. (26), which
uses the cross section from Fabre delta Ripelle (1949) with Eq. (22), as the best for-
mulation, including the extreme conditions of very high or very low overvoltages:
R 1
_
1
w
0
dZ
dw
_
wE
0
E
c
A
B
dEdw
_
_
_
_
_
_
_
E
0
E
c
A
B
dE
_
_
_
_
_
_
1
26
where A lnE=E
c
and B 1=E
c
1:32=E ln1:166=J E.
In a multielement system, the factor R for element i can be calculated from Eq. (27),
which was proposed by Duncumb and Reed (1968):
R
i
j
C
j
R
ij
27
Myklebust and Newbury (1988) compared results from Monte Carlo calculations to re-
sults from Eq. (25) for a 10% CuAu alloy to determine the accuracy of mass con-
centration averaging in multielement targets. Their results indicate that Eq. (27) is valid
for the alloy studied. A detailed discussion of the R factor can be found in Myklebust and
Newbury (1991).
Evaluation of k
Z
. The results of analysis for copper in 2 wt% CuAl alloy have
been used by Goldstein et al. (1981d) to demonstrate the magnitude of the atomic-number
correction. The authors used both pure elements and a 46% AlCu alloy as standards for
the analysis. The results, given in Table 2, indicate that the atomic-number correction for
this analysis at 30-keV incident electrons is as high as 16% for the pure-element standards
and is reduced to 8% when the alloy is used as the standard.
b. Absorption Correction, k
A
The primary electron beamgenerates x-rays at varying depths within the sample. As a result,
the x-rays must pass through that portion of the specimen that lies between the x-ray gen-
eration point and the detector before they escape the sample and are measured. As shown
Figure 11, the distance ABis referred to as the absorption path length because a percentage
of the generated x-rays undergo photoelectric absorption, interacting with specimen atoms
prior to escape. The eect of this absorption is an attenuation of the generated x-ray in-
tensity that is measured by the detector. Note that inelastic scattering of the x-rays is not a
signicant process over the path lengths involved. The energy of the unabsorbed char-
acteristic x-ray is not modied during its passage through the specimen. The magnitude of
the attenuation is dependent on the composition of the specimen and a correction must be
considered when the sample and the standards used for the analysis are dissmilar.
Castaing (1951) described the characteristic x-ray intensity (without absorption)
generated in a layer of thickness dz, at a depth z below the specimen surface, in a sample of
density r as
dI frz drz 28
Copyright 2002 Marcel Dekker, Inc.
Figure10 Comparison of different functions for R versus Z for K (a), L (b), and M (c) x-ray lines.
(From Myklebust and Newbury, 1991.)
Copyright 2002 Marcel Dekker, Inc.
where frz is the distribution of characteristic x-rays as a function of depth [density 6
distance (mg=cm
2
)] in the sample. A typical curve for CuKa radiation is shown in
Figure 12. The total generated x-ray intensity for a given line can be obtained by in-
tegrating the area under the curve over the entire x-ray range:
I
_
1
0
frz drz 29
The introduction of x-ray absorption into Eq. (29) results in the following expression for
the x-ray intensity after absorption, I
0
:
I
0
_
1
0
frz expm=rrz cscC drz 30
where rz cscC is the absorption path length for the x-rays in the specimen, C is the
detector emergence angle, and m=r is the mass absorption coecient of the specimen for
the characteristic line of the element of interest.
The absorption term for EPMA is referred to as fw or f
p
, where w m=r cscC
and fw I
0
=I. From Eqs. (29) and (30), fw can be expressed in terms relating to the
specimen as
fw
_
1
0
frz exp
m
r
rz csc C
_ _ _ _
drz
_
_
_
_
_
_
I
1
_
1
0
frz drz 31
Table 2 k
Z
Correction for Copper in a 2 wt% Copper in Aluminum
Alloy
15 keV 30 keV
Standard k
Z
(Cu) k
Z
(Cu)
Elemental Cu 1.16 1.11
Alloy 1.08 1.05
Source: From Goldstein et al., 1981d.
Figure11 X-ray absorption path length.
Copyright 2002 Marcel Dekker, Inc.
The absorption correction for EPMA can then be expressed as
k
A
fw
std
fw
sam
32
The basic formulation of the fw term was derived by Philibert (1963), who considered the
number of ionizations produced in a layer of element i with thickness dz at some depth z
within a specimen:
fw
1
f0hw
4f0hs
1 w=sf1 h=1 hw=sg
33
In this equation, f0 is the surface ionization function, s is Lenards (1895) coecient for
a given incident electron energy, and h is given by
h 1:67 10
6
A
Z
2
sE
2
0
34
Figure12 A f(rz) curves for CuKa radiation. The solid line are measurements by Castaing and Des-
camps and the dashed line is based on Philiberts f (w) with the DuncumbShields s. (From Criss, 1968.)
Copyright 2002 Marcel Dekker, Inc.
Noting that fw was not sensitive to either f0 or s, Philibert (1963) simplied Eq. (34)
by setting f0 0 and h 1:2A=Z
2
. This results in the following expression for fw:
1
fw
1
w
s
_ _
1
h
1 h
w
s
_ _
35
where s 4:5 10
5
=E
1:65
0
E
1:65
c
.
Duncumb and Shields (1996) modied s to take into account that electrons with
energies less than E
c
cannot generate x-rays from the line associated with E
c
. They pro-
posed the following expression for s:
s
2:39 10
5
E
1:5
0
E
1:5
c
36
Heinrich (1970) t experimental values of fw from Green (1962) and proposed the
following formulation for s:
s
4:5 10
5
E
1:65
0
E
1:65
c
37
The most common form of the absorption term found in the various analytical schemes is
Eq. (35) with h 1:2A=Z
2
and Heinrichs s. This form of Eq. (35) is often referred to as
the PhilibertDuncumbHeinrich equation.
Heinrich et al. (1972) empirically derived a simplied absorption term that was based
on experimental data and the Philibert equation. In the formulation of this term, it was
noted that the compositional dependence in Philiberts h term was small compared to the
scatter in the available experimental data. The resulting formulation for fw, given in
Eq. (38), is referred to by Heinrich as the quadratic model. In this equation, the
compositional dependence in the h term has been eliminated, making the model in-
dependent of the atomic weight and atomic number of the target:
1
fw
1 1:2 10
6
gw
_ _
2
38
where g is the quantity E
1:65
0
E
1:65
c
.
In Eq. (35), h is dependent on target composition and must be averaged for the
various elements in multielement targets. A compositionally weighted average for h,
Eq. (39), is generally employed in analytical procedures:
h
t
j
C
j
h
j
39
In addition, the mass absorption coecient, m=r, for the characteristic line of element i
in a multielement target is the weighted sum over all elements in the target including i:
m
r
_ _
sam
i
j
C
j
m
r
_ _
j
i
40
where m=r
j
I
is the mass absorption coecient for the line of element i in element j and C
j
is the concentration of element j.
The calculation of x-ray absorption from Eq. (35) is most accurate when the value of
fw is greater than 0.7. Toward this end, it is important in setting up an experiment to
minimize the degree of x-ray absorption by the following:
Copyright 2002 Marcel Dekker, Inc.
1. Minimizing the absorption path length by selecting a low overvoltage E
0
=E
c
2. Selecting high-energy x-ray lines that have low-absorption cross sections. It is
important to note that selecting a high-energy x-ray line requires a high
acceleration potential and will, therefore, increase the absorption for elements in
the sample that have relatively low-energy analytical lines
3. Measuring the intensities at the highest possible x-ray emergence angle (which
typically is xed for any given instrument)
A study was done by Small et al. (1991) comparing experimental measurements of fw
with four dierent absorption models. The experimental data were collected for Si at both
20 and 30 kV and for Fe and S in FeS
2
at 20 kV acceleration potential. The models studied
included the following:
1. Heinrichs quadratic model, Eq. (38) (Heinrich, 1972)
2. Heinrichs duplex model, Eq. (41) (Heinrich, personal communication, 1987)
3. Philiberts model, Eq. (35) (Philibert, 1963)
4. Pouchou and Pichoirs model, Eq. (42) (Pouchou and Pichoir 1991a)
Duplex model:
fw
1 1:65 10
6
gw
2
1 a1:65 10
6
gw
41
where a 0:18 2=g 8 10
9
E
c
0:005
Z
p
with E
c
in electron volts.
Puchou and Pichior model:
fw
F
1
w F
2
w
F
42
where the formulation for F, F
1
, and F
2
are given in Pouchou and Pichoir (1991a).
Acomparison of the dierent models and the experimental data is shown in Figure 13.
The results of the study indicated that no one model is best for all targets and conditions.
The analyst interested in the highest accuracy should select the absorption correction that
works the best for the particular type of sample and experimental conditions.
c. Fluorescence Correction, k
F
The characteristic uorescence correction is necessary when the analysis involves the
following conditions:
1. The characteristic x-ray peak from element j at energy E
j
is greater than the
critical excitation potential E
c;i
for element i.
2. The energy dierence E
j
E
c;i
< 5 keV.
When these conditions exist, the characteristic line of element j will excite the characteristic
line of element i in the specimen as shown in Figure 14. This uorescence results in an
increased intensity for element i that must be taken into account in obtaining a quanti-
tative analysis. In this case, the excitation is caused by x-rays and not electrons; hence, the
generation of the secondary x-rays will originate from a much larger volume of the sample
that if it were generated directly by electrons. In addition, the measured intensity of the
uorescence radiation is proportional to the x-ray emergence angle.
The basic formulation of the uorescence correction is given by
F
i
1
j
I
f; j
I
i
_ _
std
1
j
I
f; j
I
i
_ _
1
sam
43
Copyright 2002 Marcel Dekker, Inc.
Figure13 Plots of the experimental data and literature models for f (w), where the residuals 100f f w
model
f w
experimental
=
f w
experimental
g and high absorption w>10,000, medium1000<w<10,000, and loww<1000. (From Small et al., 1991.)
Copyright 2002 Marcel Dekker, Inc.
The factor I
f
=I
i
is the ratio of the x-ray-excited intensity for the characteristic line of
element i to the electron-excited intensity. The summation over j is required because the
total correction must be summed over all the elements in the specimen.
The most commonly used uorescence correction factor was originally developed by
Reed (1965) and is given in Eq. (44) [from Reed (1997)]:
I
f
I
i
0:5C
j
m
i
j
m
j
_ _
r
i
1
r
i
_ _
o
j
A
i
A
j
_ _
U
j
ln U
j
U
j
1
U
i
ln U
i
U
i
1
_ _
ln1 m
m
ln1 n
n
_ _
44
where C
j
is the concentration of element j; m is the mass absorption coecient, r is the
absorption jump ratio dened as the ratio of m=r on the high-energy side of the ab-
sorption edge to m=r on the low-energy side, the factor r
i
1=r
i
is 0.88 for K line
uorescence and 0.74 for L line uorescence, o
j
is the uorescent yield of element j,
A
i
is the atomic weight, U is the overvoltage, m m
i
cosec c=m
j
and n s (Lenards
coecient)=m
j
.
In practice, when calculating the uorescence correction from Eq. (43), the standard
is either a pure element or there is no signicant uorescence of element i by other elements
in the standard. For this situation, Eq. (43) reduces to
F
i
1 1
j
I
f; j
I
i
_ _
1
sam
45
which is the most common form of the correction found in analytical procedures. Gold-
stein et al. (1981e) demonstrated the magnitude of the uorescence correction with the
analysis of a 10 wt% Fe90 wt% Ni alloy. The results of the analysis are summarized in
Table 3 and show that the intensity of the uoresced iron, I
f
=I
Fe
ranges from 16.8% to
34.6% of the observed iron x-ray intensity. The magnitude of the correction is lower at the
smaller accelerating potential and detector takeo angle.
Figure14 Secondary fluorescence in a specimen.
Copyright 2002 Marcel Dekker, Inc.
In addition to the secondary excitation by characteristic x-rays, x-rays can also be
excited by the continuum x-rays produced in the sample. The continuum x-rays are the
result of the deceleration of beam electrons in the Coulombic eld of the specimen
atoms. This radiation forms an x-ray background that is slowly varying with energy
and ranges from E
0
to zero. Although there has been some eorts by Heinrich (1987) to
simplify the calculations for determining continuum uorescence, the basic formulation
for the added intensity from the continuum uorescence, I
c
, was derived by Henoc
(1968):
I
c
f
"
Z; o; r;
m
r
_ _
; C
_ _
46
The calculation of the continuum uorescence is relatively complicated, involving
integration over the range of E
0
to E
c
for each element in the specimen. Because the
correction can be as large as 24%, it should be included for highest accuracy. Myk-
lebust et al. (1979), investigating the continuum uorescence, determined the following
experimental conditions for which a correction for the uorescence by the continuum is
required:
fw > 0:95
C
i
< 0:5
"
Z from the standard much dierent from
"
Z of the sample
These requirements translate to a signicant continuum uorescence correction in
the analysis of a small amount of an element with a high-energy x-ray line such as zinc, in a
light matrix such as carbon. In Figure 15, Heinrich compared the uorescence intensities
caused by the x-ray continuum to the intensities generated by the primary electron beam
for Al and Cu at dierent electron beam energies and takeo angles.
2. Analysis Method Based on Integrationof frz Curves
In the classical ZAF procedures, the absorption correction, as mentioned earlier, was
based on Philiberts parameterization of a limited set of frz curves, including several
simplications. The shape of his curve, although not correct, works well for higher-
energy x-rays for which the shape of the curve is not critical due to the relatively long
absorption path lengths. However, for light elements in specimens with high absorption,
Al in Cu, or elements with low-energy x-rays, such as B, C, N, and O, where absorption
is signicant, the shape of the frz curve is critical. As a result, the classical ZAF
routines based on the simplied Philibert model produce large errors in the concentra-
tions for these elements.
Table 3 Fluorescence of FeKa by Ni in 10 wt% Fe90 wt%
Ni alloy
C (deg) E
0
(keV) I
f
=I
Fe
F
Fe
52.5 15 0.263 0.792
15.5 15 0.168 0.856
52.5 30 0.346 0.743
15.5 30 0.271 0.787
Source: From Goldstein et al., 1981e.
Copyright 2002 Marcel Dekker, Inc.
Several researchers, including Love (1983), Love and Scott (1988), Bastin et al.
(1986), Pouchou and Pichoir (1988), and Sewell et al. (1985a) have made experimental
measurements of x-ray depth distributions, frz curves, in a large number of targets,
including pure elements, alloys, and oxides. In addition, other researchers [e.g., Heinrich
et al. (1988), Gauvin et al. (1995), Murata et al. (1983), and Karduck and Rehbach (1988)]
have made Monte Carlo calculations of x-ray depth distributions. The increased knowl-
edge regarding the behavior of these curves for dierent materials has made it possible to
empirically develop corrections based on the integration of frz curves. These procedures
are particularly attractive for the analysis of low-energy x-ray lines from the elements
boron through uorine, where fw is much less than 0.7 and the accuracy of the ZAF
method is low. In these procedures, frz refers to the depth distribution of x-rays in a
bulk sample normalized to the x-ray intensity produced in an innitely thin, unsupported
lm of the same composition. In principle, the frz curve as described by Castaing (1951)
contains the information necessary to correct experimental data for both the atomic-
number correction, Z (integrated area under curve), and absorption correction, A (shape
of the curve). The Gaussian model proposed by Packwood and Brown (1981) combines
the Z and A corrections in their model for frz producing a single emitted intensity. It
should be noted, as pointed out by Armstrong (1988a), that the use of frz expressions
Figure15 Comparison of x-ray intensities fluoresced by the x-ray continuum to the intensities gen-
erated by the primary electron beam. I
0
c
=I
0
p
refers to the relative emergent fluorescent intensity at a
given angle and I
c
=I
p
refers to the relative generated fluorescent intensity. (From Heinrich 1981d. Re-
produced with permission of Van Nostrand Reinhold.)
Copyright 2002 Marcel Dekker, Inc.
for the correction of x-ray absorption requires only that a given expression produce the
correct shape for the frz curve. The use of the frz expression for the atomic-number
correction is risky because this requires that the thickness of the tracer lms and the
normalizing thin lms in the frz experiments for dierent matricies be known to a high
degree of accuracy (Armstrong 1988a).
In practice, several of the frz methods use of the shape of the frz curve for the
absorption correction and combine that with separate Z and F corrections. Independent of
the procedure used, the analysis of specimens using the frz correction method is de-
pendent on the derivation of an accurate expression that describes the experimental frz
curves and the degree to which that expression can be universally applied to systems with
unknown frz curves.
a. Analytical Expressions for frz
Listed below are a series of parameterizations that have been used by various authors to
represent the shape and area of frz curves.
Packwood and Brown. Packwood and Brown (1981) proposed a modied Gaussian
function to describe the shape of frz curves. In their procedure, both the atomic-number
correction and the absorption correction are parametrized by the model. The Gaussian is
centered at the sample surface and modied by a transient function that was introduced to
model the near-surface x-ray distribution. Figure 16 is a plot of the modied Gaussian
function showing the inuence of the various parameters a; b; g
0
, and f
0
. In the model given
in Eq. (47) [from Packwood (1991)], the a term related to the width of the Gaussian and g
0
relates to the amplitude. The b termin the transient is related to the slope of the curve in the
near-surface region (i.e., the rate at which the focused electron beamis randomized through
scattering in the target and the y intercept, f
0
, is related to the surface ionization potential):
frz g
0
expa
2
rz
2
..
Gaussian term
f1 g
0
f
0
=g
0
expbrzg
..
Transient term
47
The formulas for the dierent terms can be found in Packwood (1991).
Figure 16 Modified Gaussian function for fitting f(rz) curves. (From Packwood and Brown,
1981. Copyright John Wiley & Sons Limited. Reproduced with permission.)
Copyright 2002 Marcel Dekker, Inc.
Pouchou and Pichoir. Pouchou and Pichoir (1991b) proposed a frz procedure,
PAP that was designed to provide a broad range of analytical capabilities including bulk
and stratied sample analysis. In this procedure, the authors separate the parameters that
dene the shape of the frz curve (i.e., the absorption correction) from the parameters
that dene the total integral of the frz curve (i.e., the generated primary intensity or
atomic-number eect). The authors refer to their procedure as neither like a ZAF
structure with separate Z, A, and F corrections nor like a global structure such as the
model of Packwood and Brown (1981) in which the emerging intensity is obtained directly
from the parametric description of frz.
Area parameter: In the PAP procedure, the authors dene the integral of the frz
curve from zero to innity, Eq. (29), as F. The analytical expression for F which
is used for the atomic-number correction is given in Eq. (48), where R is the
backscatter factor based on the mean-backscatter coecient and the mean-
reduced backscattered electron energy for the electrons, 1=S is the deceleration
factor, and Q
A
1
E
0
is the cross section for x-ray production at level 1 for
element A and voltage E
0
:
F
R
S
_ _
Q
A
1
E
0
48
A detailed description of the analytical expression used to calculate F is given
in Pouchou and Pichoir (1991b).
Shape parameter: For their procedure, Pouchou and Pichoir placed the following
conditions on their analytical expression for the shape of the frz curve:
The integral of the shape function must be equal to F.
The shape function must have a nite value for f0.
The parameterization of the function must have a maximum, designated
as R
m
.
The function must vanish with a horizontal tangent at a predetermined
range, R
x
.
The shape of the frz curve is dened by a pair of intersecting parabolas. For this
model, the curve is dened by the following parameters:
f0 the surface ionization function
R
m
depth at which the maximum in the frz curve occurs
R
x
the x-ray range
The area under the frz curve (i.e., the total generated intensity)
The equation for the rst parabola nearest the target surface is given in Eq. (49) for
values of rz between 0 and R
c
and the second parabola by Eq. (50) for values
of rz between R
c
and R
x
:
f
1
rz A
1
rz R
m
2
B
1
49
f
2
rz A
2
rz R
x
2
50
where the parameters A
1
; A
2
, and B
1
are functions of R
m
; R
x
, and f0, and R
c
is the crossover point of the parabolas. The equation for R
c
is a function of the
generated X-ray intensity, I, which is the integral of the area under the frz
curve, Equation (29). Figure 17 is the plot of the double-parabola curve with
the various parameters located on the curve. Pouchou and Pichoir (1991b)
introduced a simplied version of their model based on
Copyright 2002 Marcel Dekker, Inc.
frz Aexparz Brz f0 A expbrz 51
The coecients A; B; a, and b can be determined from the following parameters:
1.
_
1
0
frzdrz
A
a
f0 A
b
B
b
2
2. f0 is the surface ionization.
3.
"
Rmean ionization depth,
"
R
A
a
2
f0 A
b
2
2B
b
2
4. The initial slope B aA bf0 A.
Merlet. Merlet (1994) developed a parameterization of frz based on a double
partial Gaussian expression for the description of the frz curve as shown in Figure 18.
The parameterization is dependent on dening the depth distribution at three points of the
frz curve, the surface at 0; f0, the curve maximum at rz
m
; fm, and the x-ray
range at rz
x
; fm=100. Equation (52) denes the partial Gaussian f
1
for rz values
between the surface and rz
m
, and Eq. (53) denes the partial Gaussian f
2
for rz values
from rz
m
to rz
x
:
f
1
f
m
exp
rz rz
m
b
_ _
2
_ _
52
where
b
rz
m
lnfm=f0
0:5
and
f
2
f
m
exp
rz rz
m
a
_ _
2
_ _
53
where
a 0:46598rz
x
rz
m
2
54
For rz
m
rz 1
f
r
rz f
m
expa
2
rz rz
m
2
55
Figure18 Double partial Gaussian for fitting f(rz) curves. (From Merlet, 1994). Data points are
taken from the work of Castaing and Henoc. (From Castaing and Henoc, 1965; Reproduced with
permission of Springer-Verlag.)
Copyright 2002 Marcel Dekker, Inc.
In this case, the b term in Eq. (54) refers to the Gaussian rate of increase from 0 to rz
m
and
is dierent from the b term in the PROZA formulation. Likewise, the a term in Eq. (55) is
the Gaussian rate of decrease from rz
m
out to innity.
Bastin et al. adopted the new PROZA96 formulation for frz because of the
constraints placed on the position and height of the maximum in the frz curve by the g
term in the PROZA formulation. The area under the frz curve in PROZA is xed by
the F term. The main advantage of the new formulation is in absorption. The shape of the
frz curve in PROZA96 results in an increased exibility for analyzing ultralight ma-
trices such as Be, B, and C, particularly for thin-lm calculations where the shape of the
frz curve is critical, and at high overvoltages, the maximum will be shifted deeper into
the sample.
Armstrong. Armstrong (1991) developed a frz model, Eq. (56), based on the
Gaussian equation from Packwood and Brown:
frz g
0
e
a
2
rz
2
1 qe
brz
56
where
a 2:97 10
5
"
Z
1:05
"
AE
1:25
0
ln1:166E
0
=J
E
0
E
c
_ _
0:5
b
8:5 10
5
"
Z
2
"
AE
2
0
g
0
1
g
0
5pU
0
U
0
1 lnU
0
lnU
0
5 5U
0:2
0
q
g
0
f0
g
0
"
Z and
"
A are the atom-concentration weighted atomic number and atomic weight,
respectively. The surface ionization function, f0, can be determined from the expression
of either Reuter (1972) or Love and Scott (1978).
Love and Scott. Love et al. (1984) proposed a frz method in which they in-
troduced a separate atomic-number correction. With the separate treatment of the atomic-
number correction, the frz curve is used only for target absorption. As mentioned
earlier, target absorption requires only that the shape of the curve be correct and not the
absolute height. The authors proposed a quadrilateral prole, Figure 19, to t the shape of
the curve. The quadrilateral is dened by the y intercept f0, the position and amplitude
of the peak, rz
m
; fm, and the x-ray range, rz
r
. From the quadrilateral model, the
analytical expression for fw can be written as
fw 2A
1
B C 57
where
A rz
r
rz
m
rz
m
hrz
r
w
2
B expwrz
m
h expwrz
r
wrz
r
rz
m
h 1
C
expwrz
m
rz
r
hrz
r
hrz
r
rz
r
rz
m
h
fm
f0
Copyright 2002 Marcel Dekker, Inc.
To constrain the quadrilateral model when fw values exceed 0.5, the authors expressed
both rz
m
and rz
r
in terms of the mean depth of x-ray generation, " rz (Scott and Love,
1991). In conjunction with the quadrilateral t to the frz curve for target absorption,
the atomic-number correction combines the stopping power factor from Eq. (14) with the
backscatter factor R, Eq. (58), from Love et al. (1978):
1 R
Z
_ _
0:6
IU
0
ZGU
0
58
where U
0
is the overvoltage, Z is the backscatter coecient,
IU
0
0:3
1
U
0
exp1:5 1:5U
0:25
0
_ _
and
GU
0
0:368 0:075 ln U
0
exp1 2:3U
4
0
A full description of the quadrilateral model can be found in Scott and Love (1991).
3. Evaluationof ZAFand frz Analytical Correction Procedures
The relationship describing a ZAF or frz correction procedure gives the impression that
the concentration of a given element can be calculated directly from the k ratio. This,
however, is not the case because the parameters used in the dierent correction schemes
are dependent on the composition of the sample, including the element of interest. For a
simple binary system, the analyst could construct a calibration curve relating measured
x-ray intensity to concentration. However, this does not apply to more complicated sys-
tems and the primary method used in correction procedures is an iterative method.
A general owchart for a typical ZAF or frz correction method is shown in
Figure 20. The measured k ratio is used as the initial estimate of composition by taking
C
i
i
k
i
=C
i
. The dierent ZAF or frz corrections are then applied to the initial es-
timate of concentration to obtain the corresponding estimate of k. Next, a hyperbolic
iteration, Eq. (59), is used to obtain a value for a:
Figure 19 Quadrilateral function for fitting f(rz) curves. (From Scott and Love, 1991. Repro-
duced with permission of Plenum Press.)
Copyright 2002 Marcel Dekker, Inc.
Figure 20 General flowchart for a typical ZAF or f(rz) correction method.
Copyright 2002 Marcel Dekker, Inc.
k
i
a
i
C
i
a
j
C
j
59
The a value is then used with the measured k ratio to obtain a new estimate of con-
centration that is then normalized and the iteration continued. Convergence with the
hyperbolic function is rapid and seldom requires more than three iterations.
The practical lower limit for quantitative x-ray microanalysis is on the order of
0.01% by weight for an element homogeneously distributed in an innitely thick, polished
specimen. The x-ray excitation cross-sections and the signal-to-noise ratio of the x-ray
detection systems primarily set this limit. The accuracy of a classical ZAF electron-probe
analysis is shown in Figure 21, which is a histogram of the relative errors from 264
analyses of binary metal alloys.
The accuracy of several dierent analytical models, including both ZAF and frz
models is shown in Figures 2224, which are adapted from a table compiled by Pouchou
and Pichior (1991b) for 826 analyses of binary compounds. In these gures, the horizontal
line for a given model is 1 standard deviation for the mean value of k (calculated)=k
(experimental), which is represented by a lled circle. The length of the horizontal line for
a given model is an indicator of the variation in the accuracy for the performance of that
model on the given dataset. Similarly, the proximity of the mean with respect to unity is an
indicator of any systematic bias in the model.
Figure 22 displays the results from the entire 826 analyses. The frz models 1 and 2
have the lowest variations in accuracy, both less than 2%. frz models 3 and 4 and ZAF
model 5 are similar with variations less than 3%. The remaining two ZAF models (6 and
7) have variations greater than 4%. The systematic biases for the techniques are small and
fall within 1%, except for the ZAF method 5, which has a systematic bias of about
1.5%. Figure 23 displays the results of 577 of the 826 analyses involving samples where
the atomic-number eect is greater than the x-ray absorption eect. The variances for all
the models are similar generally less than 2%, except the frz model 4, which has a
variance of about 3%. The biases are also small, except for model 4, which has a positive
systematic bias of about 1.5%. Figure 24 shows the result for 242 analyses involving
samples where x-ray absorption is the predominate correction. For these samples, there is
a signicant dierence between the ZAF and frz models. Three of the fz models (1, 2
and 4), have variances in accuracy less than 3% and biases less than 1%. ZAF model 5 and
frz model 3 are similar for both variation, about 4%, and systematic bias, about 2%.
The remaining two ZAF models (6 and 7) have signicantly larger variances, about 6%,
and systematic biases of around 4%.
In addition to the above dataset, Pouchou and Pichior (1991b) also compiled data
for a series of boride analyses, the results of which are given in Figure 25. The variance in
accuracy for the dierent methods range from about 4% for several of the frz methods
to about 25% for the worst ZAF method. Five of the seven methods show a slight sys-
tematic bias of 12% with models 7 and 8 having signicant positive biases of 8% and
15%, respectively. Finally, Table 4 lists the results from the analyses of compounds
containing ultralight elements with PROZA96 from Bastin et al. (1998).
B. Empirical Approach to Quantitative Analysis
Ziebold and Ogilvie (1964) developed a quantitative analysis procedure for emitted x-rays
that is based on Castaings third approximation for generated x-rays given in Eq. (59)
(Castaing, 1951). In this empirical procedure, the relationship between the k ratio and
Copyright 2002 Marcel Dekker, Inc.
Figure 21 Histogram of the relative errors for a classical ZAF analysis from Goldstein et al.
(From Goldstein et al., 1975b. Reproduced with permission of Plenum Press.)
Figure 22 Results from the analysis of 826 binary compounds with several different quantitative
analysis procedures. (Adapted from Pouchou and Pichoir, 1991b.)
Copyright 2002 Marcel Dekker, Inc.
Figure23 Results for 577 of the 826 binary compounds where the atomic-number effect is greater
than the x-ray absorption effect. (Adapted from Pouchou and Pichoir, 1991b.)
Figure 24 Result for 242 analyses involving samples where x-ray absorption is the predominate
correction. (Adapted from Pouchou and Pichoir, 1991b.)
Copyright 2002 Marcel Dekker, Inc.
concentration, expressed in Eq. (2), can be expanded to include real samples by the in-
troduction of an eciency factor a for element a in a sample:
k
a
a
a
C
a
Sa
i
C
i
where the summation is over all elements in the sample.
Ziebold and Ogilvie, using Eq. (59), expressed the relationship between C and k for a
binary system containing elements i and j as
C
i
k
i
a 1 aC
i
60
or
k
i
C
i
1
a
1
1
a
_ _
k
i
61
where a for the binary is referred to as the a factor and equals a
j
=a
i
.
Figure 25 Results from the analysis of a series of borides. (Adapted from Pouchou and Pichoir,
1991b.)
Table 4 Performance of PROZA96 on Ultralight Element Data: Comparison of the Calculated k
ratio (k
0
) with the Measured One (k)
Element No. of measurements Mean k
0
=k Root mean square deviation (%)
Boron 192 1.0033 3.3114
Carbon 117 1.0001 3.2092
Nitrogen 144 0.9989 3.6820
Oxygen 294 1.0004 2.0090
Source: From Bastin et al., 1998.
Copyright 2002 Marcel Dekker, Inc.
Finally, solving for a yields Eq. (62) which is the formulation for the hyperbolic
approximation because the plots of the calibration curves of C versus k described by this
equation are hyperbolas:
a
C
1 C
_ _
1 k
k
_ _
62
Equations (60) and (61) indicate a linear relationship between C=k and C, and k=C
and k so that a plot of C=k versus C or k=C versus k are straight lines. The slope of the line
will be positive if a is greater than 1 and negative if a is less than 1. Experimental con-
rmation of the hyperbolic approximation is shown in Figure 26, which is a plot of k=C
versus k for experimental measurements on AgLa x-rays from four dierent AgAu alloys
at several dierent beam energies. Except for the 5-keV line, the plots for the dierent
alloys as predicted by the hyperbolic approximation are straight lines.
The concentrations C
i
and C
j
, for elements i and j in an unknown binary, can be
determined from Eq. (62) if the appropriate values for a
i
; a
j
; k
i
, and k
j
are known. In
theory, the k values do not present a problem because they are experimentally measured.
The determination of a
i
and a
j
, however, requires the analysis of a standard binary con-
taining the same elements as the unknown. In the procedure for determining the a factors,
k
i
and k
j
are obtained for the standard binary by comparison to pure-element standard of i
and j. Because the values of C
i
and C
j
, in the standard binary are known, it follows that a
i
Figure 26 k=C versus k for AgAu alloys at various electron beam energies. (From Goldstein
et al., 1981f. Reproduced with permission of Plenum Press.)
Copyright 2002 Marcel Dekker, Inc.
and a
j
can be calculated from Eq. (59). In the case of a binary, Eq. (63) can be used to
calculate the concentration for element i in the unknown binary:
C
i
k
i
a
1 k
i
a 1
63
The values for a are specic for a given experimental setup, and if the analysis
parameters are changed, the a values must be recalculated. As long as the analyst is
concerned with measurements on a given binary system under constant conditions, then
Eq. (63) provides a very rapid and easy method for the determination of elemental
composition with an accuracy that is comparable to that of the full ZAF procedure.
In addition to their work on binary systems, Ziebold and Ogilvie (1964) also derived
the analytical expression for a ternary system. The basic equation for the ternary remains
Eq. (63), except that the combined a factor for element 1 in a ternary compound con-
taining elements 1, 2, and 3 is given by the following expression:
a
1;2;3
a
1;2
C
2
a
1;3
C
3
C
2
C
3
64
An important aspect of the ternary derivation is that the combined a factor for the ternary
system in Eq. (64) can be expressed as a combination of the individual a factors for the
dierent binaries.
Bence and Albee (1968), building on the work of Ziebold and Ogilvie, extended the
application of the hyperbolic approximation to multielement systems containing six to
eight elements. They were interested in obtaining a rapid, accurate procedure for the
analysis of specimens of geological origin. In such specimens, it was not practical in 1968
to apply the full ZAF corrections due to the limited computational resources available.
Even with modern computer capabilities, the empirical approach oers a simple,
straightforward approach to quantitative analysis that can be programmed and executed
on a small pocket calculator. In addition, because of its simplicity and speed, it can be used
for real-time processing of spectra image data (see, Sec. V). The general formula for ele-
ment i in a system of n components is given by
C
i
k
i
b
i
65
where
b
i
k
2
a
i2
k
3
a
i3
k
n
a
in
k
2
k
3
k
n
In Eq. (65), the a
i2
factor refers to the a factor for element i in a binary of element 2 and
element i. Similarly, k
2
refers to the intensity ratio of element 2 in the specimen to the
element 2 in the standard. In practice, the analyst is required to measure nn 1 values
for a, which requires nn 1 binary standards.
The owchart for the empirical method of quantitative analysis is shown in Figure 27.
The analyst, having determined the needed a values, measures the appropriate k values for
the unknown and calculates the b termfromEq. (65). The b values are then used to calculate
a rst approximation of the elemental concentrations. From Eq. (65), a second set of b
values are determined with the approximate concentrations used for the ks. The iteration
loop is continued until the dierence in successive b terms is below a predetermined limit.
As previously mentioned, the analysis of multielement systems requires nn 1
standards. Because the large number of standards for the determination of a factors in a
complex system may not be available, a combination ZAF, frz, or Monte Carlo a factor
Copyright 2002 Marcel Dekker, Inc.
approach has been developed. In this procedure, the analyst calculates the necessary a
factors by assuming an appropriate composition and then running a ZAF, frz, or
Monte Carlo procedure to back-calculate the corresponding k ratios. The a factors can
then be determined from Eq. (60). It is important that the analyst realize that the total
uncertainty in this combined approach includes the uncertainties of both the empirical and
the ZAF methods. Albee and Ray (1970) have used this procedure to determine the a
factors for 36 elements relative to simple oxides. Laguitton et al. (1975) and Bence and
Holzwarth (1977) proposed additional corrections along with more elaborate polynomial
representations of the a factors. Armstrong (1988b) calculated a large number of oxide a
factors using a combination of ZAF, frz, and Monte Carlo methods combined with a
second-order polynomial for the a factors, Eq. (66):
a
i
i; j
c d
C
i
C
i
C
j
_ _
e
C
i
C
i
C
j
_ _
2
66
where C
i
and C
j
are the oxide weight fractions of elements i and j in the binary and c; d,
and e are the coecients of the polynomial t. The results indicated that the C=k values
Figure 27 An empirical analysis procedure.
Copyright 2002 Marcel Dekker, Inc.
from the second-order polynomial a factor were essentially identical to the ZAF, frz, or
Monte Carlo method they were based on and the results can be expected to have the same
accuracy and precision as the conventional procedures.
The empirical method of analysis is most accurate when the dierence between the
sample and standards is mainly the result of x-ray absorption. The procedure is less ac-
curate when there are large dierences in atomic number and is least accurate when there
is a signicant contribution from uorescence. Given these limitations, the empirical
procedure works best for the analysis of materials such as geologic specimens and oxide
systems that have relatively low mean atomic numbers and minimal secondary uores-
cence.
C. Standardless Quantitative Electron Probe X-ray Microanalysis with EDS
An increasing trend in recent years in performing quantitative electron probe x-ray
microanalysis with energy-dispersive x-ray spectrometry has been the substitution of
standardless methods in place of the traditional approach of measuring standards
containing the elements of interest on the same analytical instrument under the same
excitation and detection conditions (Russ, 1974). In standardless methods, the appropriate
standard intensities necessary for quantication are either calculated from rst principles,
considering all aspects of x-ray generation, propagation through the solid target, and
detection (true standardless), or else standard intensities are derived from a suite of
experimental measurements performed remotely and adjusted for the characteristics of the
local instrument actually used to measure the unknowns (tted standards). With either
route to standard intensities, the resulting k values (unknown=standard) are then
subjected to matrix corrections with one of the usual approaches [ZAF, frz, etc.].
The apparent advantages of standardless analysis are considerable. Instrument op-
eration can be extremely simple. There is no need to know the beam current, and, indeed,
it is not even necessary for the beam current to be stable during the spectrum accumu-
lation, a real asset for instruments such as the cold-eld-emission gun scanning electron
microscope (SEM), FEG-SEM, where the beam current can be a strong function of time.
Moreover, the detector solid angle is of no consequence to the quantitative procedure.
Both the beam current uncertainty and the detector solid angle uncertainty are eectively
hidden by forcing the analytical total to a predetermined value (e.g., unity when all
constituents are measured). When the spectrum has been accumulated to the desired level
of statistical precision, the analyst needs to specify only the beam energy, the x-ray takeo
angle, and list of elements to be quantied (of this list can be derived from an automatic
qualitative analysis). The software then proceeds to calculate the composition directly
from the spectrum, as described below, and in the resulting output report, the con-
centration is often specied to three or four signicant gures. The measurement precision
for each element, calculated from the integrated peak and background counts, is also
reported. The precision is really only limited by the patience of the analyst and the stability
of the specimen under electron bombardment, so that precision values below 1% relative
(1s can be readily achieved, even for minor constituents. Although such excellent precision
is invaluable when comparing dierent locations on the specimen, the measurement pre-
cision is independent of the accuracy of the analysis. For the traditional approach of
measuring standards and calculating matrix corrections, there exists an extensive literature
describing testing of the quantitative procedures. These test results are usually presented in
the form of error histograms, which the analyst can use to make an estimate of the ac-
curacy of the result. The relative error is dened as
Copyright 2002 Marcel Dekker, Inc.
Rel: error 100%
Measured True
True
_ _
67
An example of such an error histogram is shown in Figure 21, from which it can be de-
duced that, under the particular conditions used [binary alloys measured against pure
element standards with NISTZAF matrix corrections (1975 formulation)], 95% of the
analyses lie within 5% relative of the correct value. Unfortunately, such error histo-
grams are virtually never available for standardless analysis procedures as implemented in
commercial EDS systems. Recent work indicates that these standardless analysis proce-
dures may, in fact, be subject to large errors (Newbury et al., 1995; Newbury, 1998).
1. First-Principles Standardless Analysis
Calculating an equivalent standard intensity from rst principles requires solution of
I
std
rN
0
A
_ _
o
_
E
c
E
0
Q
dE=ds
dE
_
_
_
_Rfwe 68
where the terms in brackets represent the excitation function: r is the density, N
0
is
Avogadros number, A is the atomic weight, o is the uorescence yield, Q is the ionization
cross section, dE=ds is the stopping power, E is the electron energy, E
0
is the incident beam
energy, and E
c
is the critical excitation energy. The other terms correct for the loss of x-ray
production due to electron backscattering (R), the self-absorption of x-rays propagating
through the solid fw, and the eciency of the detector, e. It is useful to consider the
condence with which each of these terms can be calculated.
a. Excitation
Three terms are critical in the excitation function: the ionization cross section, the uor-
escence yield, and the stopping power.
1. Ionization cross section: Several parameterizations of the K shell ionization cross
section are plotted in Figure 28. The variation among these choices exceeds 25%.
Although we cannot say that any of these is correct, it is certain that they cannot
all be correct. Moreover, because of the continuous energy loss in a solid target,
the cross section must be integrated from E
0
to E
c
, through the peak in Q and
the rapid decrease to U1. This region of the cross section is poorly
characterized, so that it is dicult to choose among the cross-section
formulations based on experimental measurements. The situation for L and M
shell cross sections is even more unsatisfactory.
2. Fluorescence yield: Various experimental determinations of the K shell
uorescence yield are plotted in Figure 29. Again, a variation of more than
25% exists for many elements. The situation for L and M shell transitions is
substantially more incomplete.
3. Stopping power: The classic Bethe formulation of the stopping power becomes
inaccurate at low beam energies (<5 keV), and eventually with decreasing
energy, it becomes physically unrealistic with a sign change. The accuracy of the
stopping power matters for calculating Eq. (68) because the cross section must
be integrated to E
c
, which for low-energy x-rays (e.g., C, N, O, F) involve
electron energies in this regime. As discussed earlier in this chapter, several
authors have suggested modications to the Bethe formulation to correct for the
Copyright 2002 Marcel Dekker, Inc.
low beam energy regime. Unfortunately, the base of experimental measurements
necessary to select the best choice is just being developed.
b. Backscatter Loss
The backscatter-loss correction factor R was initially formulated based on experimental
measurements of the total backscatter coecient and the dierential backscatter coe-
cient energy. Although careful, extensive measurements of total backscatter were available
in the literature, the database of the dierential backscatter coecient as a function of
energy, a much more dicult experimental measurement, was limited to a few elements
and was available at only one emergence angle. The development of advanced Monte
Carlo simulations has permitted the rigorous calculation of R over all scattering angles
and energy losses so that this factor is probably known to an accuracy within a few percent
across the periodic table and the energy range of interest.
c. X-ray Self-absorption
The self-absorption of x-rays in the hypothetical standard is calculated with the for-
mulation of the absorption factor used in matrix corrections. Fortunately, the absorption
correction is generally small for the x-rays of a pure element, so that at least for higher-
energy characteristic x-rays (e.g., greater than 3 keV), there is little uncertainty in this
factor. However, the self-absorption increases both as photon energy decreases and in-
Figure 28 Ionization cross section as a function of overvoltage U, as formulated by different
authors.
Copyright 2002 Marcel Dekker, Inc.
cident electron energy increases, so that the error in calculating the intensity emitted from
an element emitting low-energy photons (e.g., carbon) could be signicant.
d. Detector Eciency
The last term in Eq. (69) is one of the most dicult with which to deal:
e exp
_
m
r
_ _
win
r
win
t
win
m
r
_ _
Au
r
Au
t
Au
m
r
_ _
SiDL
r
Si
t
SiDL
m
r
_ _
con
r
con
t
con
m
r
_ _
ice
r
ice
t
ice
_
1 exp
m
r
_ _
Si
r
Si
t
Si
_ _
69
In the traditional k-value approach, the detector eciency cancels quantitatively in the
intensity ratio because the same x-ray peak is measured for the unknown and the standard
under identical (or at least accurately reproducible) spectrometer conditions. When
standardless analysis is performed, this cancellation cannot occur because x-ray peaks of
dierent energies are eectively being compared and, therefore, accurate knowledge of the
detector eciency becomes critical. Detector eciency is mainly controlled by absorption
losses in the window(s) and detector structure. The expression for detector eciency, Eq.
(69), consists of a multiplicative series of absorption terms for each component: detector
window (win), gold surface electrode (Au), semiconductor dead layer (DL, actually a
partially active layer below the electrode and the source of incomplete charge phenomena),
and a transmission term for the detector thickness. Additionally, for most practical
measurement situations, there may be absorption contributions from contamination on
Figure 29 K Shell fluorescence yield as a function of atomic number.
Copyright 2002 Marcel Dekker, Inc.
the detector crystal, usually arising from ice buildup due to pinhole leaks in the window or
support (ice) and from contamination on the detector window usually deposited as hy-
drocarbons from the microscope environment (con).
An example of the detector eciency as a function of photon energy for several
window materials is shown in Figure 30. The choice of the window material has a strong
eect on the detector eciency for photon energies below 3 keV, and accurate knowledge
of the window and detector parameters is vital for accurate interelement eciency cor-
rection across the working range of the detector, typically 100 eV to 12 keV. The change in
the detector eciency with the accumulation of ice is illustrated in Figure 31. The buildup
of ice and other contaminants and the resulting loss in eciency is referred to as detector
aging. Detector aging can result in a signicant loss of the low-energy photons (<3 keV)
relative to the higher-energy photons (312 keV) (Fiori and Newbury, 1978).
2. Fitted Standards Standardless Analysis
The tted standards technique is the more widely used approach for implementing
standardless analysis on commercial computer-assisted EDS analyzer systems. In the
tted standards technique, a suite of pure-element standards covering K-, L-, and M- family
x-rays is measured at one or more beam energies on an electron beam instrument equipped
with an EDS detector whose eciency is known fromindependent measurements or at least
can be estimated. An example of such a measurement for a portion of the Kseries frompure
Figure 30 Detector efficiency for various windows (Al window coating 0.02 mm; Au electro-
de 0.01 mm; Si dead layer 0.03 mm): (a) beam energies 05 keV; (b) beam energies 01 keV.
Copyright 2002 Marcel Dekker, Inc.
elements measured at 20 keV is shown in Figure 32. Missing elements can be calculated by
simply tting the available peaks and interpolating (e.g., in Figure 32, the intensity for
gallium could be estimated by tting the smoothly varying data and interpolating). From
the smooth change in peak height with atomic number seen in Figure 32, such an inter-
polation should be possible with reasonable accuracy. The situation is not as satisfactory in
the L and M families, as illustrated in Figures 33 and 34, respectively, because the uor-
escence yield is a much more complicated function of atomic number.
If the analysis must be perfomed at a beam energy other than that of the spectral
database, then Eq. (68) must be used to shift the intensities appropriately. Similarly,
if a dierent EDS detector is used, the detector eciency must be corrected using Eq. (69).
The tted standards standardless procedure is expected to be more accurate than the
rst-principles standardless procedure because it is tied to actual experimental measure-
ments which directly incorporate the eects of the cross section and the uorescence yield,
at least over the range of the elements actually measured.
a. Testing the Accuracy of Standardless Analysis
The accuracy of standardless analysis procedure has been tested by carrying out analysis
on microhomogeneous materials of known composition: NIST Microanalysis Standard
Reference Materials (SRM), NIST Microanalysis Research Materials (RM), stoichio-
metric binary compounds (e.g., IIIV compounds such as GaAs and IIVI compounds
such as SrTe), and other materials such as ceramics, alloys, and minerals for which the
Figure 30 Continued.
Copyright 2002 Marcel Dekker, Inc.
composition was available from independent chemical analysis and for which micro-
homogeneity could be established. Compositions were carefully chosen to avoid serious
spectral overlaps (e.g., PbS, MoS
2
). Light elements such as B, C, N, O, and F were also
eliminated from consideration because of large errors due to uncertainties in mass ab-
sorption coecients. In oxidized systems, the oxygen was calculated by means of assumed
stoichiometry, but the resulting oxygen values were not included in the error histograms
because of their dependence on the cation determinations.
Figure 35 shows an error histogram for the rst-principles standardless analysis
procedure embedded in the National Institute of Standards and TechnologyNational
Institutes of Health Desktop Spectrum Analyzer (DTSA) x-ray spectrometry software
engine (Fiori et al., 1992). The error distribution shows symmetry around 0% error, but in
comparing this distribution with that for the conventional standards=ZAF procedure
shown in Figure 21, the striking fact is that the error bins are 10 times wider for the rst-
principles standardless analysis procedure. Thus, the 95% error range is approximately
50% relative rather than 5% relative.
The error distribution for a commercial standardless procedure based on the tted
standards approach is shown in Figure 36. This distribution is narrower than the rst-
principles standardless approach, but the error bins are still ve times wider than those of
the conventional standards=ZAF procedure, so that the 95% error range is 25%
compared to 5%. It must be emphasized that this distribution represents a test of only
one of the many implementations of standardless analysis in commercial software systems
and more extensive testing is needed.
Figure 31 Effect of ice buildup on the detector.
Copyright 2002 Marcel Dekker, Inc.
3. Using Standardless Analysis
Given the width of these error distributions for standardless analysis, it is clear that re-
porting composition values to three or four signicant gures can be very misleading in the
general case. At the same time, the error distributions show that there are signicant
numbers of analyses for which the errors are acceptably small. Usually, these analyses
involve elements of similar atomic number (e.g., Cr, Fe, and Ni in stainless steel) for which
the x-ray peaks are of similar energy and are therefore measured with similar eciency.
The user of standardless analysis must be wary that any condence obtained in such
analyses does not extend beyond those particular compositions. The most signicant er-
rors are usually found when elements must be measured involving a mix of K, L, and M
shell x-rays. The errors in the tails of the distributions are so large that the utility of any
such concentration values is limited in terms of solving problems.
Standardless analysis does have legitimate value. If a microhomogeneous material
with a known composition similar to the unknown specimen of interest is available, then
the errors due to standardless analysis can be assessed and included with the report of
analysis. For example, if the FeS system is to be studied, then the minerals pyrite (FeS
2
)
and troilite (FeS) would form a good test to challenge the standardless analysis procedure.
The analyst should never attempt to estimate relative concentrations merely by in-
specting a spectrum. There are simply too many complicated physical eects of relative
Figure 32 K-Family peaks from transition elements; E
0
20 keV.
Copyright 2002 Marcel Dekker, Inc.
excitation, absorption, and eciency of detection to allow a casual inspection of a spec-
trum. Standardless analysis incorporates enough of the corrections to allow a sensible
classication of the constituents of the specimen into broad categories:
Major: greater than 10 wt%
Minor: 110 wt%
Trace: less than 1 wt%
In the absence of a known material to test a standardless analysis procedure, it is
recommended to use broad classication categories instead of numerical concentration
Figure 33 L-Family peaks; E
0
20 keV.
Copyright 2002 Marcel Dekker, Inc.
Figure 34 M-Family peaks; E
0
20 keV.
Figure 35 Error distribution for the first-principles standardless analysis procedure embedded
in the NISTNIH Desktop Spectrum Analyzer x-ray spectrometry software system.
Copyright 2002 Marcel Dekker, Inc.
values, which may imply far more apparent accuracy than is justied and which may lead
to a loss of condence in quantitative electron-probe microanalysis when an independent
test is conducted.
III. MICROANALYSIS AT LOWELECTRONBEAMENERGY
A. What Is to Be Gained at LowBeamEnergy?
1. Lateral and Depth Resolution
Throughout the history of electron beam x-ray microanalysis, analysts have relied upon
the strong dependence of the electron range (R
0
) on the incident energy to control the
spatial resolution of analysis, which is critical in such problems as characterizing inclu-
sions in a matrix or a thin layer on a substrate:
R
0
kE
n
0
E
n
c
70
where k depends on matrix parameters [atomic number (Z), atomic weight (A), and
density (r)], E
0
is the incident beam energy, E
c
is the critical excitation energy (critical
ionization potential), and the exponent n is in the range 1.51.7. The conventional
energy range for quantitative electron beam x-ray microanalysis can be thought of as
beginning at 10 keV and extending to the upper limit of the accelerating potential, typically
3050 keV, depending on the instrument. The lower limit of 10 keV for the conventional
operating range is selected because this is the lowest incident beam energy for which there
is at least one satisfactory analytical x-ray peak excited from the K, L, or M shells for every
element in the periodic table that is accessible to x-ray spectrometry (Goldstein et al.,
1992c). Only H, He, and Li are excluded due to a lack of x-ray emission or, in the case of
Li, because the photon energy is so low (54 eV) that it is completely reabsorbed in the
Figure 36 Error distribution for a fitted standards standardless analysis procedure embedded
in the commercial x-ray spectrometry software system.
Copyright 2002 Marcel Dekker, Inc.
target or the components of the x-ray spectrometer. Electron-excited x-ray microanalysis
thus extends from Be (E
K
0.116 keV) to the transuranic elements. This elemental range
is based on establishing the minimum acceptable overvoltage as U E
0
=E
c
! 1:25. At
E
0
10 keV, this overvoltage criterion involves the use of K lines for 4 (Be) Z27 (Co)
L lines for 28 (Ni) Z67 (Ho), and M lines for 68 (Er) Z to the transuranics, as il-
lustrated by the periodic table in Figure 37, shaded according to the appropriate shell for
analysis with E
0
10 keV and U!1.25. This broad elemental coverage of electron beam
x-ray microanalysis is one of its most important features.
No strict denition exists as to what energy value constitutes the beginning of the
low-beam-energy microanalysis regime, but from the previous discussion, the low-energy
regime can be considered those beam energies below 10 keV, and especially below 5 keV.
Figure 38 shows the excitation range for several characteristic x-rays in a silicon matrix as
a function of incident beam energy from 0 to 10 keV, as calculated with the Kanaya
Okayama formulation of the excitation range (Kanaya and Okayama, 1972):
Rmm
0:0276A
Z
0:89
r
_ _
E
1:67
0
E
1:67
c
_ _
71
Near E
0
10 keV, the excitation range is substantially greater than 1 mm for the various
elemental lines plotted, whereas at 5 keV and below, the range for these x-rays is below
500 nm. The excitation range as a function of overvoltage is shown in Figure 39. The range
is less than 150 nm for AlK in Si for an overvoltage of 2 and diminishes sharply as the
overvoltage is reduced, so that for an overvoltage of 1.25, the excitation range is
approximately 2030 nm.
Figure 37 Selection of x-ray peaks available for analysis with E
0
10 keV for conventional EDS.
Note that elements in black squares either do not emit x-rays or else are too low in energy for prac-
tical detection.
Copyright 2002 Marcel Dekker, Inc.
Low-electron-beam-energy microanalysis below E
0
5 keV has been possible for
many years with scanning electron microscopes (SEMs) based on tungsten and LaB
6
lament electron guns. However, the relatively low brightness of these sources severely
limits resolution in the low-beam-energy regime. The emergence of the high-performance
FEGSEM which can still achieve nanometer-scale beam diameters at low energy, has
made it possible to actually make use of the potential analytical resolution given by
Eq. (71), and this has greatly increased interest in low-energy x-ray microanalysis. The
typical analytical FEGSEM as currently delivered is equipped with a semiconductor (Si
or Ge) energy-dispersive x-ray spectrometer with a limiting spectral resolution of 125
150 eV. Although useful analyses can certainly be performed with the semiconductor EDS,
the limited resolution of EDS (approximately 50100 times the natural x-ray linewidth)
and aspects of the physics of x-ray generation with low-energy excitation impose some
severe constraints. As the electron beam energy is reduced, x-ray microanalysis becomes
insensitive to certain elements because of these factors.
B. What Are the Negative Aspects of Low-Beam-EnergyAnalysis?
1. Limits of Detection
At high overvoltage U ! 3, experience with semiconductor EDS x-ray spectrometry has
shown that the concentration limit of detection, C
MDL
, for most elements in most matrices
is approximately 0.001 mass fraction in the absence of peak interference (Goldstein et al.,
Figure38 Range of excitation of various lines (AlK, FeL, and AuM) in a silicon matrix as a func-
tion of the beam energy, as calculated with the KanayaOkayama range.
Copyright 2002 Marcel Dekker, Inc.
1992d). The limit of detection depends strongly on the characteristic peak-to-brems-
strahlung background (P=B) and upon the characteristic x-ray peak counting rate (P)
(Ziebold, 1967):
C
MDL
3:29a
ntPP=B
1=2
72
where a is the zieboldOgilvie factor in the expression relating concentration to measured
intensity ratio [unknown sample to pure-element standard (Ziebold and Ogilvie, 1964); a
can be taken as approximately unity for general estimation purposes], n is the number of
measurements, t is the integration time per measurement, P is the peak counting rate, and B
is the background counting rate. As the overvoltage is lowered, both the peak counting rate
and the peak-to-background decrease. Figure 40a shows the experimentally measured
behavior of P=B and C
MDL
as a function of overvoltage for SiKa from pure silicon.
Figure 39 Range of excitation of AlK in silicon.
Figure 40 (a) Experimental measurement of the P=B for copper as a function of overvoltage and
calculation of the concentration limit of detection for SiKa in a matrix of similar atomic number
(e.g., Al). Detector: Si EDS (150 keV); 100 s integration. The vertical arrow designates the low-
keV SEM regime. (b) Calculated limit of detection (mass fraction), derived from experimental mea-
surements of P=B and peak counting rate, as a function of overvoltage, U. Si EDS (150 eV FWHM)
operated at 3040% dead time. Three different accumulation times (nt) of 100 s, 200 s, and 1000 s are
plotted. The vertical arrow designates the low-keV SEM regime.
"
Copyright 2002 Marcel Dekker, Inc.
Copyright 2002 Marcel Dekker, Inc.
Measurement conditions were 100-s accumulation times (nt product) with a beam current
and detector solid angle chosen so that the EDS system dead time was approximately 30%
(total spectrum) for a 30-mm
2
Si detector with a resolution of 150 eV (MnKa). As can be
seen in Figure 40a, operation at low overvoltage results in restrictions on the concentration
limit of detection. Although major (arbitrarily dened as C>0.1 mass fraction) and minor
(dened as 0.01 C0.1 mass fraction) constituents can still be detected with over-
voltages as low as U2 at an accumulation time of 100 s, trace constituents (dened as
C<0.01 mass fraction) cannot be reliably measured with U below 9.0. To reduce the
overvoltage requirement to U3 for trace levels above 0.001 mass fraction requires an
accumulation time of 1000 s, as seen in Figure 40b. A higher-resolution semiconductor
EDS detector (e.g., 130 eV at MnKa) would improve the situation by a factor of 1020%.
Any peak interferences would serve to make the detection situation much more un-
favorable.
2. Limitations Imposedby FluorescenceY|eld
When an energetic electron scatters inelastically with a bound atomic electron resulting in
the ejection of the atomic electron, the ionized atom is left in an excited state. The atom
returns to the ground state through processes that involve electron transitions. The
dierence in energy between the electron states can be released in the form of an x-ray
photon, or the energy can be transferred to another atomic electron, which is ejected as an
Auger electron. The fraction of ionizations that leads to photon emission is called the
uorescence yield, o, which is related to the Auger yield, a:
o a 1 73
For low-electron shells with low binding energy, the Auger process is strongly favored
over x-ray emission. For example, for carbon K shell ionization (E
K
0.284 keV), the
uorescence yield is 0.00198. Table 5 lists some typical values of the uorescence yield for
low-ionizaiton-energy shells. Examination of these data reveals that an important con-
sequence of having to select low-energy L and M shells to analyze for intermediate and
heavy atoms is that the uorescence yields are substantially lower than those of similar K
shells, a factor of 35 poorer for L shells and 1030 poorer of M shells. Thus, the x-ray
spectra of these elements with low-beam-energy excitation are likely to have characteristic
peaks with low intensity relative to the background (x-ray bremsstrahlung), a problem that
is further exacerbated by the poor resolution of the semiconductor EDS, which acts to
spread the available peak information over a wide range of background.
Table 5 Fluorescence Yields
Element Shell Edge energy (keV) o
C K 0.284 0.00198
Na K 1.080 0.0192
Mg K 1.303 0.0265
Si K 1.848 0.0603
Zn L 1.022 0.00736
Ga L 1.117 0.00875
Ge L 1.217 0.0103
Sm M 1.080 0.00133
Eu M 1.130 0.00137
Copyright 2002 Marcel Dekker, Inc.
The combination of low overvoltage and the low yield from L and M shell x-ray
peaks can result in spectra which look extremely unfamiliar to analysts who are used to
spectra excited in the conventional beam energy range. As an example, Figure 41 shows
the situation for (a) SiO
2
and (b) SiO excited with an incident beam energy of 5 keV and
measured with a Si EDS (150 eV FWHM at MnKa). Both O and Si are measured with K
lines, and peaks for both are prominent in the spectrum, with the O peak reduced relative
to the Si due to stronger absorption eects. By comparison, Figure 42 shows the spectrum
for GeO
2
. The GeL family of peaks (unresolved) is much reduced in intensity relative to
the OK and SiK peaks due to the relative uorescence yield dierences, despite the fact
that the overvoltage is actually greater for the GeL edge (E
LIII
1:217 keV) than the SiK
edge (E
K
1:848 keV). Similarly, Figure 43 shows the situation for CaF
2
excited with an
electron beam energy of 3 keV, which is below the CaK edge (4.038 keV). Note that the FK
peak is very prominent, but the Ca can only be detected through the CaL peak, which is a
factor of 10 lower relative to FK, due to both the lower uorescence yield from the L edge
and the increased absorption of the lower-energy CaL radiation.
An extreme example of the eects of excitation and uorescence yield is shown in
Figure 44, which presents the spectrum of BaCO
3
excited at 5 keV and measured with a Si
EDS (150 eV FWHM at MnKa). The BaM family of peaks is just visible above back-
ground. In fact, the series of BaM peak structures is partially a result of the BaM
absorption edges, which exist as discontinuities in the x-ray continuum (bremsstrahlung)
and which are broadened into peaklike features as a result of the action of the detector
broadening function.
An overall assessment of the impact of both low excitation and the uorescence yield
on elemental accessibility with a 5-keV incident beam is presented in Figure 45 and the
situation with 2.5 keV is shown in Figure 46. At 5 keV, a few intermediate and heavy
elements are inaccessible (black squares), whereas at 2.5 keV, the situation is much worse,
with extensive sections of the periodic table eectively lost to SEM=semiconductor EDS
measurement.
3. Quantitative Analysis at LowBeamEnergy
Quantitative analysis procedures for the low-beam-energy regime is a subject of some
controversy and ongoing development. The classic ZAF methods are based on a combi-
nation of empiricism and physical theory which were developed in the conventional high-
beam-energy regime and should not be expected a priori to work well in the low-energy
regime (Goldstein et al., 1992e). Very often, the analyst must employ x-ray peaks for which
there are little data on mass absorption coecients and ts to the available data are
subject to signicant error. The more recently developed frz quantitation methods have
the advantage that at least some of the database of experimental measurements on which
the algorithms are based were indeed measured with low incident beam energy for low-
energy photons, so that these methods should be inherently better for quantitation in the
low-energy regime. An example of the analysis of transition metal suldes at 5 keV with
conventional ZAF matrix corrections is given in Table 6. Signicant relative errors
(overestimates) are found in the measurement of the metal components; consequently, the
totals are anomalously high. By comparison, under conventional beam energy analysis
(20 keV), the relative errors are less than 2% for this system.
A particular problem in performing analysis at low beam energy is the actual
structure of the specimen being measured. Most specimens have a surface layer that is
dierent from the interior due, for example, to surface oxidation, a conductive coating
applied to minimize charging, or contamination. At high beam energy, this surface layer
Copyright 2002 Marcel Dekker, Inc.
Copyright 2002 Marcel Dekker, Inc.
Figure41 (a) SiO
2
excited with a beam energy of 5 keV and measured with a Si EDS (150-eV FWHM at MnKa). (b) SiO excited
with a beam energy of 5 keV and measured with a Si EDS (150 eV FWHM at MnKa).
Copyright 2002 Marcel Dekker, Inc.
Figure 42 GeO
2
excited with a beam energy 5 keV and measured with a Si EDS (150 eV FWHM at MnKa).
Copyright 2002 Marcel Dekker, Inc.
Figure 43 CaF
2
excited with an incident electron beam energy of 3 keV and measured with a Si EDS
(150 eV at MnKa).
Copyright 2002 Marcel Dekker, Inc.
Figure 44 BaCO
3
with minor Sr ($0.05) excited at 5 keV and measured with a Si EDS (150 eV at MnKa). The BaM and BaL
lines are marked. Note the low peak intensity for the BaM series; the BaL lines are not excited.
Copyright 2002 Marcel Dekker, Inc.
Figure 45 Selection of x-ray peaks available for analysis with E
0
5 keV for conventional EDS.
Note that elements in black squares are not sufficiently excited to be practically measured.
Figure 46 Selection of x-ray peaks available for analysis with E
0
2.5 keV for conventional EDS.
Note that elements in black squares are not sufficiently excited to be practically measured.
Copyright 2002 Marcel Dekker, Inc.
thickness is a small fraction of the excitation range, so the eect on quantitation is neg-
ligible. However, at low beam energy, the total sampling depth becomes so shallow that a
surface layer constitutes a signicant fraction of the analyzed depth. The specimen should
thus be considered as a case of special geometry, that of an overlayer of C (or oxide,
etc.) on a substrate that is actually the specimen composition of interest.
Rigorous testing for the accuracy of low-beam-energy microanalysis is similarly
aected by the issue of proper test specimens. The situation is dicult enough at high
beam energy, where the lateral homogeneity is the key issue. At low beam energy, the
measurement becomes profoundly sensitive to the surface region of both the specimen and
the standards, especially if these are dierent. Thus, any testing of quantitation procedures
can only sensibly proceed when a suitable series of test specimens and standards has been
produced that do not suer from limitations imposed by the quality of the surface.
C. What Future Developments Are Possible?
The physics of electron-excited x-ray production and propagation is determined by the
choice of beam energy, the element(s) to be measured, and the matrix composition. Once
the analysts room to maneuver is restricted through the choice of low beam energy, the
only possible route to improve the measurement process is to improve the x-ray spec-
trometry. From Eq. (72) and the spectra presented in Figures 4144, it is clear that im-
proving the spectral resolution will increase the peak-to-background ratio, thereby
improving the visibility of low-uorescence-yield peaks, reducing interferences, and low-
ering the limit of detection. Increasing the detector eciency (i.e., solid angle) and redu-
cing the pulse processing time will improve the peak counting rate, also contributing to
improving the limit of detection.
1. ImprovementstoWavelength-Dispersive Spectrometry
Wavelength-dispersive x-ray spectrometry, based on Bragg diraction of x-rays from a
crystal, is the oldest form of x-ray spectrometry and is commonly incorporated in the
instrument conguration referred to as the electron-probe x-ray microanalyzer. The WDS
is a high-resolution (10 eV or less), single-channel, narrow-bandpass spectrometer that
must be mechanically scanned to cover a range of x-ray wavelengths (energy). The higher
resolution of the WDS results in a higher P=B, yielding a lower limit of detection than
semiconductor EDS. Typically, the WDS limit of detection is 10
74
10
75
mass fraction.
Table 6 Low Beam Energy (5 keV) Analysis of Metal Suldes
True Analyzed Rel. error (%)
FeS (meteoritic troilite)
Fe 0.635 0.667 5.0
S 0.365 0.365 0
CuS
Cu 0.665 0.730 9.8
S 0.335 0.334 0.2
ZnS
Zn 0.671 0.732 9.1
S 0.329 0.320 2.7
Copyright 2002 Marcel Dekker, Inc.
The high spectral resolution has the added advantage of diminishing the likelihood of
spectral interferences, although interferences may still occur for certain combinations of
x-ray lines. However, the quantum eciency of the WDS is signicantly lower than the
EDS due to the relative ineciency of the diraction process. Consequently, beam cur-
rents of the order of 100 nA or more are normally used to achieve adequate peak counting
rates. Such current requirements are incompatible with small beam size, even with the
thermally assisted eld-emission gun. Additionally, when low-beam-energy excitation is
considered, another problem becomes signicant. Low-energy x-rays are emitted as a
result of electron transitions involving electron energy levels whose energy and population
can be altered due to chemical bonding eects. These chemical bonding eects are man-
ifested in the x-ray spectrum as shifts in the peak position and shape (Bastin and Heijligers,
1991). Because the WDS only detects a narrow band of energy, for an accurate mea-
surement of a peak that is subject to alteration due to chemical bonding eects, it is ne-
cessary to scan across the peak to accurately determine its position and the total intensity
under the peak above background. This requirement for scanning further reduces the
measurement eciency and increases the need for beam current. Thus, WDS at low beam
energy is normally performed under conditions that result in poorer spatial resolution than
EDS.
Two developments may lead to improved WDS eciency, which can benet low-
beam-energy x-ray microanalysis. First, the WDS is a focusing device, wherein the x-rays
diracted at various points along a curved crystal are brought to a focus within the en-
trance slit of a gas proportional detector placed on the focusing circle of the spectrometer
(Rowland circle). If the focused rays are allowed to diverge beyond the Rowland circle,
they form a spatially dispersed image of the peak (with a single focusing geometry
WDS). By placing an imaging detector array o the Rowland circle in the dispersion
plane, a wider energy window could be simultaneously measured, providing a direct image
of the full peak and adjacent background regions (Fiori et al., 1991). Second, the eciency
of the WDS can be improved by increasing its solid angle of collection. For low-energy
x-rays (01 keV), a parabolic reective x-ray optic has been described that gathers a wide
solid angle and presents a parallel beam to a at crystal diractor that scatters x-rays into
a large-window proportional gas detector. With this augmented WDS, limits of detection
as low as 10
5
mass fraction have been reported for light elements, such as C (Agnello
et al., 1997).
2. Microcalorimetry Energy-Dispersive Spectrometry
The ideal x-ray spectrometer would have the resolution of the WDS combined with the
energy-dispersive character of EDS over the useful analytical range, 100 eV10 keV. This
seemingly impossible requirement has actually been realized with the development of a
practical fast microcalorimeter EDS at NIST, Boulder (Wollman et al., 1998). In the
microcalorimeter spectrometer (see also Sec. III.F.4 of Chapter 3), the energy of an
individual photon is measured as the temperature rise in an absorbing target maintained
near 100 mK. The NIST microcalorimeter design incorporates critical features such as
the use of an absorber maintained in the normal conducting state and a temperature
sensor based on a superconducting transition sensor which both contribute to a faster
pulse response than earlier microcalorimeters. A novel superconducting quantum in-
terference device (SQUID) is used as a stable, fast, high-gain amplier. With regard to
analysis, the chief operating characteristics of the NIST microcalorimeter EDS are the
following:
Copyright 2002 Marcel Dekker, Inc.
1. Energy resolution <10eV. As an example of the energy resolution applied in a
dicult analytical situation, Figure 47 shows the separation of the BaLlines fromthe
TiK lines in BaTiO
3
achieved with the NIST microcalorimeter EDS and compares it
with the unresolved peaks in the spectrum obtained with a semiconductor EDS.
2. Limiting count rate (with beam blanking) $1 kHz.
3. Photon energy coverage: 250 ev10 keV. An example of this energy coverage is
given in Figure 48, which gives a spectrum for YBa
2
Cu
3
O
7
obtained with the
microcalorimeter EDS, showing L line detection for Cu, Y, and Ba and K line
detection for O. Carbon K shell x-rays at 0.282 keV have been successfully
measured, as shown in Figure 49.
The NIST microcalorimeter EDS represents the leading edge of a revolution in x-ray
spectrometry. In particular, microcalorimeter EDS technology should have a major and
immediate impact on the application of the FEGSEM to microanalysis at low beam
Figure 47 Comparison of spectra of BaTiO
3
obtained with the microcalorimeter EDS
(FWHM10 eV) and with the semiconductor EDS (FWHM140 eV).
Copyright 2002 Marcel Dekker, Inc.
Figure 48 Microcalorimeter EDS spectrum of YBa
2
Cu
3
O
7
; E
0
10 keV.
Figure 49 Microcalorimeter EDS spectrum of carbon; logarithmic scale. Note the carbon coinci-
dence peak, and the discontinuities in the background due to absorption in the spectrometer materi-
als (Al window thermal protection coating and AG absorber).
Copyright 2002 Marcel Dekker, Inc.
energy, for example, in the analysis of particles during defect review in simiconductor
processing. As an example, Figure 50 shows the spectrum obtained at 5 keV from WSi
2
with a 10-eV microcalorimeter EDS. The separation of the SiKa line from the WM family
is virtually complete, thus eliminating the severe interference encountered with semi-
conductor EDS. The high P=B ratios evident in the spectra in Figures 4750 will result in
limits of detection similar to WDS. Finally, the resolution of the microcalorimeter EDS
will make it possible to actually detect low-energy L and M lines that have low uores-
cence yield. As an example, Figure 51a demonstrates the detection of several of the low-
yield peaks of the titanium L family and the barium M family in BaTiO
3
, which can be
compared to Figure 51b, obtained with semiconductor EDS. With conventional EDS, the
TiL peaks are unresolved from the OK peak, and the BaM peaks, although visible, are
actually convolved with the BaM edge discontinuities in the background.
Figure 50 Microcalorimeter EDS spectrum of WSi
2
; E
0
5 keV.
Figure 51 (a) Microcalorimeter EDS spectrum of BaTiO
3
, low-energy region. Note detection of
BaM-family members; E
0
10 keV. (b) Si EDS spectrum of BaTiO
3
, low-energy region. Note detec-
tion of BaM-family members; E
0
10 keV.
"
Copyright 2002 Marcel Dekker, Inc.
Copyright 2002 Marcel Dekker, Inc.
IV. ANALYSIS OF SAMPLES WITHNONSTANDARD GEOMETRIES
As mentioned earlier, in classical electron-probe microanalysis, the samples are polished
at, homogeneous, and opaque with respect to the electron beam. Although these con-
straints on sample geometry enable the analyst to perform high-quality quantitative
analysis on a wide variety of samples, there are many samples which cannot be prepared
with the classical geometry. These samples include, among others, layered samples, thin
sections from samples that have inclusions ore phase domains smaller than the excitation
volume of a 1030-keV electron beam, particles, and samples with rough surfaces. To
obtain some degree of quantitative information from these samples, the analyst must
modify the analytical schemes outlined in the previous sections.
A. Quantitative Analysis of Layered Specimens
In the early days of electron microbeam analysis, characterization of specimens that were
compositionally nonuniform within the analytical volume (the volume excited by the
primary electron beam), such as particles, thin lms, or layered specimens, was either not
attempted or involved use of simple empirical correction procedures. For example, it was
noted that the thickness of thin lms (particularly single-element lms) on a common
substrate could be determined by measuring two or more lms of the same composition as
the unknown with dierent known thicknesses and then using the results to construct a
linear calibration curve [e.g., Sweeney et al. (1960), Bolon and Lifshin (1973)]. This pro-
cedure has reasonable accuracy as long as the lms are quite thin with respect to x-ray
absorption and electron penetration and if the standards substrate compositions are
identical or very similar in atomic number to those for the samples.
As microprobe instrumentation and correction procedures continued to develop, the
importance of considering the eect of surface coatings (such as surface oxide layers or the
thin conductive lms evaporated onto insulating samples) on analytical results obtained at
low accelerating potentials was noted [e.g., Sweatman and Long (1969)]. At the same time,
thin-lm analysis progressed by using variable accelerating potential analysis to estimate
the thickness of the lms and by using low accelerating potential analysis conditions to
analyze only the surface lm [e.g., Hutchins (1966), Colby (1968), Warner and Coleman
(1973)].
With the development of increasingly more accurate algorithms expressing the re-
lative production of x-ray intensity as a function of depth in the sample, it is possible to
derive theoretical correction methods that enable accurate quantitative analysis of thin-
lm or layered samples. Today, a number of laboratories are actively engaged in routinely
analyzing complicated layered specimens and in developing and rening the correction
procedures that enable the quantitative determination of both layer thicknesses and
compositions for such materials. A variety of through review papers have been recently
written on layered-specimen analysis [e.g., Pouchou and Pichoir (1991a), Packwood and
Remond (1992), Willich (1992), Pouchou (1993), Karduck, 1997]. One particularly de-
tailed and cogent treatment of the subject can be found in the textbook by Scott et al.
(1995).
The two most commonly used approaches to correct for electron scattering, electron
energy loss, and x-ray absorption in the quantitative analysis of thin lms and layered
specimens are those based on (1) analytical expressions of the relative production of x-rays
as a function of depth, f(rz) and (2) Monte Carlo modeling of electron scattering. In their
typical development, neither of these approaches predicts the contribution of secondary
Copyright 2002 Marcel Dekker, Inc.
x-ray uorescence, either by characteristic lines or the continuum. Several authors have
separately described corrections for secondary uorescence in layered specimens that can
be incorporated with either of the two approaches. The following subsections will outline
the current status of these approaches and give examples of the types of results that can be
obtained by their use.
1. Methods Based on Calculationof f(rz)
The most commonly used procedure used to analyze thin lms or layered specimens is to
measure the emitted x-ray intensity from the line(s) of interest in the sample and compare
them with the emitted intensities of those lines from one or more thick, polished standards.
In the simplest case, the standards employed are pure elements. In that case, when
uorescence by characteristic lines and the continuum can be ignored, the relative intensity
can be related to concentration and layer composition by the equation
k
A
I
thin
I
bulk std
C
A
_
rt2
rt1
f
0
A
rz expw
0
A
rz drz
_
1
0
f
A
rz expw
A
rz drz
_
_
_
_
1
74
where f
0
A
and f
A
are the x-ray depth distribution functions for the layer of interest in
the specimen and the bulk standard, respectively, w
0
A
and w
A
are the respective ab-
sorption factors, and rt1 and rt2 are the depths from the surface of the top and
bottom, respectively, of the layer of interest (i.e., rt1 0 for the topmost layer, rt2 1
for a bulk substrate). Only in a case in which the layer of interest is very similar in
average atomic number both to its underlying (and overlying) layer(s) and to the bulk
standard will f
0
A
and f
A
be similar. In most cases, f
0
A
needs to be modied from ex-
pressions developed for bulk specimens to account for the variable composition within
the excited volume of the electron beam. Similarly, w
0
A
and w
A
are seldom identical, and
w
A
for all underlying layers needs to be modied to account for the dierent degrees of
absorption in the various overlying layers. If the order, thicknesses, and compositions of
the various layers in the specimen are known, it is simple algebra to determine the
absorption factor [cf. Pichou and Pichoir (1991a), Scott et al. (1995)]. However, even if
the sequence of thickness and compositions were known, conventional parameteriza-
tions of f(rz) for bulk specimens cannot be used for layered specimens without sig-
nicant modication.
The fundamental dierence between f
0
A
rz and f
A
rz lies in the number and en-
ergy distribution of the electrons that scatter through the layer rz Drz; these, in turn,
depend on the average atomic number, distance, and thickness of the layer(s) that lie
above and below rz Drz. For example, at the top of a reasonably thick surface layer
(with respect to the electron beam depth of penetration) on a substrate of considerably
dierent Z, both the upward traveling and downward traveling electron components of
the f(rz) are similar for both f
0
A
rz and f
A
rz. At the bottom of a relatively thin surface
layer over a bulk substrate, the downward-traveling electron component of f
0
A
rz is
similar to a bulk specimen of the surface lm composition, whereas the upward-traveling
component is similar to a bulk specimen of the substrate composition. The situation gets
increasingly complex as the layers get thinner and=or more numerous.
A number of investigators have tried to come up with simplied expressions to ac-
count for the dierences in electron scattering between layered samples and thick speci-
mens, with varying degrees of success. Hutchins (1966) tried to modify the original
Copyright 2002 Marcel Dekker, Inc.
Castaing expression for bulk with expressions involving the initial slope of frz curves
and the backscatter coecient, whereas Colby (1968) modied Castaings expression with
functions relating to the dierences in the backscatter coecients, backsatter correction
factors, and stopping powers between lm and substrate compositions [cf. Scott et al.
(1995)]. Warner and Coleman (1973) modied Hutchins methods to apply it to biological
materials, and Duzevic and Bonefacic (1978) modied his procedure to measure alloy lm
compositions on elemental metal substrates. Oda and Nakajima (1973) modied Colbys
procedure and showed good results in analyzing AgCu alloys on FeNi substrates. None
of these early attempts, however, resulted in widespread application.
As more realistic models for frz for bulk specimens were developed, further at-
tempts were made to apply them to thin-lm and layered-specimen analysis. Reuter (1972)
modied the frz parameterization of Philibert (1963) to better t experimentally de-
termined frz curves and applied it to the determination of the thickness of elemental
lms on single-element and multielement substrates. His frz parameterization was a
function of the electron backscatter coecient, Z. In order to account for the eects of
variable average atomic number in the dierent layers, he introduced an eective back-
scatter coecient, Z
eff
, which varied between the limiting values of the backscatter coef-
cient of a bulk specimen of the composition of the lm, Z
F
, and the backscatter coecient
of the bulk substrate, Z
sub
:
Z
eff
Z
sub
n
rz
n
0
_ _
Z
F
1
n
rz
n
0
_ _ _ _
75
Where n
rz
=n
0
is the ratio of the number of electrons transmitted through layer rz com-
pared to the number of electrons passing through an innitely thin lm at the surface of a
sample and is expressed by
n
rz
n
0
expsrz or 1
4 10
4
Z
0:5
rz
E
1:7
0
76
whichever is larger (the Lenard coecient s 4:5 10
5
=E
1:65
0
; Z is the average atomic
number of the layer, E
0
is the accelerating potential in kiloelectron volts). As noted by the
author, this approximation accounts for dierences between the number of backscattered
electrons from lm and substrate, but not for dierences in their energy distributions.
Nevertheless, the author showed good agreement between calculated and experimental
results for elemental lms on substrates, both having a wide range in Z. A number of
investigators implemented various versions of this correction [e.g., Armstrong and Buseck
(1975) in their original frz expressions for thin lms and particles].
Love and Scott (1978) modied the square frz model of Bishop (1974) for
application to thin lms on substrates:
k
A
I
thin
I
std
1 expwrt
1 exp2wrz
m
77
where I
thin
and I
std
are emitted intensities from a thin lm and bulk standard of the same
material, t is the lm thickness, and z
m
is mean depth of x-ray penetration which Scott et al.
(1995) proposed should be expressed by the equation of Sewell et al. (1985b):
rz
m
rs
m
2:4 0:07Z 1:04 0:48Z= ln U
0
78
where, in turn, rs
m
is the mean electron range given by
Copyright 2002 Marcel Dekker, Inc.
rs
m
7:87E 6J
0:5
E
1:5
0
7:35E 7E
2
0
C
i
Z
i
=A
i
79
Bushby and Scott (1993) reported results of analyses of oxide lms on metallic substrates
using these corrections. The authors noted that this procedure is limited to analysis of
lms on substrates of similar atomic number, as the procedure does not correct for
dierences in electron scattering. This correction, however, is particularly easy to calculate
and is thus useful for back of the envelop calculations during experiment design.
The various forms of the modied, surface-centered Gaussian frz expression
[see Eq. (46)], as originally developed by Packwood and Brown (1981), have been widely
adapted for thin-lm and multilayer-specimen analysis. As described earlier, these
equations have the general form
frz g
0
expa
2
rz
2
g
0
f0 expa
2
rz
2
brz 80
with a variety of functions proposed for a; b; g
0
and f0. Packwood et al. (1987) noted
that for very thin lms on substrates, the eect of the lm(s) on the x-ray distribution
would be negligible and, thus, to a rst approximation, a frz expression for a sample of
the average atomic number of the substrate could be used for all layers. Similarly, for very
thick homogeneous lms on substrates (or lms analyzed at an accelerating potential low
enough so that few, if any, energetic electrons penetrate to the substrate), the frz ex-
pression for a sample of the average atomic number of the lm could be used for both lm
and substrate. The problem lies in the formulation of frz expressions for layers of in-
termediate thickness between the two extremes.
Packwood and co-workers (Packwood et al., 1987; Packwood, 1991; Packwood and
Remond, 1992) note that most expressions for the g
0
termin the Gaussian frz equationare
not functions of sample atomic number and so do not need to be modied for intermediate-
layer analysis. They further note that most expressions for the a term are only weakly de-
pendent on the atomic number of the specimen and that, in most cases, simply using an a
term calculated for a sample of the average atomic number of the substrate is sucient
(assuming most electrons in the analysis penetrate into the substrate). They propose two
alternate formulations of the b and f0 terms and one formulation of the backscatter
coecient, Z, to account for the change in electron scattering across layer boundaries:
b
specimen
b
layer
rd
1=2a
_ _
b
substrate
1 rd
1=2a
_ _
81
and
f0
specimen
f0
layer
rd
1=2a
_ _
f0
substrate
1 rd
1=2a
_ _
82
(where any layer of thickness greater than 1=2a is considered innite), or
b
specimen
b
layer
erf ard b
substrate
erfcard 83
and
f0
specimen
f0
layer
erf 2ard f0
substrate
erfc2ard 84
Z
specimen
Z
layer
erf 2ard Z
substrate
erfc2ard 85
(where the layer and substrate contributions are weighted under the electron distribution
curve). In these equations, d is the thickness of the layer, r is the density, and a is the term
Copyright 2002 Marcel Dekker, Inc.
in Eq. (80). Using Eqs. (83)(85), Packwood et al. (1987) [based on the original work of
Packwood and Milliken (1985)] proposed the following general expressions for the emitted
x-ray intensities for the various types of layered specimens:
I
surface layer
C
A
sqrtp
2a
_ __
g
0
exp
w
2a
_ _
2
erf ad
w
2a
_ _ _ _
erf
w
2a
_ _ _ _
g
0
f0 exp
w b
2a
_ _
2
erf ad
w b
2a
_ _
erf
w b
2a
_ _ _ __
86
I
buried layer
C
A
sqrtp
2a
_ __
g
0
exp
w
2a
_ _
2
erf ad
w
2a
_ _ _ _
erf ad
0
w
2a
_ _ _ _ _ _
g
0
f0 exp
w b
2a
_ _
2
erf ad
w b
2a
_ _
erf ad
0
w b
2a
_ _ _ __
expd
0
w
0
w 87
I
covered substrate
C
A
sqrtp
2a
_ __
g
0
exp
w
2a
_ _
2
1 erf ad
0
w
2a
_ _ _ _ _ _
g
0
f0 exp
w b
2a
_ _
2
1 erf ad
0
w b
2a
_ _ _ __
expd
0
w
0
w 88
compared to their expression for conventional thick, at specimens:
I
bulk specimen
C
A
sqrtp
2a
_ __
g
0
exp
w
2a
_ _
2
1 erf
w
2a
_ _ _ _
g
0
f0 exp
w b
2a
_ _
2
1 erf
w b
2a
_ _ _ __
89
where d is the mass thickness of the analyzed layer, d
0
is the mass thickness of the overlying
layer, w is the absorption term ( m
Aby specimen
csc c for x-rays in the analyzed area, and w
0
is the absorption term by the overlying lm(s).
Bastin and co-workers [e.g., Bastin et al. (1992, 1998), Bastin and Heijligers (1991)]
proposed modifying their PROZA formulation of the surface-centered Gaussian f(rz)
model to enable layered-specimen analysis in a manner similar to that done by Packwood
and co-workers. They noted, in agreement with Packwood et al. (1987), that the f(rz)
expression for a thin layer lies between the two extremes of a very thin layer, that has a
f(rz) expression consistent with a composition with the surface layer elements dissolved
in the substrate, and a very thick layer (with respect to electron penetration), whose f(rz)
is similar to a bulk specimen having the lms composition. Similar to Packwood et al.,
they proposed modifying the expressions for the a, b, g
0
, and f(0) terms in the f(rz)
equation for surface layers of intermediate thickness by empirical smoothing functions
accounting for the dierences in electron scattering among the dierent layers and sub-
strate:
a a
bulk layer
a
dissolved
a
bulk layer
exp1:5a
bulk layer
T 10
6
2
90
b b
bulk layer
b
dissolved
b
bulk layer
exp1:5a
bulk layer
T 10
6
91
g
0
g
0;bulk layer
g
0;dissolved
g
0;bulk layer
exp2:0a
bulk layer
T 10
6
2
92
f0 f0
bulk layer
f0
dissolved
f0
bulk layer
exp2:5a
bulk layer
T 10
6
2
93
Copyright 2002 Marcel Dekker, Inc.
where bulk layer refers to a bulk specimen with the same composition as the analyzed
layer, dissolved refers to the composition that would be calculated from an analysis as-
suming the elements in the surface layer(s) were dissolved in the substrate, and T is the
layer thickness (in mg=cm
2
). Unlike Packwood et al., Bastin and co-workers assume that
the composition of the surface layer(s) has no eect on the f(rz) expression for the
substrate. This latter assumption is certainly a source of error in cases of surface layers and
substrates of dissimilar average atomic number analyzed under conditions where a sub-
stantial fraction of the electron trajectories occur in the surface layers.
The f(rz) model of Pouchou and Pichoir [e.g., Pouchou and Pichoir (1985, 1987,
1991a), Pouchou (1993)] was initially designed to include the analysis of layered speci-
mens. As discussed earlier, the model is treated in two parts, each of which is expressed by
a parabolic function:
frz
1
A
1
rz R
m
2
B
1
94
frz
2
A
2
rz R
x
2
95
where R
m
is the depth at which f(rz) is at a maximum and R
x
is the x-ray range. These
terms need to be modied for layered specimens, along with F, the total integral under the
f(rz) curves, and f(0), the surface ionization potential. The expressions given by Pouchou
and Pichoir for these terms were briey discussed earlier and can be found extensively
derived in their articles. To account for layered samples of intermediate thickness, they
suggested modifying these expressions in order to calculate the f(rz) for a ctitious
homogeneous composite sample having an appropriate average atomic number to produce
electron scattering similar to that which occurs in the real multiphased sample. They utilize
a weighting law of the form
frz Nrz L
2
rz R
2
96
where L and R are the low and high rz intercepts of the f(rz) expression with L<0 and
N is a normalizing factor so that the area under the f(rz) curve equals F. The various
parameters are sequentially calculated using various estimates of L and R, and compo-
sitions of the dierent layers are then calculated by an iterative process. For the details of
this weighting procedure, see Pouchou and Pichoir (1991a).
Willich (1992) proposed modifying the weighting procedure of Pouchou and Pichoir
with a simple eective atomic number, Z
combination
, for a ctitious homogeneous sample
with the same electron scattering properties as the real multilayered specimen. Based on
the work of Willich and Overtop (1988), he proposed
Z
combination
Z
substrate
Z
film
Z
substrate
rT
film
2rz
max
_ _
0:65
97
where Z
substrate
and Z
lm
are the average atomic numbers of the substrate and lm, rT
lm
is the mass thickness of the lm, and rz
max
is the x-ray range.
In all of the above procedures, the layer and substrate thicknesses and composi-
tions are calculated by an iterative procedure. Typically, elements are assigned to each of
the layers; rst, estimates of the layer compositions are determined by running con-
ventional bulk corrections and normalizing to 100%. Next, the thin-lm expressions are
evaluated to determine the thickness at which the corrected k-ratio data would yield
concentrations summing to 100%. Finally, using this thickness for a new estimate of
layer compositions, followed by reiteration until convergence in both thickness and
composition is achieved.
Copyright 2002 Marcel Dekker, Inc.
If elements are present in more than one layer, it is necessary to either assign pro-
portionality of the element among the various layers on some basis or to collect data at
more than one accelerating potential and try either by trial and error or by simultaneous
solution of the f(rz) expressions at the various E
0
s. Public-domain and commercial
computer programs exist that utilize these corrections [e.g., Waldo (1988), Bastin et al.
(1992), Pouchou (1993)].
2. Methods Based on Monte Carlo Calculations
The Monte Carlo procedures that have been used by numerous investigators to calculate
electron deceleration trajectories in compositionally homogeneous bulk materials can be
readily adapted to simulate x-ray emission from multilayer samples. The algorithms em-
ployed inconventional Monte Carlocalculations for thick specimens [e.g., Joy (1995)] donot
have to be modied in any signicant way for thin lms or layered specimens. Only the
boundary conditions need to be changed. In a layered specimen, each time an electron
trajectory is calculated, the programneed to be changed. In a layered specimen, each time an
electron trajectory is calculated, the program needs to determine in which layer(s) the
electronis moving andthen use the appropriate scattering, emission, and absorption data for
the assumed composition of that layer. The nature of the equations used for scattering,
energy loss, ionization probabiliity, and so forth does not change. Unlike the case for the
analytical expressions for f(rz) described earlier, no assumptions need be made regarding
dierences inthe nature andamount of electronbackscattering fromthe various layers; these
are directly and independently calculated for each electron in the Monte Carlo method.
Thus, at least to a rst approximation, the analytical accuracy of thin-lm and layered-
specimen analysis using Monte Carlo methods is comparable to that for bulk specimens.
Calculations of x-ray emission from Monte Carlo calculations are made by de-
termining a f(rz) curve based on the number, directions (path lengths), and energies of
electrons that are calculated to pass through the layer rz D(rz):
frz
C
A
o
A
N
AV
A
A
_ __
E
c
E
0
nE; rzQ
A
E
drz
drs
_ _
drs
dE
_ _
dE 98
[e.g., Armstrong (1991)]. For a multilayer specimen, a series of f(rz) functions are cal-
culated, one for each element of interest in each layer. These are constructed by keeping
track of the arrays of summed electron path lengths through each rz layer corresponding
to the dierent array components of energy. For details of the procedures, consult Joy
(1995) or Armstrong (1991).
One of the rst groups to utilize Monte Carlo calculations for thin-lm analysis was
Bolon and Lifshin (1973). Utilizing the multiple scattering Monte Carlo method of
Curgenven and Duncumb (1971), they produced calculated calibration curves of relative
intensity versus thickness for thin elemental lms on elemental substrates (e.g., Au on Si)
and showed good agreement with experimental data. Kyser and Murata (1974, 1976) used
the single-scattering Monte Carlo model of Murata et al. (1971) to construct similar ca-
libration curves for both elemental and compound (alloy) lms on elemental and com-
pound targets. Using the examples of Mn
x
Bi
y
and Co
x
Pt
y
lms on SiO
2
, they were able to
show self-consistent agreement in both concentration and lm thickness between mea-
sured and Monte Carlo data.
Numerous groups more recently have used Monte Carlo calculations to predict
emitted x-ray intensities from a wide variety of layered samples, including thin surface
Copyright 2002 Marcel Dekker, Inc.
coatings, thin and thick multiple-layer structures, surface layers containing implanted
species, surface and depth diusion proles, and surface-layer analysis during ion sput-
tering [e.g., Amman and Karduck (1990), Amman et al. (1992), Karduck and Rehback
(1991), Armstrong (1991, 1993), Armigliato et al. (1996), Valamontes and Nassiopolos
(1996), Karduck (1997), Chan and Brown (1997a, 1997b)]. These studies have emphasized
the advantage of analyzing the same specimens at dierent analytical conditions (such as
electron beam accelerating potential and incidence angle) in order to elucidate their
structure. If there are no elements in common in the dierent analyzed layers, then each of
these measurements is an independent estimate of layer thickness and composition. If
there are elements that are present in more than one layer, then multiple measurements are
essential to deconvolve the composition and structure. Joy (1995) and co-workers and
Gauvin and co-workers (e.g., Hovington et al., 1997a,b; Drouin et al., 1997) have de-
veloped public-domain Monte Carlo programs that are easily adapted to the analysis of
layered specimens.
3. Correction for Secondary Fluorescence
All of the above discussion of quantitative analysis of layered specimens has only con-
sidered the case of electron excitation of inner-shell ionization resulting in production of
characteristic x-rays. However, characteristic x-rays can also be produced by uorescence
from any absorbed x-ray photons (characteristic or continuum=bremsstrahlung) having
energies greater than the inner-shell ionization potential. These uorescing x-rays can be
produced from any of the layers in the specimen. Because the range of energetic x-rays
typically greatly exceeds that of the beam electrons, x-rays may be uoresced in layers
(e.g., the substrate) that are considerably deeper than the electron range.
As noted above, uorescence by characteristic or continuum x-rays can contribute
up to 60% or 70% of the total x-ray emission from bulk specimens in extreme cases.
Typically, secondary uorescence contributes to 10% or less of the emitted intensity. In
the case of layered specimens, the amount of secondary uorescence will depend on the
thicknesses and relative positions of the layers containing the uoresced and uorescing
species. The typical x-ray mean free path between absorption=uorescence events is much
greater than the typical electron mean free path between inelastic scattering=inner-shell
ionization events. As a result, the relative contribution from secondary uorescence events
will be small if either the uorescing or uoresced x-ray line is from even a major element
in a very thin layer. The greatest contribution from secondary uorescence occurs for a
minor element in a thick layer from a major element in that or another thick layer or
substrate. Thus, in many applications, just as in the case of particles, correction for sec-
ondary uorescence can be safely ignored.
Several investigators have proposed corrections for uorescence by characteristic
lines and=or the continuum in layered specimens. Armstrong and Buseck (1985) derived a
rigorous, theoretically based general expression for a variety of layersubstrate conditions.
This expression requires numerical integration in multiple dimensions and is too lengthy to
show here. The reader is referred to the original paper for the details. Waldo (1991)
simplied these equations and adapted them for use in his public-domain thin-lm analysis
software (Waldo, 1988). Cox et al. (1979) developed a simplied characteristic uorescence
correction for multielement thin lms on multielements substrates. Their method was
based on the assumptions that the lm was thin with respect to the mean depth of x-ray
generation, and that the x-rays were generated isotropically from a point source located in
the substrate at a distance from the lmsubstrate boundary equal to the mean depth of
Copyright 2002 Marcel Dekker, Inc.
x-ray production. The details of the procedure can be found in the original paper and a
good summary can be found in the original paper and a good summary can be found in
Scott et al. (1995). Armigliato et al. (1982) incorporated this procedure into there thin lm
correction and reported improved results in selected cases. However, by its nature, it is
inappropriate for the analysis of thick lms on substrates or multiple uorescing layers.
More recently, Pouchou and Pichoir (1991b) proposed corrections for uorescence
by both characteristics lines and the continuum in layered specimens. The authors make
some simplifying assumptions regarding the depth distribution of the primary x-rays and
relative line intensities. Their derivation is beyond the scope of this book, and the reader is
once again referred to the original paper for the details. Their corrections are employed
in the commercial versions of their program. The continuum uorescence correction has
been integrated as well into the public-domain program of Walso (1988, 1991).
a. Analytical Results: Testing the Methods
X-ray emission analysis, particularly involving wavelength-dispersive detectors at
microprobe beam currents (i.e., 11000 nA) can produce thin-lm analysis results of very
high precision. Detection limits of less than 10% of an atomic monolayer can be measured
[e.g., Packwood and Remond (1992)]. Film thickness variations can be measured at a level
of 0.1 nm. Compositional variations of 0.2% for major and minor element concentrations
at a level of 100 ppm can be detected, under favorable conditions, in lms as thin as 10 nm.
(For examples of the measurement sensitivities and detection limits, see the review papers
of Pouchou and Pichoir (1991b), Packwood and Remond (1992), Willich (1992), Pouchou
(1993), and Karduck (1997)]. Thus, x-ray emission analysis provides an important
complementary tool to the more conventional techniques used in thin-lm and layered-
specimen characterization, such as Auger electron spectrometry (AES), Rutherford-
backscattering spectrometry (RBS), and secondary ion mass spectrometry (SIMS) [e.g.,
Cazaux (1992)] and, in certain circumstances, approaches or exceeds their detectability.
X-ray emission analysis has the potential to be a more accurate, routine technique
for quantitative determination of thin-lm=layered-specimen composition (and, in some
cases, layer thickness) than techniques like AES, RBS, and SIMS. However, a serious
problem in evaluating the comparative accuracy of thin-lm analysis with any of these
techniques is the lack of certied thin-lm standards (both in terms of composition and
thickness). Most of the papers cited above that have proposed quantitation schemes for
lm analysis or reviewed the applications of thin lm=layered specimen analysis showed
examples of the accuracy or consistency=plausibility of the technique by results of ana-
lyses of laboratory samples of unknown composition and thickness. In some cases, the
samples were analyzed and thicknesses were determined by another analytical technique.
In other cases, an expected range of composition and thickness was stated and the results
compared to that range. In no case that we have discovered, have thin lm or layered
specimens of known, certiable pedigree (i.e., having been analyzed and had thicknesses
measured by more than one analytical procedure that also included analyses of certied
reference materials for which limits of accuracy could be given) been used to evaluate the
proposed correction procedures. In most instances, the lm=layer composition is assumed
from the starting materials and the thickness may be measured by wet chemistry or RBS.
Many of the papers containing raw thin-lm and layered-specimen data that have been
used to formulate and evaluate the analytical procedures discussed above date back to
when RBS was a new technique with relatively large uncertainties and when evaporation
and sputtering techniques for thin lm manufacture did not have nearly the precision and
accuracy that exists today.
Copyright 2002 Marcel Dekker, Inc.
Although the necessary published data does not exist to make a precise evaluation of
the relative accuracy of the various proposed thin-lm analysis techniques, we present
several examples of published raw data from the literature processed through various of
the correction procedures we have discussed to give an idea of the variation of the pro-
cessed results. Three types of thin lm=layered specimen analysis are given.
Table 7 presents the results of thickness determinations for monoelemental thin lms
deposited on monoelemental substrates, taken from the raw data of Reuter (1972) and
processed through various thin-lm correction procedures. The procedures used for this
table are the f(rz) correction procedures of Reuter (1972), Pouchou and Pichoir (1991a)
[PAP], Bastin et al. (1986), Bastin and Heijlingers (1990) [both using implementation of
Table 7 Experimental Data of Reuter (1972) Processed Through GMRFilm and CASINO
Film and x-ray line: AlKa AlKa AlKa AlLa
Substrate: B B Be Fe Au Pt
Accelerating potential (keV) 12 12 4 30 30 30 30
Film thickness (mg=cm
2
)
Chemical measurement 14.5 4.4 4.4 73.0 73.0 73.0 77 193
Thin-film corrections
Reuter (1972) phi(PZ) 14.2 4.3 4.3 75.0 83.0 85.0 94 198
Monte Carlo (CASINO) 100 207
PAP w contin. fluor. corr. 13.0 4.2 4.2 81.3 79.0 73.5 109 233
PAP w=o contin. fluor. corr. 13.0 4.2 4.2 81.3 79.0 73.5 103 217
Bastin (1986) 12.3 3.7 4.4 79.2 80.4 79.9 102 224
Bastin (1990) 13.2 4.2 4.7 82.1 79.1 59.6 107 225
Packwood and Brown 11.6 3.9 4.8 72.5 74.7 74.2 (138) (303)
Mean 12.9 4.0 4.5 79.3 78.4 72.1 104 221
Std. dev. 0.9 0.2 0.3 3.9 2.2 7.5 3.7 9.8
% Sigma 6.8% 5.7% 6.3% 5.0% 2.8% 10.4% 3.6% 4.4%
Meanchem. 11.1% 8.2% 1.4% 8.6% 7.5% 1.2% 35.3% 14.6%
Film and x-ray line: CuKa CuLa AuLa
Substrate: SiO
2
Au SiO
2
Au Pt
Accelerating potential (keV) 30 30 10 10 15 18 22 30
Film thickness (mg=cm
2
)
Chemical measurement 83.0 83.0 83.0 83.0 26.0 26.0 26.0 26.0
Thin-film corrections
Reuter (1972) phi(PZ) 72.0 85.0 71.0 82.0 23.0 23.0 24.0 23.0
Monte Carlo (CASINO)
PAP w contin. fluor. corr. 89.2 82.4 77.6 86.4 26.4 24.5 26.6 26.1
PAP w=o contin. fluor. corr. 85.3 82.3 77.6 86.4 25.0 23.4 25.4 25.1
Bastin (1986) 82.9 89.1 78.1 93.0 26.5 24.9 26.6 26.0
Bastin (1990) 86.5 82.1 86.1 73.8 26.1 24.2 25.9 25.2
Packwood and Brown 84.7 96.0 80.8 95.9 30.6 26.5 26.8 24.1
Mean 83.4 86.4 80.0 87.1 26.9 24.7 26.3 25.3
Std. dev. 6.0 6.1 3.6 8.5 2.1 1.1 0.6 0.8
% Sigma 7.2% 7.1% 4.5% 9.8% 8.0% 4.6% 2.2% 3.2%
Meanchem. 0.5% 4.1% 3.6% 4.9% 3.5% 5.0% 1.0% 2.7%
Copyright 2002 Marcel Dekker, Inc.
Bastin et al. (1992)], Packwood (1991), and the Monte Carlo procedure of Gauvin and
co-workers (Hovington et al., 1997a, 1997b; Drouin et al., 1997) (CASINO). The results of
the Reuter procedure were taken from the article. The Monte Carlo results were de-
termined by developing calibration curves of k
el
versus thickness from simulations using
the public-domain CASINO program of Gauvin et al. The other procedures were tested
using a modied version of the public-domain GMRFilm program of Waldo (1991).
Examples are chosen to include cases of elements of very dissimilar atomic number as lm
and substrate, cases of both soft and hard x-ray lines, and cases of variation of accelerating
potential. None of the cases shown involve a large correction for uorescence either by the
continuum or by characteristic lines. The results show (with the exception of a few out-
liers) an agreement of about 1015% relative among the thicknesses calculated by the
various correction procedures and an agreement of about 510% relative between the
mean corrected value and the thickness determined by wet chemistry.
Table 8 shows the results of the next more complicated type of thin-lm analysis,
multielement thin lms on a multielement substrate of dissimilar average atomic number.
The data were taken from the work of Kyser and Murata (1974). The data were processed
through the same correction procedures as used for Table 7, except the Monte Carlo re-
sults of Kyser and Murata were used instead of the CASINO program and the early f(rz)
expression of Reuter was not evaluated. The alloy lms deposited on SiO
2
substrates were
of composition Mn
x
Bi
y
and Co
x
Pt
y
. To conserve space, only the Mn and Co concentra-
tions are shown in the Table 8; the Bi and Pt concentrations can be determined by dif-
ference from 100%. As was the case in Table 7, the PAP expression was evaluated with
and without a continuum uorescence correction; all other expressions were evaluated
without correction for continuum uorescence. The data in Table 8, indicates only a minor
dierence in calculated thickness and composition (typically 1%) due to continuum
uorescence in these cases. The agreement among the various corrections in the alloy
compositions is better than 5% relative; the agreement in thickness is better than 8%. The
agreement in the lm composition between the mean of the correction procedures and the
RBS measured value is better than 5%; the agreement in thickness is typically better than
15%, perhaps reecting the larger uncertainties in these early RBS thickness measure-
ments.
Table 9 gives the results of a still more complicated type of analysis, two multi-
element layers of dissimilar average atomic number over a monoelemental layer with an
atomic number signicantly dierent from both of the layers. The data were taken from
the work of Pouchou (1993). The data were processed through the same corrections as
employed for Table 8, except the STRATA data from the article is given instead of results
from the Monte Carlo expressions of Kyser and Murata. (STRATA, as described in the
article, is a commercial program that does a best-t determination of data collected at
multiple accelerating potentials using the PAP correction method.) The case studied is a
NlCr lm deposited on a layer of FeGdPt that has been deposited on Si. There are
small corrections for characteristic uorescence but no signicant continuum uorescence
eects. Data have been collected at three dierent accelerating potentials to show the eect
of electron range. In most cases, agreement among the various models for data collected
at a given E
0
exceeds that for the same correction procedure at varying E
0
s. Agreement in
both calculated composition and thickness among the various procedures at a given
accelerating potential is typically similar to those measured in Table 8. The agreement
between RBS and calculated compositions and thickness is typically better than that
observed in Table 8, possibly reecting improvements in the accuracy of RBS measure-
ments in this later work.
Copyright 2002 Marcel Dekker, Inc.
Close examinations of Tables 79 reveals dierences in the levels of agreement of the
various correction schemes, both with respect to the independent measurement data and
with respect to internal consistency (e.g., giving the same answer for data collected at
dierent accelerating potentials) that might be used to choose one procedure over another
as providing superior accuracy. However, due to the absence of independently analyzed
and certied reference standards as well as to the absence of comparable raw data col-
lected by more than one laboratory on these samples, no best correction scheme can be
unambiguously determined. The internal agreement of the various procedures, however,
is generally reasonably goodtypically at a level of about 5% relative. This level is ap-
proximately twice what might be expected for these types of materials if they were
Table 8 Experimental Data of Kyser and Murata (1974) Processed Through GMRFilm
MnBi lms
X115-3 X115-5 X115-7
CMn rt CMn rt CMn rt
RBS measurement 26.6 48.2 53.5 33.0 80.1 26.2
Thin-film corrections:
Monte Carlo (Kyser & Murata) 25.5 53.0 51.0 38.0 80.0 29.0
PAP w continuum fluorescence correction 26.5 52.1 52.1 37.2 80.0 28.6
PAP w=o continuum fluorescence correction 26.3 51.4 51.8 36.5 79.9 27.9
Bastin (1986) 26.8 49.0 52.7 35.4 80.5 27.6
Bastin (1990) 26.8 51.8 52.5 36.7 80.3 28.3
Packwood and Brown 26.4 53.5 51.7 37.6 79.7 28.5
Mean 26.4 51.8 52.0 36.9 80.1 28.3
Std. dev. 0.5 1.6 0.6 0.9 0.3 0.5
% Sigma 1.8% 3.0% 1.2% 2.5% 0.4% 1.8%
MeanRBS 0.8% 7.5% 2.9% 11.8% 0.0% 8.1%
CoPt Films
10-2 13-2 Mn lm Bi lm
CCo rt CCo rt rt rt
RBS measurement 19.4 49.0 9.6 33.6 40.1 48.6
Thin-film corrections:
Monte Carlo (Kyser & Murata) 18.5 53.0 10.1 40.0 42.0 54.0
PAP w continuum fluorescence correction 19.0 52.5 10.0 39.5 41.2 54.7
PAP w=o continuum fluorescence correction 18.6 51.7 9.7 39.1 40.1 54.1
Bastin (1986) 19.0 49.5 10.1 37.6 39.6 51.3
Bastin (1990) 19.0 52.3 9.9 39.0 41.3 54.3
Packwood and Brown 19.6 54.0 10.3 40.4 41.2 56.6
Mean 18.9 52.2 10.0 39.3 40.9 54.2
Std. dev. 0.4 1.5 0.2 1.0 0.9 1.7
% Sigma 2.1% 2.9% 2.0% 2.5% 2.2% 3.1%
MeanRBS 2.4% 6.5% 4.3% 16.9% 2.0% 11.5%
Note: Concentration in element weight %, lm thickness (rt) in mg=cm
2
. Analyses performed at 20 keV, with
psi 52.5. MnKa and CoKa, BiMa and PtMa lines measured. Films are on SiO
2
substrates.
Copyright 2002 Marcel Dekker, Inc.
Table 9 Experimental Data of Pouchou (1993) Processed Through GMRFilm
Layers on Si
Layer 1 Layer 2
E
0
CNi CCr T (nm) CFe CGd CPt T (nm)
RBS meas. 14.4 85.6 68.3 51.4 28.6 20.0 24.6
STRATA pgm 14.7 85.4 67.1 52.0 28.7 19.3 24.2
StrataRBS 2.1% 0.2% 1.8% 1.2% 0.3% 3.5% 1.6%
20 keV PAP 15.0 85.0 70.8 51.0 29.1 19.9 24.6
Bastin (1990) 15.2 84.8 72.2 51.4 29.3 19.3 26.3
Bastin (1986) 15.5 84.5 69.8 52.3 27.8 19.9 22.7
Packwood 16.1 83.8 73.1 50.6 30.4 19.0 27.2
Mean 15.5 84.5 71.5 51.3 29.2 19.5 25.2
Std dev. 0.5 0.5 1.5 0.7 1.1 0.5 2.0
% Sigma 3.1% 0.6% 2.0% 1.4% 3.7% 2.3% 7.9%
MeanRBS 7.3% 1.3% 4.6% 0.1% 1.9% 2.4% 2.4%
25 keV PAP 14.6 85.4 69.6 50.8 29.8 19.6 24.1
Bastin (1990) 15.4 84.6 68.3 51.8 28.9 19.6 22.5
Bastin (1986) 14.7 85.3 69.3 51.0 29.9 19.4 24.8
Packwood 15.5 85.5 67.4 49.9 31.3 19.4 25.0
Mean 15.1 85.2 68.7 50.9 30.0 19.5 24.1
Std dev. 0.5 0.4 1.0 0.8 1.0 0.1 1.1
% Sigma 3.1% 0.5% 1.5% 1.5% 3.3% 0.6% 4.7%
MeanRBS 4.5% 0.5% 0.5% 1.0% 4.8% 2.5.% 2.0%
30 keV PAP 14.3 85.7 68.8 51.1 29.3 19.6 24.2
Bastin (1990) 13.5 86.5 67.2 51.9 28.5 19.6 22.7
Bastin (1986) 14.4 85.6 68.3 51.2 29.4 19.4 24.4
Packwood 15.0 85.0 64.0 49.9 30.7 19.4 24.0
Mean 14.3 85.7 67.1 51.0 29.5 19.5 23.8
Std dev. 0.6 0.6 2.2 0.8 0.9 0.1 0.8
% Sigma 4.3% 0.7% 3.2% 1.6% 3.1% 0.6% 3.2%
MeanRBS 0.7% 0.1% 1.8% 0.7% 3.1% 2.5.% 3.2%
Means PAP 14.6 85.4 69.7 51.0 29.4 19.7 24.3
Std dev. 0.4 0.4 1.0 0.2 0.4 0.2 0.3
% Sigma 2.4% 0.4 1.4% 0.3% 1.2% 0.9% 1.1%
MeanRBS 1.6% 0.3% 2.1% 0.8% 2.8% 1.5% 1.2%
Bastin (1990) 14.7 85.3 69.2 51.7 28.9 19.5 23.8
Std dev. 1.0 1.0 2.6 0.3 0.4 0.2 2.1
% Sigma 7.1% 1.2% 3.8% 0.5% 1.4% 0.9% 9.0%
MeanRBS 2.1% 0.4% 1.4% 0.6% 1.0% 2.5% 3.1%
Bastin (1986) 14.9 85.1 69.1 51.5 29.0 19.6 24.0
Std dev. 0.6 0.6 0.8 0.7 1.1 0.3 1.1
% Sigma 3.8% 0.7% 1.1% 1.4% 3.8% 1.5% 4.7%
MeanRBS 3.2% 0.5% 1.2% 0.2% 1.5% 2.2% 2.6%
Packwood 15.5 84.8 68.2 50.1 30.8 19.3 25.4
Std dev. 0.6 0.9 4.6 0.4 0.5 0.2 1.6
% Sigma 3.5% 1.0% 6.7% 0.8% 1.5% 1.2% 6.4%
MeanRBS 7.9% 1.0% 0.2% 2.5% 7.7% 3.7% 3.3%
Note: Concentration in element weight %, layer thicknesses in nm. Continuum uorescence correction applied in
all cases. Lines analyzed: NiKa, CrKa, FeKa, GdLa, PtMa.
Copyright 2002 Marcel Dekker, Inc.
prepared and analyzed as bulk specimens. The level of agreement between the corrected
electron microbeam data and that determined by other techniques ranges from about 5%
to 15% relative and certainly includes a considerable component of uncertainty from the
comparative technique. In both cases, the level of agreement in the layer composition
appears to be somewhat better than the level of agreement in the layer thickness. The data
certainly suggest that electron microbeam analysis has the potential for determining thin-
layer compositions and thickness with comparable accuracy to the more conventionally
employed surface techniques, particularly when appropriate certiable standards be-
come available for correction procedure evaluation.
B. Quantitative Analysis of Thin Specimens at HighVoltages
The quantitative analysis of thin samples such as thin lms, thin sections of samples
that are inhomogeneous on a submicrometer scale, or small particles with dimensions less
than about 0.5 mm represents a unique class of analysis that must be considered separately
from conventional probe analysis. Unlike conventional electron-opaque samples, this
class of sample is best analyzed in the analytical electron microscope by a high-energy
electron beam with an accelerating potential in excess of 100 kV. If the sample is thin
enough, there is minimal electron backscattering and the electrons lose very little of their
energy in the sample (Goldstein, 1979). As a result, the atomic number correction k
Z
can
be neglected and the generated x-ray intensity, f
Drt
a
, from a characteristic line of element a
in an isolated thin lm of mass thickness Drt can be calculated from Eq. (99) according to
Williams and Goldstein (1991):
f
Drt
a
N
o
a
a
a
Q
a
A
a
_ _
C
a
Drt 99
where N is Avagadros number, C
a
is the weight fraction of element A, o
a
is the uor-
escence yield for the analytical line of element A, a
a
is the fraction of total K, L, or M line
intensity measured, A
a
is the atomic weight of element A, Q
a
is the ionization cross section
of the analytical line of element a, and t is the lm thickness.
In addition to a simplied generation expression, if the sample can be approximated
as an innitely thin lm, the eects of x-ray absorption and uorescence can be neglected.
This condition is often referred to as the thin-lm approximation and makes it possible
to express the relationship between the measured I
a
and the generated x-ray intensities for
element a by Eq. (100):
I
a
f
Drt
a
e
a
100
where e
a
is an eciency factor related to the overall eciency of the Si(Li) detector for the
detection of x-rays from element a (Goldstein et al., 1986). Because the determination of
sample thickness at each analytical point is impractical and the value of e
a
is not constant,
the analysis schemes for thin samples involve the measurement of elemental ratios in which
the relative concentration of one element to another can be expressed by the following
equation:
C
a
C
b
k
ab
I
a
I
b
_ _
101
The sensitivity factor approach to analysis is common to many analytical techniques and
the factor k
ab
in Eq. (101) is referred to as the CliLorimer factor or k
ab
factor (Cli and
Copyright 2002 Marcel Dekker, Inc.
Lorimer, 1975; Nasir, 1976). It is related to the generation and eciency terms for element
b ratioed to those for element a. In many analytical schemes, the analyst determines, from
binary standards, a set of k
ab
factors. Then, for a binary system containing unknown
concentrations of elements a and b, C
a
and C
b
can be determined from Eq. (100) and the
knowledge that C
a
C
b
1. In ternary and higher-order systems, the relative con-
centrations for the various elements can be determined if the k
ab
factors are known for
combinations of various binary compounds as shown in the following equations for a
ternary system:
C
a
C
b
k
ab
I
a
I
b
_ _
102
C
c
C
b
k
cb
I
c
I
b
_ _
103
C
a
C
b
C
c
1 104
In analytical electron microscopy, particularly in analyses related to geological speci-
mens, the convention is to express the k
ab
factors relative to silicon (i.e., k
aSi
). In this
form, they are referred to as simply k factors and are related to the k
ab
factors as shown
in
k
ab
K
aSi
K
bSi
K
a
K
b
105
Figure 52 shows the k
aSi
values from the work of Wood et al. (1984) and Schreiber and
Wims (1981).
In addition to the k factors relative to silicon, several researchers, particularly those
involved in metallurgy, have found it useful to report the k factors relative to iron. Table 10
lists the k
aFe
for the K and L lines of several elements.
It is also possible to calculate k factors from Eq. (99). A detailed discussion of the
equations used to calculate the various terms inEq. (99) canbe foundin Williams and Goldstein
Figure 52 Measured k
aSi
factors for thin-sample analysis from Wood et al. (From Wood et al.,
1984. Reproduced with permission of Blackwell Science Ltd.)
Copyright 2002 Marcel Dekker, Inc.
(1991) and Goldstein et al. (1986). Figures 53 and 54 show a comparison of the calculated and
measured k factors for K and L line radiation from various elements (Wood et al., 1984).
Sheridan (1989) has determined a large set of experimental k
aSi
factors for a wide
range of Ka, La, Ma lines. The measurements were performed on a series of sub-
micrometer particles ground from NBS multielement research glasses. After applying an
absorption correction, the experimental results were compared to theoretical k
aSi
factors
determined with several dierent ionization cross sections.
As the thickness of the sample increases, the electron transparency of the thin sample
or small particle decreases, eventually reaching a thickness for which the sample no longer
Table10 Experimental k
aFe
Factors for Various Elements
K Lines
Lorimer et al. (1976, 1977) McGill and Hubbard (1981) Wood et al. (1984)
Element (100 keV) (100 keV) (120 keV)
Na 2.46
Mg 1.23 0.08 1.16 0.96 0.03
Al 0.92 0.08 0.8 0.86 0.04
Si 0.76 0.08 0.71 0.76 0.004
P 0.77 0.005
S 0.83 0.03
K 0.79 0.77 0.86 0.014
Ca 0.81 0.05 0.75 0.88 0.005
Ti 0.86 0.05 0.86 0.03
Cr 0.91 0.05 0.90 0.006
Mn 0.95 0.05 1.04 0.025
Co 1.05 0.98 0.06
Ni 1.14 0.05 1.07 0.06
Cu 1.23 0.05 1.17 0.03
Zn 1.24 1.19 0.04
Nb 2.14 0.06
Mo 3.38 3.80 0.09
Ag 6.65 9.52 0.03
L Lines
Wood et al. (1984) Goldstein et al. (1977)
(120 keV) (100 keV)
Sr
a
1.21 0.06
Zr
a
1.35 0.01
Nb
a
0.90 0.06
Ag
a
1.18 0.06 1.04
Sn 2.21 0.07 2.39
Ba 2.18
W 2.43
Au 3.10 0.09 3.27
Pb 4.14
a
k Factors for these elements are for combined La and Lb lines.
Source: From Goldstein et al., 1986.
Copyright 2002 Marcel Dekker, Inc.
Figure 53 Calculated and measured k
aFe
factors for K line radiation from several elemental thin
films. Solid lines represent range of theoretical values. (From Wood et al., 1984. Reproduced with
permission of Blackwell Science Ltd.)
Figure 54 Calculated and measured k
aFe
factors for L-line radiation from several elemental thin
films. Solid lines represent range of theoretical values. (From Wood et al., 1984. Reproduced with
permission of Blackwell Science Ltd.)
Copyright 2002 Marcel Dekker, Inc.
conforms to the thin-lm criteria. Under these circumstances, corrections for x-ray
absorption should be included in the analysis scheme. According to Williams and
Goldstein (1991), the limits for the failure of the thin-lm criterion are an x-ray absorption
correction, A, <0.97 or >1.03 and=or a uorescence correction >5%. For these cases,
Eq. (101) can be expanded to include absorption and uorescence:
C
a
C
b
k
ab
I
a
I
b
_ _
F
abs
1
1 F
fl
_ _
106
where F
abs
is the absorption factor given in Eq. (107) and F
fl
is the ratio of the uoresced
to primary intensity, I
a
=I
0
, given in Eq. (108):
F
abs
m
r
_ _
a
sam
m
r
_ _
b
sam
_
_
_
_
_
_
1 exp w
a
rt
1 exp w
b
rt
_ _
107
I
a
I
0
C
b
o
b
r
a
1
r
a
_ _
A
a
A
b
m
r
_ _
b
a
E
c
a
E
c
b
_ _
lnU
b
lnU
a
_ _
rt
2
0:932 ln
m
r
_ _
b
sam
rt
_ _
sec c 108
where r
a
is the absorption jump ratio for element a and A
a
and A
b
are the atomic numbers
for a and b, respectively.
The absorption factor was derived by Goldstein et al. (1986) from the work of Tixier
and Philibert (1969) and Konig (1976). Williams and Goldstein (1991) calculated the
thicknesses of materials above which the absorption of x-rays was outside the 0.971.03
limits. The results of these calculations are given in Table 11.
As pointed out by Williams and Goldstein (1991), the eect of uorescence must be
investigated in cases where the concentration of the uoresced element is at a minor to
trace level, <10 wt%, in a sample composed mainly of the exciting element. In a study of
Fe10 wt% Cr (Nockolds et al., 1980), the uorescence correction was only about 5%.
The equations for the absorption and uorescence involve the measurement of the
lm or particle thickness and the calculation of the mass absorption coecients for ele-
ments a and b in the specimen. As the result, it is necessary to measure the thickness at
each analysis location and to calculate the concentrations in an iterative loop similar to
that used for the conventional ZAF schemes.
C. Quantitative Analysis of Particles with Energy-Dispersive X-ray
Spectrometry
Although some initial particle studies were done on the electron probe using wavelength-
dispersive spectrometers (WDS), the variablity and complexity of many particle
compositions coupled with the high currents necessary for WDS made elemental analysis
Table 11 Calculated Maximum Thickness of Thin Samples for
Which the Absorption Correction Is Less Than 3%
Element of interest=x-ray line Sample t
max
nm
AlKa NiAl 9
PKa Fe3P 22
MgKa, OKa MgO 25
Source: Data from Williams and Goldstein, 1991.
Copyright 2002 Marcel Dekker, Inc.
of particles by WDS dicult at best (Armstrong and Buseck, 1978). In addition, the use of
multiple spectrometers, each with a dierent view of the particle and, therefore, a dierent
particle geometry, as shown in Figure 55, limited the quantitative capabilities of the
technique. The evolution of the various quantitative procedures resulted in the separation
of particles into roughly three size categories that are based on the procedures and in-
struments used for analysis (see Fig. 56). The particles less than 0.5 mm in diameter are best
analyzed in the analytical electron microscope using the procedures described in the
previous subsection. The particles with dimensions below about 0.1 mm require very little,
if any, particle-geometry corrections for elements with atomic numbers greater than about
11 and can be analyzed as innitely thin lms using k
AB
factors. As the particle size in-
creases, approaching 0.5 mm and larger, the eects of particle geometry must be included in
the analysis procedure (Goldstein, 1979).
Once the particle dimensions are greater than approximately 0.5 mm, the particles are
too large to be analyzed in the AEM and are best analyzed in the electron-probe or
scanning electron microscope. From 0.5 to about 20 mm in diameter, particle geometry
may have a pronounced eect on the generation and emission of x-rays and can sig-
nicantly aect quantitative analytical results. Many dierent analytical schemes have
been developed for the analysis of the particles in this size range. Finally, when the dia-
meter of the particle is greater than about 20 mm, the particle is large enough with respect
to the electron beam than it is essentially a bulk sample and classical analytical corrections
apply. The remainder of this subsection is concerned with the quantitative analysis of
particles in the 0.510-mm size range.
Figure 55 Images of a particle taken at the angles for three different WD spectrometers.
Figure 56 Particle size ranges and associated analytical techniques.
Copyright 2002 Marcel Dekker, Inc.
In classical electron-probe analysis, schemes employing either a ZAF, bulk-sample
f(rz), or BenceAlbee approach, both sample and standard must be innitely thick with
respect to the penetration of the electron beam and have at surfaces. By controlling
sample dimensions and shape, the corrections for the interaction of the electron beam with
the sample and the subsequent x-ray emission can be calculated from simple geometric
relationships. In the quantitative analysis of particles, the shape and thickness of the
specimen often cannot be controlled or measured. the diculties in quantitative analysis of
particles result from three dierent eects that inuence the generation and measurement
of x-rays from these samples (Small, 1981).
The rst eect is the result of the nite size (mass) of the sample. The mass eect is
related to the elastic scattering of the electrons and is strongly aected by the average
atomic number of the sample. The mass eect is important when the sample thickness or
particle size is smaller than the range of the primary electron beam so that a fraction of the
beam escapes the sample before exciting x-rays. This is shown in Figure 57, which is a
Monte Carlo simulation of electron trajectories. The majority of the primary electron
trajectories terminate within the boundaries of larger sphere, but very few terminate within
the boundaries of the smaller sphere. As the size of the sample decreases, the mass of
material from which x-rays are excited drops and results in a reduction of x-ray intensities
from the specimen compared to a bulk sample of the identical composition. The mass
eect can be demonstrated by comparing the x-ray emissions from a bulk target to the
emissions from a particle of the same composition. This eect can be seen in Figure 58,
which is a plot of the BaLa x-ray intensity from particles normalized to the intensity from
a bulk material of the same composition plotted versus particle diameter. The energy of
the Ba x-rays is 4.47 keV, which is high enough so that the absorption eects are minimal.
The mass eect is demonstrated by the decrease in the intensity measured on the particles
compared to the bulk for particles less than 3 mm in diameter. The net result of not
Figure 57 Monte Carlo simulation of electron trajectories in a large and small particle. (From
Small, 1981.)
Copyright 2002 Marcel Dekker, Inc.
correcting for the mass eect in the analysis of particles less than about 3 mm in size will be
an underestimation of the composition for all elements analyzed in the sample.
The second eect that must be corrected for is the absorption eect. In the analysis
of most particles, the x-ray emergence angle and, therefore, the absorption path length
cannot be predicted accurately as it can for polished specimens. The magnitude of this
eect is largest when there is high absorption, as is typically the case for soft x-rays from
elements like Al or Si that have energies less than 2 keV. The dierence between the ab-
sorption path length in a particle and a bulk at sample can result in widely dierent
values of emitted x-ray intensities. For the particle shown in Figure 59, the path length A
B in the particle is less than the path length AC in the polished sample. The path length
A*D in the particle, however, is greater than the path length AC in the polished sample.
The result of the varying absorption path lengths is shown in Figure 60, which is the SiKa
x-ray intensity from particles normalized to the intensity from a bulk material of the same
composition plotted versus particle diameter. For these spherical particles, the absorption
path lengths in the particles are less than the bulk material, resulting in a higher emitted
x-ray intensity form the particles compared to bulk. This eect is detectable primarily for
the lower-energy x-ray lines, such as that of aluminum at 1.49 keV, which are highly ab-
sorbed. This increased intensity for the low-energy x-rays results in an overestimation of
the concentration for these elements and a corresponding underestimation of elements
with high-energy x-rays for which absorption is not signicant.
The third eect is caused by the uorescence of x-rays by either the continuumor other
characteristic x-rays. Because x-ray absorption coecients in solids are relatively small
compared to electron attenuation, the secondary x-ray uorescence occurs over a much
Figure 58 Normalized BaLa x-ray intensity plotted as a function particle diameter. (From Small,
1981.)
Copyright 2002 Marcel Dekker, Inc.
larger volume than the primary electron excitation. In bulk samples and standards, the
x-rays, for the most part, remain in the specimens. In the case of particles however, the
particle volume may be only a small fraction of the x-ray excitation volume. As a result, the
exciting x-rays will uoresce relatively few x-rays before leaving the particle. In those
samples where the uorescence is important, the eect of comparing a particle to a bulk
standard may be signicant. This eect is shown in Figure 61, which is a plot of the range for
NiKa x-rays causing uorescence of FeKa x-rays in a NiFe alloy. The net eect of not
correcting for the secondary excitation of x-rays in particle analysis is as follows:
Figure 59 X-ray absorption in a particle compared to the absorption in bulk material. (From
Small, 1981.)
Figure 60 Normalized SiKa x-ray intensity plotted as a function particle diameter. (From Small,
1981.)
Copyright 2002 Marcel Dekker, Inc.
1. An underestimation of the concentration for elements that have a signicant
contribution to their characteristic line or lines from excitation by other
characteristic x-rays
2. In the case of continuum uorescence, an underestimation of the concentrations
for all elements, particularly those with higher-energy lines, which are excited by
the higher-energy, longer-range continuum
1. Normalization
One of the simplest methods for the quantitative analysis of particles is to normalize to
100% the concentrations, determined with bulk standards, from a conventional procedure
such as ZAF Eq. (109) (Wright et al. 1963):
C
n
a
C
a
i
C
i
109
where C
n
a
is the normalized concentration of element a, C
a
is the concentration of a de-
termined from the analytical procedure, and
i
C
i
is the sum of the concentrations for all
elements in the sample.
The analyst, in selecting this method of correction, makes the assumption that x-ray
absorption and uorescence are the same for the particle as for a bulk specimen and that
the mass eect is the same for all elements. No elements can be determined by dierence;
that is, oxygen must be analyzed or calculated from an assumed stoichiometry. In addi-
tion, the analyst cannot determine, by obtaining an analysis total of less than 100%, the
presence of any undetermined elements. In practice, normalization of results is most
eective for the correction of the mass eect because the decrease in intensity as a function
of particle size is nearly the same for all elements. Figure 62 shows that the dierent
elemental curves merge together for particle diameters less than about 2 mm. Because this
Figure 61 Range of secondary x-ray fluorescence in a NiFe alloy. (From Small, 1981.)
Copyright 2002 Marcel Dekker, Inc.
procedure does not accurately compensate for the absorption and uorescence eects, the
most accurate results will be obtained on particle systems that meet the following con-
ditions: (1) systems for which all the analytical lines for the elements are above 4 keV
where the absorption eects are minimal; (2) if any of the analytical lines are below about
4 keV in energy, then the lines for all the elements should be as close together in energy as
possible so that the matrix absorption is approximately the same in all cases; (3) systems
for which there is no signicant uorescence.
Table 12 lists the results from the analysis of lead silicate glass particles. The rst set of
results are taken fromthe analysis of the PbMa line at 2.3 keV, which is close in energy to the
SiKa line at 1.74keV. Because the absorption and mass corrections are similar for these two
lines, the lead and silicon concentrations are in good agreement with the true values. The
second set of results were determined by analyzing the PbLa line at 10.6keV. In this case, the
two analytical lines have very dierent energies and the particle absorption eect is not similar
inmagnitude. As expected, the errors associatedwiththis analysis are considerablyhigher than
those associated with the PbMa analysis. The lack of an eective absorption correction by
simple normalization can also be seen in Figure 63, which shows the relative error distribu-
tions, by element, for the analysis of K-411 glass microspheres (Roberson et al., 1995) with a
bulk K-411 glass standard. The error distributions showthat the concentration of Mg with the
lowest energy x-ray, 1.25keV, is overestimated, Si at 1.74keV is centered at about zero error,
and the elements with harder x-ray lines (Ca, 3.7 keV and Fe, 6.4 keV) are underestimated.
2. Particle Standards
The analyst can use a conventional analysis scheme and substitute particle standards for
the normal polished standards (White et al., 1966). In this procedure, the assumption is
Figure62 Normalized x-ray intensities from several elements plotted as a function of particle dia-
meter. (From Small, 1981.)
Copyright 2002 Marcel Dekker, Inc.
that the particle eects, particularly the absorption eect, will be approximately the same
for the sample and standard. This assumption is reasonably valid, providing the sample
and standard are close in composition and shape, and the particle diameter is above about
2 mm. Below 2 mm, as shown in Figure 60, any dierence in size and shape between un-
known and standards will be critical because a small change in eective diameter will result
in a large change in x-ray intensity. Figure 64 shows the results of using particle standards
for the analysis of a series of K-411 glass microspheres. Figure 64a is the error distribution
for the normalized analysis of a population of spheres ranging in size from about 1 to
10 mm analyzed with a piece of bulk K-411 glass as the standard. The error distribution in
Figure 64a shows three distinct modes that correspond to the overestimation of under-
estimation of dierent elemental concentrations because of the lack of an eective ab-
sorption correction (see Sec. IV.C.1). In contrast, Figure 64b is the error distribution for
the normalized analysis of the same particle spectra used in Figure 64a, except a 5-mm
sphere was used as the standard rather than a piece of the bulk glass. The distribution in
Figure 64b has a single maximum, is centered on zero, and is narrower, 0.1, than the
distribution in Figure 64a.
3. Geometric Modelingof Particle Shape
The procedure for the quantitative analysis of particles by geometric modeling of particle
shape is based on Eq. (110), which denes the emitted x-ray intensity of element a from a
layer at depth z in the sample, I
0
a
rz (Armstrong, 1991):
I
0
a
rz f
a
rz exp m
a
csc Crz 110
where r is the density, f
a
rz is the generated primary x-ray intensity from a layer at depth
z in the sample, m
a
is the mass absorption coecient for element a, and C is the take-o
angle.
Table12 Analysis of Lead Silicate Glass K-229 by Normalization of ZAF Results
Analysis Si (wt%) % Error Pb (wt%) % Error
Analysis done with PbMa line (meets conditions)
1 0.155 10.7 0.620 4.6
2 0.144 2.9 0.643 1.0
3 0.136 2.7 0.658 1.2
4 0.138 1.1 0.653 0.5
5 0.127 9.0 0.675 3.8
6 0.170 22 0.588 9.5
7 0.137 2.5 0.657 1.1
Analysis done with PbLa line (does not meet conditions)
1 0.134 4.5 0.663 1.9
2 0.177 26.3 0.578 11
3 0.159 14 0.612 5.8
4 0.166 18 0.602 7.4
5 0.017 88 0.894 37
6 0.100 29 0.731 12.4
7 0.157 12.3 0.616 5.2
Note: Nominal composition: Si 0.140; Pb0.650.
Copyright 2002 Marcel Dekker, Inc.
Figure 63 Relative error distributions, by element, for the analysis of K-411 glass with a bulk K-411 glass stan-
dard.
Copyright 2002 Marcel Dekker, Inc.
Copyright 2002 Marcel Dekker, Inc.
Modifying Eq. (110) to take into account the particle absorption eect and the loss
of x-ray intensity due to the mass eect, this equation becomes
I
a
rz
_
a
2
rz
ya
1
rz
_
b
2
r
y;r
z
xb
1
r
y;r
z
f
a
r
x
; r
y
; r
z
exp m
a
gr
x
; r
y
; r
z
dr
x
dr
y
111
for particles, where gr
x
; r
y
; r
z
is the distance from the point of x-ray generation to the
particle surface in the direction of the detector and f
a
r
x
; r
y
; r
z
is the generated x-ray
intensity in the particle. Calculation of f
a
r
x
; r
y
; r
z
requires the determination of the
number of x-rays, DI of element a, product per electron path length Ds as shown in
Eq. (112), which is identical to Eq. (99) except Ds is substituted for D(pt):
DI
C
a
o
a
p
ia
N
A
a
Q
a
Ds
112
Introducing the integration limits for the particle volume and electron energy into
Eq. (112) yields Eq. (113) for the calculation of fr
x
; r
y
; r
z
:
fr
x
; r
y
; r
z
C
a
N
A
a
o
a
p
ia
_ _ _
E
c
E
0
_
2p
y0
_
2p
g0
nE; r
x
; r
y
; r
z
; y; g
drs=dV
Q
a
E
dE=drs
_ _
dg dy dE 113
where nE; r
x
; r
y
; r
z
; y; g is the number of electrons of energy E, scattering at angle y, the
angle relative to the beam axis, and at angle g, the azimuthal angle in plane normal to beam
axis, dE=drs is the mean electron energy loss over distance drs, and drs=dV is the distance
traveled by the electron going through the volume element dV at point x; y; z. The equation
for including secondary uorescence is complex and is given in Armstrong (1991).
The emitted x-ray intensity for element a, corrected for particle mass and adsorption,
can be determined from Eqs. (111) and (112) if the function dening the particle shape=
volume is known. For irregular particles, this is not the case and the method developed by
Armstrong and Buseck (1977, 1978) is based on the determination of a simple geometric
shape or combination of shapes such as a square or pyramid which denes the boundaries
of the particle of interest. The various particle eects are then calculated for the chosen
geometric shapes dening the shape of the particle. The procedure back-calculates the
x-ray production from the particle to an appropriate value for an innitely thick sample
for which all primary electrons remain in the sample and which has the same composition
as the unknown. The various mechanisms responsible for x-ray are corrected as follows:
1. Electron transmission: The amount of primary radiation lost as a result of electron
transmission through the particle and sidescatter requires an analytical expression
for fr
x
; r
y
; r
z
. This expressionis calculated fromthe frz expression developed
by Armstrong (1991) (see Sec. II.2 in this chapter). In the modied form, the
Armstrong frz expression can be used to calculate x-ray production as a
function of position within a particle by setting the integration limits of x; y, and z.
Figure 64 (a) Error distribution for the normalized analysis of a population of spheres ranging in
size from about 110 mm analyzed with a piece of bulk K-411 glass as the standard. (b) Error distri-
bution for normalized analysis of the same particle spectra used in (a), except a 5 mm sphere was used
as the standard rather than a piece of the bulk glass.
3
Copyright 2002 Marcel Dekker, Inc.
2. X-ray absorption: The correction for x-ray absorption is done by numerical
integration of Eq. (111), where x, y, and z are determined for the geometric
shape or shapes which dene the overall particle. The shapes which have been
included in the program include cube or rectangular prism, tetragonal prism,
triangular prism, square pyramid, cylinder, hemisphere, and sphere.
3. Atomic number: The expression of Duncumb and Reed (1968) can be used to
calculate the loss of x-rays as a result of electron backscatter. The standard
equation for bulk materials can be used because this correction is applied at a
point where the eects of particle shape have been removed.
4. Secondary uorescence: Armstrong and Buseck (1985) showed that the ratio of
the emitted, uoresced x-ray intensity to the primary x-ray for a particle relative
to the ratio for a thick polished sample of the same composition was about 0.5
for a 10-mm particle, 0.33 for a 5-mm particle, and 0.1 for a 1-mm particle. In
addition, they also found the ratio to be insensitive to beam energy and mostly
dependent on absorption of the excited and exciting x-rays.
Based on the series of polynomial ts to calculated particle data, Armstrong and
Buseck (1985) proposed Eq. (114) to calculated the ratio of the emitted, uoresced x-ray
intensity to the primary x-ray intensity for a particle relative to the ratio for a thick po-
lished sample of the same composition, R
f
:
R
f
A Bx Cx
2
114
where X 1 expm
bj
rr, A0.0260, B 1:1409 0:2012y; C 0:2471 0:2741y
0:01315y
2
; y w
ak
=m
bj
, r is the particle radius, r is the density, m
bj
is the mass absorption
coecient for the exciting element by matrix, and w
ak
is the mass absorption coecient for
the excited element by matrix.
Armstrong (1991) and Myklebust (1975) (see Figs. 57, 58, and 60) have also used
Monte Carlo procedures to study the generation and emission of x-rays from particles.
These calculations involve setting boundary limits for the Monte Carlo calculations based
on particle shapes. Armstrong has also proposed using Monte Carlo calculations based on
the particle shapes in his ZAF program to develop analytical procedures for the analysis of
the particles based on working curves or on a-factor approach (Armstrong, 1982, 1991).
For this procedure, the particles are assumed to be homogeneous and are analyzed
by scanning over the electron beam rather than positioning the beam at a single spot on
the particle. The required input includes the following:
Elements and lines analyzed
Beam voltage and takeo angle
Standards and compositions
Standard morphology (i.e., bulk polished, thin lm, or particle of a particular shape
as described earlier)
Size, estimated shape (as described earlier), an estimated density of the unknown
particle
Measured standard and unknown x-ray intensities for the analyzed elements
The program will also accept elemental k ratios. The performance of the geometric
modeling procedure is shown in Figure 65. Figure 65a is the relative error distribution for
particles analyzed with a thick-specimen ZAF procedure (results normalized to 100%).
The distribution is quite broad with the majority of the analyses falling between 30%
relative error. In contrast, Figure 65b calculated from the geometric-modeling program is
much narrower with almost the entire set of analyses falling within 10% relative error.
Copyright 2002 Marcel Dekker, Inc.
Figure 65 (a) Relative error distribution for particles analyzed with a thick-specimen ZAF proce-
dure (results normalized to 100%). (b) Relative error distribution for particle concentrations calcu-
lated from the geometric-modeling program. (From Armstrong, 1991.)
Copyright 2002 Marcel Dekker, Inc.
4. Peak-to-Background Ratios
A fourth method for the quantitative analysis of particles was developed by Small et al.
(1978) and Statham and Pawley (1978). This method, derived from work on biological
specimens by Hall (1968), is based on the following observation: To a rst approximation,
the ratio of a characteristic x-ray intensity to the continuum intensity of the same energy
for a at, innitely thick target is equivalent to the ratio from a particle or rough surface of
the same composition.
In the form of an equation, this observation can be expressed as
I
B
_ _
particle
I
B
_ _
bulk
115
where I is the background-corrected peak intensity and B is the continuum intensity for
the same energy window as the peak. It is assumed that the spatial distribution of char-
acteristic x-ray excitation is identical to the distribution for continuum x-ray excitation. As
a result, the eects of particle shape and size on measured x-ray intensity will be the same
for the continuum and the characteristic x-rays. It therefore follows that by taking the
ratio of the two intensities, the particle mass and absorption eects will cancel each other.
In the procedure developed by Small et al. (1978), Eq. (115) is rearranged and the
peak intensities for the particle or rough surface are scaled up to values similar to a bulk
material of the identical composition:
I
particle
I
bulk
I
particle
B
bulk
B
particle
116
The values of I
for each element in the unknown can then be used as input for a standard
quantitative analysis scheme. In practice, a bulk material will not exist for particle of
unknown composition and the value of B
bulk
at any given characteristic peak energy must
be estimated as part of the analysis scheme. In the procedure, B
bulk
is determined from
Eq. (117), and for each characteristic peak energy, it must be summed over all elements in
the unknown:
B
bulk
i
F
"
Z
bulk
F
"
Z
std
B
std
fw
bulk
fw
std
117
The F
"
Z terms describe the continuum intensity as a function of the concentration-
weighted average atomic number for the hypothetical bulk material and standards. For
the simplest case, Kramers relationship (Kramers, 1923), the rst part of Eq. (117) would
be
"
Z
bulk
=
"
Z
std
, where the current estimation of concentration from the iteration loop in the
ZAF procedure is used for the calculation of
"
Z
bulk
B
std
is the measured continuum in-
tensity for a given standard at the energy of the analyzed x-ray. For multielement stan-
dards, B
std
must be multiplied by the weight fraction of the element of interest. Once the
values for B
bulk
are determined, they can be used in Eq. (116) to obtain the rst estimates
for I
. A set of k ratios can then be calculated and used as input to the ZAF routine. The
set of concentrations from each iteration is used to calculate new values of B
bulk
and the
sequence is repeated until successive iterations agree within a predetermined limit.
Various mineral particles have been analyzed with the peak-to-background method,
FRAME P, (Small et al., 1979a). The results of these analyses are reported in Table 13
along with the result from the conventional ZAF routine, FRAME C (Myklebust et al.,
1979). In all cases, the analyses with the peak-to-background routine are within 10% and
usually better than 5% relative error of the stoichiometric values. In contrast, the errors
Copyright 2002 Marcel Dekker, Inc.
with the conventional ZAF routine range from 7.9% for S in ZnS to 47% for Mg and Si in
talc. In addition, standard deviations for individual measurements are less for the peak-to-
background routine than they are for the conventional ZAF routine.
In the analysis of particles, the assumption that the generation volumes for char-
acteristic and continuum x-rays are identical is only valid for particles larger than about
2 mm in diameter. Below this size, the anisotropic generation of the continuum results in a
signicantly dierent excitation volume for the continuum compared to the isotropically
generated characteristic x-rays (Albee and Ray, 1970). This eect is shown in Figure 66,
in which isotropic and anisotropic cross sections have been used to calculate peak-to-
background ratios from K-309 glass particles normalized to bulk glass. The composition
of K-309 is 7.9% Al, 18.7% Si, 10.7% Ca, 13.4% Ba, and 10.5% Fe. These plots show that
the introduction of an anisotropic cross section for the continuum results in signicantly
higher peak-to-background ratios for the smaller particles. As a result, it is necessary to
introduce a correction for anisotropic generation of the continuum for quantitative ana-
lysis with peak-to-background procedures.
In the peak-to-background method developed by Statham and Pawley (1978), the
peak-to-background ratio from a given element is compared to the ratio from a second
element:
C
a
C
b
f
ab
P=B
a
P=B
b
118
In this case, f
ab
is a correction factor that should have a minimal dependence on particle
size and can be calculated or determined empirically from standards. Table 14 lists the f
ab
Table13 Analysis of Mineral Particles by the Peak-to-Background Method
Talc (Mg
3
[Si
4
O
10
](OH)
2
)
Wt% Mg % Error
a
Wt% Si % Error
Nominal 19.3 29.8
Normalized FRAME C 19.4 0.5 29.7 0.3
PB FRAME P 18.5 4.0 29.7 0.3
FeS
2
Wt% S % Error Wt% Fe % Error
Nominal 53.4 46.6
Normalized FRAME C 52.7 1.3 47.3 1.5
PB FRAME P 52.9 0.9 46.4 0.4
ZnS
2
Wt% S % Error Wt% Fe % Error
Nominal 32.9 67.1
Normalized FRAME C 35.4 7.8 64.5 3.8
PB FRAME P 36.0 9.0 67.7 0.7
Note: The results are the average from seven analyses.
a
% Error {[(nom.calc.)]=nom.}6100.
Source: From Small, 1979a.
Copyright 2002 Marcel Dekker, Inc.
values for ve particles of K-961 glass, the composition of which is listed in Table 15.
These results show that the f
ab
values do not exhibit any noticeable trend with particle size
and the standard deviation of f
ab
is less than 8% relative for elements with a concentration
greater than 1 wt%. For the elements Mn and P, with concentrations less than 1%, the
standard deviation in f
ab
is small considering the relatively poor counting statistics.
Figure 66 Peak-to-background ratios from glass spheres normalized to bulk ratios showing the
effects of isotropic and anisotropic cross sections for continuum generation. (From Small, 1981.)
Copyright 2002 Marcel Dekker, Inc.
D. Analysis of Rough Surfaces
Results from the FRAME P analysis of rough surfaces show a similar improvements in
accuracy with the peak-to-background method compared to conventional ZAF. Table 16
lists the results of analyses on fracture surfaces of AuCu alloys and a Fe3.22% Si alloy.
A micrograph of one of the AuCu fracture surfaces is shown in Figure 67. The use of the
peak-to-background method for the analysis of rough surfaces can lead to signicant
improvement in accuracy and precision for the analysis of these types of samples.
V. SPATIALLYRESOLVEDX-RAYANALYSIS
A. xy Mapping
In spatially resolved x-ray analysis, the position of the electron beam on the sample is
couple to the output of an x-ray spectrometer. The x-ray spectrometer, in turn, is coupled
to the output of the display on a cathode-ray tube, photographic plate, or computer
memory such that when the beam is interrogating a point on the sample, the output from
x-ray spectrometer is displayed or stored at the corresponding point on the storage
medium. In this way, multiple analyses can be taken on a sample and the spatial
Table14 f
ab
Values for K-961 Microspheres
a
Diameter (mm) Mg Al Si P K Ca Ti Mn Fe
2 1 1.24 1.74 1.72 2.43 3.12 3.01 3.77 3.06
3 1 1.36 1.68 1.57 2.44 3.46 3.13 2.56 2.90
4 1 1.30 1.66 1.95 2.63 3.36 3.16 3.29 3.21
6 1 1.48 1.83 1.43 2.96 3.56 3.14 3.81 3.06
9 1 1.27 1.51 1.74 2.72 3.17 3.06 3.28 3.33
Ave. 1 1.33 1.69 1.68 2.64 3.33 3.10 3.34 3.11
s (%) 0 6.8 7.1 12. 8.3 5.7 2.2 15. 5.1
a
Na was not included in the results because of poor statistics and high ion mobility.
Source: From Statham and Pawsley, 1978.
Table15 Nominal Composition of Glass K-961
Element Wt%
Na 2.97
Mg 3.02
Al 5.82
Si 29.9
P 0.22
K 2.49
Ca 3.57
Ti 1.20
Mn 0.32
Fe 3.50
Copyright 2002 Marcel Dekker, Inc.
Table16 FRAME P Analysis of Fracture Surfaces
Actual
(wt%)
FRAME C (wt%)
(normalized) Rel. error
FRAME P
(wt %) Rel. error
SRM 482, 60% Au40% Cu
Location 1
Au 60.3 49.6 18 58.0 4
Cu 39.6 50.4 27 44.0 11
Location 2
Au 29.1 52 52.0 14
Cu 70.8 79 41.0 3.5
SRM 482,80% Au20%Cu
Location 1
Au 80.1 73.8 8 76.9 4
Cu 19.8 26.2 32 19.1 3.5
Location 2
Au 69.3 13 76.7 4.2
Cu 30.6 55 20.1 1.5
SRM 483, Fe3.22% Si
Location 1
Fe 96.8 97.0 0.2 100.0 3.3
Si 3.22 2.9 8.2 3.2 0.3
Location 2
Fe 96.4 0.4 97.7 0.9
Si 3.6 11 3.5 7.4
Source: From Small et al., 1979b.
Figure 67 Secondary electron image of a fracture surface of Standard Reference Material 482.
(From Small, 1979b.)
Copyright 2002 Marcel Dekker, Inc.
relationship between each analysis location is retained. The analyst can then manipulate
the stored data and construct an image of the sample based on its elemental composition.
One of the earliest methods of obtaining spatially resolved analysis was to deect the
analog scan of electron beam in a single direction, usually x, across the specimen surface
and simultaneously record and display the output from a WDS x-ray spectrometer on a
CRT. In this way, a line prole for the intensity of a chosen element can be overlaid on an
image of the sample. An example is shown in Figure 68.
The rst method of x-ray analysis to display elemental information in both the x and
y directions was the x-ray dot map or x-ray area scan (Cosslett and Duncumb, 1956).
Initially, this procedure, like the line scan, made use of the analog scan of the electron
beam. As the beam was scanned slowly over the sample, the x-ray intensity of a given
element as determined by a spectrometer was compared to a preselected threshold value. If
the x-ray counts are above the threshold value, then a signal is sent to the recording CRT
on the instrument, which records a full intensity spot at that location. The entire sample
area selected by the instruments magnication is scanned in this manner and the resulting
x-ray dot map is constructed. Figure 69 shows the result of such an analysis of grain-
boundary migration of Zn in a ZnCu system. In this procedure, only the presence or
absence of a given element, as determined by the preselected thereshold value can be
displayed, Minimal information is provided that reects the amount or concentration of
that element present. In addition to displaying black and white maps, the analyst could
produce color overlays of up to three dierent elements by separately photographing the
dot maps using red, green, and blue lters.
Most instruments are now equipped with digital beam control, where the electron
beam is deected such that it interrogates the sample in two-dimensional array consisting
of n6n spot analyses, where n is usually 16512 for x-ray mapping. In an instrument with
stage control, the electron beam can be static and the stage stepped in the array pattern.
Each beam or stage position in the two-dimensional array corresponds to a pixel on a map
Figure 68 Line profile of SKa x-ray intensity overlain on a target current image of a specimen
from a failed pressure vessel.
Copyright 2002 Marcel Dekker, Inc.
of the specimen. By storing WDS or EDS spectra at each position, the analyst can obtain a
full quantitative analysis at each pixel. Once the results from the analyses are stored in the
array, the variations in elemental compositions can be coupled to the gray-scale or color
output of a computer monitor. The completed array of pixels provides the analyst with a
spatially resolved map of the elemental composition across the specimen. In addition,
because the array is stored in the computer, the analytical information can be
postprocessed in several ways. For example, a row of pixels can be selected providing the
analysts with elemental x-ray line proles across the sample in any direction or pattern
desired. This information can be displayed as a simple line plot or can be used to construct
contour maps of a given x-ray distribution.
The uncertainty associated with a quantitative analysis obtained for compositional
mapping is essentially identical to the uncertainty associated with a conventional analysis
obtained at the same electron dose. For maps where high accuracy is required on minor
constituents, the dwell time per pixel and=or beam current must be suciently large to
provide the appropriate x-ray intensity in the peak of interest. Compositional maps under
these circumstances may require several hours to accumulate and care must be taken to
ensure instrument stability over the required time interval.
As previously mentioned, both EDS and WDS analysis systems can be used for
compositional mapping. WDS detection of x-rays, however, is superior to EDS detection
for compositional mapping, particularly for samples requiring the analysis of con-
stituents at the minor and trace levels (Marinenko et al., 1987). One of the most im-
portant advantages of WDS detection is that the pulse processing time for WDS is at
least an order of magnitude shorter than EDS. WDS detectors have a limiting count rate
Figure69 X-ray dot map of a grain-boundary migration in a ZnCu system. (From Piccone et al.,
1982.)
Copyright 2002 Marcel Dekker, Inc.
of about 10
5
counts per second (cps) compared with a limiting count rate of 10
4
for EDS
detectors operating at maximum resolution. In addition, the WDS systems use dirac-
tion crystals so that only a narrow energy band of x-rays are detected at any given time.
The energy is adjustable so that most of these counts are the characteristic counts from
the element of interest. This count rate compares to a limiting rate of about 10
2
10
3
cps
in the peak of interest on an EDS detector because the limiting count rate of 10
4
is
distributed over the entire energy range of the spectrum.
In addition to its higher limiting count rate, the WDS detector has a factor of 10
higher peak to background ratio than the EDS detector. Taking into account the higher
limiting count rate and peak-to-background ratio of the WDS detector, the WDS detector
has a lower detection limit of about 100 ppm compared to a lower limit of 1% or greater
for mapping on the EDS system.
One limitation of the WDS system compared to EDS systems is in the mapping of
samples which contain a large number of elements of interest. In WDS systems, the
number of elements mapped during a given digital scan is limited to the number of WDS
spectrometers. Typically, in EDS systems, all elements, with characteristic x-ray energies
above about 1 keV can be detected simultaneously.
Two important aspects that must be considered in the interpretation of composi-
tional maps are the counting statistics and resulting uncertainties associated with the
various point analyses at each pixel location. Under normal mapping conditions, it is
impractical to obtain the level of accuracy associated with conventional probe analysis
where 100 s or more are used for data accumulation.
Marinenko et al. (1987) have shown that for a 1286128 map at a magnication
below 5006, each pixel represents an area of about 2 mm in diameter on the specimen
surface. This means that the spot size of the beam can be on the order of 1 mm in diameter
and carry a current of about 1 mA. At this current density, the limiting count rate of 10
5
cps can be obtained on pure-element samples. Assuming a dwell time per pixel of 0.1 s,
a 1286128 map would require 1600 s to accumulate and have 10
4
counts per pixel. From
Poisson statistics, the 1s counting uncertainty per pixel is 1%. In the case where mulite-
lement standards are required, the dwell time per pixel must be increased to obtain the
same counting statistics, as each element is present at less than 100%. In general, dwell
times of about 0.4 s are sucient for multielement standards and require less than 2 h per
1286128 map. As in conventional analyses, the standard maps can be archived and used
indenitely, provided the instrument conditions are constant.
In the analysis of samples where one or more of the constituents are at the minor
or trace level (1% or less by weight), digital maps may require 10 h or more to ac-
cumulate. A 10-h scan on a 1286128 map translates to 2 s dwell time per pixel. If the
element of interest is at the 1% level by weight, then the x-ray intensity in the peak
would be 2000 counts, assuming a limiting count rate of 10
5
cps as stated above. The
associated counting uncertainty is 2.2% for 1s. The actual uncertainty will be larger
than this due to uctuations in the intensity of the x-ray background, which must be
taken into account for quantitative analysis. In general, it should be possible to obtain
2s condence intervals of 10% or better for constituents at the 1% weight con-
centration in an unknown sample.
The power of compositional mapping can be seen in Figure 70, which is a compo-
sitional map of the same Zn grain-boundary migration shown in Figure 69. The brighter
areas on the map correspond to a Zn concentration of 10% by weight. The faint band
extending to the left in the bottom of Figure 70 corresponds to a Zn concentration as low
as 0.2% by weight.
Copyright 2002 Marcel Dekker, Inc.
One of the major diculties in obtaining quantitative x-ray maps with crystal
spectrometers is the defocusing of the diraction crystal as the electron beam is moved o
the crystal axis. The magnitude of the defocusing eect is proportional to the distance the
beam is from a point centered on the optical focus of the electron probe. The defocusing is
most severe at the low magnications of 2008006 routinely used in digital x-ray maps.
Corrections for the defocusing fall into four categories, (Marinenko et al., 1987).
The rst method to avoid spectrometer defocusing is to employ, as mentioned above,
the digital movement in the microscope stage rather than the electron beam. In this si-
tuation, the electron beam position remains static at the optical focus of the diraction
crystal and the sample is moved in a raster pattern under the beam. For WDS mapping, it
is critical that the sample remains at the focus of the optical microscope during the xy
movement of the stage. This procedure has been used successfully by Mayr and Angeli
(1985). The success of this method, since it involves the mechanical movement of the stage,
is dependent on the reproducibility of the stage motion. If the reproducibility of the stage
movement is poor, then the accuracy of the quantitative maps will be poor except for
extremely low magnications. In the more modern instruments, which have optical en-
coding and a guaranteed stage positioning of 0.1 mm, the results from moving the stage will
be comparable to the other corrections methods.
The second method used to correct for spectrometer defocusing is to move or rock the
diraction crystal in synchronization with the raster of the electron beam, (Heinrich, 1981e).
By slightly rocking the crystal, the entire sample area under study will be maintained at the
focus of the x-ray crystal. The problems associated with crystal rocking are similar to those
involving stage motion. Because beamrocking requires mechanical movement of the crystal,
the results are dependent on the reproducibility of the crystal rocking mechanism. This
procedure has been described for multiple spectrometer by Swyt and Fiori (1986).
The third correction method is to collect a series of standard maps in conjunction
with the collection of a map from an unknown. If the standard and the unknown are
Figure 70 Compositional map of the same Zn grain-boundary migration shown in Figure 69.
(From Piccone et al., 1982.)
Copyright 2002 Marcel Dekker, Inc.
collected under the same experimental conditions, then the standard maps can be used to
construct k ratios at each pixel location, thus normalizing out the eects of spectrometer
defocusing. Although this method does not require any mechanical movement, it does
require that the instrument remain stable over the period of time necessary to collect all
the maps from the standards and unknown. This is particularly important because a
pixel-by-pixel comparison of the unknown to the standard is made which requires a
constant crystal orientation with respect to the beam raster. Unless great care is taken
during an analysis, the use of archival standard maps is not recommended because of the
possibility of introducing large errors as a result of spectrometer drift. This procedure
was used by Marinenko et al. (1987) for the analysis of the AuAg standard reference
material 481 with the results shown in Figures 71 and 72. Figure 71 presents the un-
corrected maps from the alloys and the pure-silver standard, which show the char-
acteristic banding from the defocusing of the spectrometers. Figure 72 contains the pure-
silver map and the subsequent alloy maps that have been corrected for defocusing by
normalization to the standard map. The normalized maps show very little gradation or
structure in the image gray levels indicating the artifacts from defocusing have been
removed. Table 17 contains a comparison of the average composition obtained from the
digital map and the results from a conventional point analysis. In all cases, the average
composition from each map compares quite well with the conventional point analysis as
well as with the NIST certied values (Marinenko et al., 1987).
The fourth method used is to model the spectrometer defocusing. Marinenko et al.
(1988) observed that the defocusing artifacts (i.e., banding) were equivalent to an intensity
prole from a wavelength scan across the elemental peak of interest. This is shown in
Figure 73, which compares a wavelength line scan across the CrKa
1,2
peaks to a line prole
taken from a Cr map perpendicular to the defocusing bands. The similarity of these two
traces implies that the intensity can be measured in the center band of a map and, from a
Figure 71 Elemental maps from AuAg alloys showing the effects of spectrometer defocusing
(upper left) pure Ag, (upper right) Au20Ag80, (lower left) Au60Ag40, and (lower right) Au80
Ag20. The analytical line is AgLa. (From Marinenko et al., 1987.)
Copyright 2002 Marcel Dekker, Inc.
model of the line prole, the intensity at any other pixel location can be calculated. This
procedure eliminates the need for measuring a standard map and greatly reduces the time
required for obtaining a quantitative x-ray map.
Figure 74 is a schematic showing the relationship between the x-rays emitted from a
defocused point on the sample and the corresponding angular deviation from the line of
focus for a vertical crystal spectrometer. The electron probe is designed such that the
maximum x-ray intensity obtained on a crystal corresponds to the point A in Figure 74,
which is centered on the electron optic axis with the aid of an optical microscope. The
distance between point A and the crystal, S
0
, is
S
0
2R sin y 119
Figure 72 Elemental maps, corresponding to those in Figure 72, corrected for spectrometer defo-
cusing by normalization to a standard map. (From Marinenko et al., 1987.)
Table17 Quantitative Analysis of GoldSilver Alloys, SRM 481
Alloy Element line Certied value Point beam analysis
a
Digital mapping
Au20Ag80 AgL 77.58 77.33 (0.4%)
b
77.63
AuM 22.43 21.93 (2.6%) 20.81
AuL 22.43 22.69 (2.3%) 22.94
Au60Ag40 AgL 39.92 39.44 (0.5%) 39.96
AuM 60.05 59.59 (2.4%) 57.59
AuL 60.05 61.26 (2.3%) 60.87
Au80Ag20 AgL 19.93 19.59 (0.6%) 19.72
AuM 80.05 80.56 (2.5%) 77.26
AuL 80.05 81.09 (2.0%) 80.30
Note: Concentration are in weight %.
a
Excitation potential 20 kV, Faraday cup current 38 nA, point beam, ve randomly selected samplings
averaged for each alloy.
b
One relative standard deviation of a single measurement is in parentheses.
Copyright 2002 Marcel Dekker, Inc.
where R is the spectrometer radius and y is the Bragg angle. At a defocused point, point B,
which is a distance DS from the optic axis, the angular deviation from the exact Bragg
angle is Dy. Given that Dy is small, less than about 0.01 radians, which is the case for most
mapping applications, it can be approximated as
Dy sin Dy
DS
0
S
0
120
where DS
0
DSsin C for a vertical spectrometer and C is the x-ray takeo angle for the
electron probe.
Figure 73 Comparison of a WDS line scan across the CrKa peak to a line profile taken from a Cr
map perpendicular to the defocusing bands of the spectrometer. (A) Line scan across the CrKa peak
with a LiF crystal; (B) CrKa intensity profile from the compositional map, taken along a series of
pixels perpendicular to the maximum intensity band. (From Marinenko et al., 1980.)
Copyright 2002 Marcel Dekker, Inc.
In calculating the background value for a given pixel in the compositional map, it is
necessary to determine DS, which is the distance of the pixel from the maximum focus line
of the spectrometer. The orientation and equation of the line of maximum focus for a
given spectrometer can be dened by the coordinates of two pixel points [i.e., (x
1
; y
1
) and
x
2
; y
2
] which lie on the line. The orientation of the line can be determined from any raw
intensity map by increasing the image theshold as shown in Figure 75. The line has the
general form Ax By C 0 and it follows that DS for any pixel point, represented by
coordinates x
3
; y
3
, in the x-ray map can calculated from
DS
Dx
3
Y
3
E
D
2
1
1=2
121
where D y
2
y
1
=x
2
x
1
and E y
1
Dx
1
.
DS is positive for points located above the line of focus and negative for points below
the line of focus. DS as dened in Eq. (120) is dened in units of pixel elements which must
be converted to centimeters in order to obtain the linear distances for the calculation of Ds.
Figure 74 Relationship between the x-rays emitted from a defocused point on the sample and the
corresponding angular deviation from the line of focus for a vertical crystal spectrometer. (From
Marinenko et al., 1988.)
Copyright 2002 Marcel Dekker, Inc.
Figure 75 Line of maximum x-ray intensity on a crystal spectrometer determined by increasing
the image threshold. The threshold is low in the top image and high in the bottom image. (From
Marinenko et al., 1988.)
Copyright 2002 Marcel Dekker, Inc.
Copyright 2002 Marcel Dekker, Inc.
For this purpose, the scaling factor F is used, which is dened in terms of the magnication
M, the linear dimension of the CRT display L, and the number of pixel points in the
matrix N:
F
L
MN
122
Dy can then be expressed as
Dy Ds
0
F
S
0
_ _
123
The accuracy of this procedure is demonstrated in Figure 76. Figure 76A is an experi-
mentally measured intensity prole taken perpendicular to the focal axis and Figure 76B is
the modeled prole along the same axis. The intensity proles shown together in
Figure 76C are identical within the limitations of the experiment.
The use of standard maps or modeling to characterize spectrometer defocusing is
magnication limited. At magnications of about 1506 or less the deviation in intensity
from the edge to the center of a map is about a factor of 2 and cannot be accurately
Figure 76 Comparison of experimental and modeled x-ray intensity profiles for TiKa x-ray:
(A) Experimental x-ray profile for TiKa overlain on the corresponding Ti map; (B) modeled intensity
profile; (C) profiles from (A) and (B) together. (From Marinenko et al., 1988.)
3
Figure 77 Diagram describing the CCH.
Copyright 2002 Marcel Dekker, Inc.
corrected by these methods. For very low-magnication maps where quantitative results
are important, stage mapping or crystal rocking are preferred.
Recently, Mott et al. (1995) have developed a form of spectral imaging that couples
the measurement, by EDS, of x-ray photons emitted from a sample with the xy position
of the electron beam. Unlike conventional mapping techniques, where the electron beam is
stepped relatively slowly from point to point and the full data collected for each point, this
mapping procedure, known as Position-Tagged spectrometry, operates under digital
beam control with the beam rastered at a rapid scan speed. The result is the real-time
Figure 78 Compositional maps for an aligned CuTi eutectic: (A) Cu map; (B) Ti map. (From
Bright et al., 1988.)
Copyright 2002 Marcel Dekker, Inc.
collection of elemental images from preselected energy regions of interest, a frame-aver-
aged electron image, and x-ray spectra from preselected areas (i.e., phases) of the sample
where the spectrum is a summation of the spectra from all pixels within the area. The
spectrum image is compressed and streamed in real time to disk storage. At the completion
of the Positioned-Tagged Spectrometry (PTS) run, the spectrum image consists of
complete x-ray spectra for each pixel in the digital image. This enables the analyst to
construct additional elemental maps, or spectra as needed from the stored spectrum image.
In addition, the spectra for a given image can be processed by an appropriate quantitative
routine to obtain elemental concentration maps. It is important to note that the statistical
considerations for PTS compositional mapping are the same as those mentioned pre-
viously in this section and the time requirements for high-quality compositional maps will
be similar to those mentioned above.
B. Composition^Composition Histograms
The nal aspect of electron-probe analysis that will be considered is compositioncom-
position histograms. Compositional maps provide the analyst with a visual method of
interpreting the results from the elemental analysis of some 4000 individual points on a
sample (Prutton et al., 1987, Browning, 1987, Bright et al., 1988). By associating color and
intensity, the analyst is able to spatially relate the various elements and their compositions
within the analyzed region of the sample. One problem in the interpretation of compo-
sitional maps is that it is often dicult to visualize the compositional ranges and resulting
interelement correlations from a color composite image, particularly in the case of minor
or trace constituents. An alternative method of displaying the analytical information is in
the form of a compositioncomposition histogram (CCH) (Bright et al., 1988). The CCH
provides the analyst with an image that can be used to interpret the numerical relation-
ships among the various components in the sample.
Figure 77 is a schematic diagram describing the CCH. The concentration of element
a at each pixel in the compositional map of a, is associated with the corresponding pixel
Figure 79 CCH for the CuTi aligned eutectic shown in Figure 78. (From Bright et al., 1988.)
Copyright 2002 Marcel Dekker, Inc.
and concentration of element b in the map of b. This is represented by the top two blocks
in Figure 77. The associated concentrations for a and b at each pixel in the compositional
image are then plotted in the CCH, bottom block, as a single point in a scatter diagram.
Associations between elements are visible in the CCH as features, such as lines or areas,
which have a detectable density of points above background. In addition, the overlapping
of multiple-image pixels at one CCH point can be coupled to the intensity of the recording
CRT and displayed in an appropriate gray-level or color scale.
Figure 80 CCH and corresponding map areas for the high-Cu region of the eutectic: (A) outline
of the high-Cu region of the CCH; (B) pixels in the compositional map that correspond to the out-
lined area of the CCH. (From Bright et al., 1988.)
Copyright 2002 Marcel Dekker, Inc.
An example of the type of information conveyed by the CCH is shown in Figure
78A, 78B, and 79, which are the elemental maps and the CCH for an aligned CuTi eu-
tectic alloy. The CCH (Fig. 79) shows two lobes that have a high density of pixels and
correspond to the two dierent phases of the alloy (Bright et al., 1998). In addition to the
two lobes in Figure 79, there is a distribution of points between the lobes which has a
lower pixel density compared to the density in the lobes.
Figure 81 CCH and corresponding map areas for the region of the eutectic containing both Cu
and Ti: (A) outline of the connecting pixels between the high-Cu and high-Ti lobes; (B) pixels in
the compositional map that correspond to the outlined area of the CCH. (From Bright et al., 1988.)
Copyright 2002 Marcel Dekker, Inc.
A useful feature of the CCH is the ability to relate a given region or feature on the
CCH to the corresponding pixels in the compositional map. An example of this back
trace capability is shown in Figure 80 where the high-Cu lobe of the CCH is outlined
in Figure 80A and the corresponding pixels in the compositional map are highlighted as
the bright areas in Figure 80B. Applying the back trace method to the points between
the lobes (Fig. 81) allows the analyst to determine readily that the connecting line
between the two lobes corresponds to the boundary regions between the Ti and Cu
phases.
REFERENCES
Agnello R, Howard J, McCarthy J, OHara D. Microsc Microanal, 3(suppl 2):889, 1997.
Albee AL, Ray L. Anal Chem 42:1408, 1970.
Ammann N, Karduck P. In: Michael JR, Ingram P, eds. Microbeam Analysis1990. San Francisco:
San Francisco Press, 1990, p 150.
Ammann N, Lubig A, Karduck P. Mikrochim Acta 12(suppl):213, 1992.
Armigliato A, DeSalvo A, Rosa R. J Phys. D 1:529, 1982.
Armigliato A, Lewis T, Rosa R. Mikrochim Acta 13(suppl):241, 1996.
Armstrong JT. In: Heinrich KFJ ed. Microbeam Analysis1982. San Francisco: San Francisco
Press, 1982, p 175.
Armstrong JT. In: Newbury DE, ed. Microbeam Analysis1988. San Francisco: San Francisco
Press, 1988a, p 239.
Armstrong JT. In: Newbury DE, ed. Microbeam Analysis1988. San Francisco: San Francisco
Press, 1988b, p 469.
Armstrong JT. In: Heinrich KFJ, Newbury DE, eds. Electron Probe Quantitation. New York:
Plenum Press, 1991, p 261.
Armstrong JT. Microbeam Anal 2:S13, 1993.
Armstrong JT. Microbeam Anal 4:177, 1995.
Armstrong JT, Buseck PR. Anal Chem 47:2178, 1975.
Armstrong JT, Buseck PR. Development of a characteristic uorescence correction for thin lms and
particles. Proceedings of the Twelfth Annual Conference of the Microbeam Analysis Society,
Boston, 1977, p 42A.
Armstrong JT, Buseck PR. In: Electron Microscopy and X-ray Applications to Environmental and
Occupational Health Analysis. Ann Arbor, MI: Ann Arbor Press. 1978, p 211.
Armstrong JT, Buseck PR. X-ray Spectrom 14:172, 1985.
Bastin GF, Heijligers HJM. Quantitative electron probe microanalysis of carbon in binary carbides.
Technical Report, University of Technology, Eindhoven, 1985.
Bastin GF, Heijligers HJM. Scanning 12:225, 1990.
Bastin GF, Heijligers HJM. In: Heinrich KFJ, Newbury DE, eds. Electron Probe Quantitation.
New York: Plenum Press, 1991, p 145.
Bastin GF, Dijkstra JM, Heijligers HJM., X-ray Spectrom 27:3, 1998.
Bastin GF, Van Loo FJJ, Heijligers HJM. X-Ray Spectrom 13:91, 1984.
Bastin GF, van Loo FJJ, Heijligers HJM. Scanning 8:45, 1986.
Bastin GF, Dijkstra JM, Heijligers HJM, Klepper D. Mikrochim Acta 12(suppl):93, 1992.
Bence AE, Albee A. J Geol 76:382, 1968.
Bence AE, Holzwarth W. Proc. 12th MAS Conference, 1977, p 38.
Berger MJ, Seltzer SM. Studies of Penetration of Charged Particles in Matter. NRC Publication
1133. Washington, DC: National Academy of Sciences, 1964, p 205.
Bethe H. Ann Phys (Leipzig) 5:325 1930.
Bethe HA, Ashkin J. In: Segre E, ed. Experimental Nuclear Physics. New York: Wiley, 1953, p 252.
Copyright 2002 Marcel Dekker, Inc.
Bishop H. In: Castaing R, Deschamps P, Philibert J, eds. 4th International Congress on X-Ray
Optics and Microanalysis. Paris: Hermann, 1966, p 153.
Bishop HE. J Phys D 7:2009, 1974.
Bloch FZ. Zeit Phys 81:363, 1933.
Bolon RB, Lifshin E. In: Johari O, ed. SEM=1973. Chicago, IL: IITRI, 1973, p 281.
Bright DS, Newbury DE, Marinenko RB. In: Newbury DE. ed. Microbeam Analysis1988.
San Francisco: San Francisco Press. 1988, p 18.
Browning R. In: Joy DC, ed. Analytical Electron Microscopy. San Francisco: San Francisco Press,
1987, p 311.
Bushby RS, Scott VD. Mater Sci Technol 9:417, 1993.
Castaing R. PhD thesis, University of Paris, 1951.
Castaing R, Henoc J. In: Castaing R, Deschamps P, Philibert J, eds. 4th International Congress on
X-Ray Optics and Microanalysis. Paris: Hermann, 1966, p 120.
Cazaux J. Mikrochim Acta 12(Suppl):37, 1992.
Chan A, Brown JD. X-ray Spectrom 26:275, 1997a.
Chan A, Brown JD. X-ray Spectrom 26:279, 1997b.
Cli G, Lorimer GW. J Microsc 103:203, 1975.
Colby JW. Adv X-ray Anal 11:287, 1968.
Cosslett VE, Duncumb P. Nature 177:1172, 1956.
Cox MGC, Love G, Scott VD. J Phys D 12:1441, 1979.
Criss J. In: Heinrich KFJ, ed. Quantitative Electron Probe Microanalysis. NBS Special Publication
298. Washington, DC: Department of Commerce, 1968, p 57.
Curgenven L, Duncumb P. Tube Investments Res Lab, report #303, 1971.
Derian JD, Castaing R. In: Castaing R, Deschamps P, Philibert J, eds. 4th International Congress on
X-Ray Optics and Microanalysis. Paris: Hermann, 1966, p 193.
Drouin D, Hovington P, Gauvin R. Scanning 19:20, 1997.
Duncumb P, Shields PK. In: McKinley TD, Heinrich KFJ, Wittry DB, eds. The Electron
Microprobe. New York: Wiley, 1966, p 284.
Duncumb P, Reed SJB. In: Heinrich KFJ, ed. Quantitative Electron Probe Microanalysis. NBS
Special Publication 298. Washington, DC: Department of Commerce, 1968, p 133.
Duncumb P, Shields-Mason PK, Da Casa C. In: Mollenstedt G, Gaukler KH, eds. 5th International
Congress on X-Ray Optics and Microanalysis. Berlin: Springer-Verlag, 1969, p 146.
Duzevic D, Bonefacic A. X-ray Spectrom 7:152, 1978.
Fabre de la Ripelle M. J Phys (Paris) 10:319, 1949.
Fiori CE, Swyt CR, Myklebust RL. DTSA: Desktop Spectrum Analyzer and X-Ray Data Base.
NIST Standard Reference Datbase 36 version 2.5. Washington, DC: NIST, 1992.
Fiori CE, Wight SA, Romig AD. In: Howett DG, ed. Microbeam Analysis. San Francisco:
San Francisco Press, 1991, p 327.
Fiori CE, Newbury DE. In: Johari O, ed. Scanning Electron Microscopy. Chicago IL: ITTRI, 1978,
Vol I, p 401.
Gauvin R, Hovingtonand P, Drouin D. Scanning 17:202, 1995.
Goldstein JI. In: Hren JJ, Goldstein JI, Joy DC, eds. Introduction to Analytical Electron
Microscopy. New York: Plenum Press, 1979, p 83.
Goldstein JI, Williams DB, Cli G. In: Joy DC, Romig AD, Goldstein JI, eds. Principles of
Analytical Electron Microscopy. New York: Plenum Press, 1986, p 155.
Goldstein JI, Costley JL, Lorimer GW, Reed SJB, SEM 1:315, 1977.
Goldstein JI, Newbury DE, Echlin P, Joy DC, Fiori CE, Lifshin E. In: Scanning Electron
Microscopy and X-ray Microanalysis. New York: Plenum Press, 1981c, p 319.
Goldstein JI, Newbury DE, Echlin P, Joy DC, Fiori CE, Lifshin E. In: Scanning Electron
Microscopy and X-ray Microanalysis. New York: Plenum Press, 1981a, p 121.
Goldstein JI, Newbury DE, Echlin P, Joy DC, Fiori CE, Lifshin E. In: Scanning Electron
Microscopy and X-ray Microanalysis. New York: Plenum Press, 1981b, p 305.
Copyright 2002 Marcel Dekker, Inc.
Goldstein JI, Newbury DE, Echlin P, Joy DC, Fiori CE, Lifshin E. In: Scanning Electron
Microscopy and X-ray Microanalysis. New York: Plenum Press, 1981d, p 321.
Goldstein JI, Newbury DE, Echlin P, Joy DC, Fiori CE, Lifshin E. In: Scanning Electron
Microscopy and X-ray Microanalysis. New York: Plenum Press, 1981e, p 324.
Goldstein JI, Newbury DE, Echlin P, Joy DC, Fiori CE, Lifshin E. In: Scanning Electron
Microscopy and X-ray Microanalysis. New York: Plenum Press, 1981f, p 333.
Goldstein, JI, Yakowitz H, Newbury DE, Lifshin E, Colby JW, Coleman JR. In: Goldstein JI,
Yakowitz H, eds. Practical ScanningElectronMicroscopy. NewYork: PlenumPress, 1975a, p23.
GoldsteinJI, Yakowitz H, Newbury DE, Lifshin E, Colby JW, ColemanJR. In: GoldsteinJI, Yakowitz
H, eds. Practical Scanning Electron Microscopy. New York: Plenum Press, 1975b, p 338.
Goldstein JI, Newbury DE, Echlin P, Joy DC, Romig AD Jr, Lyman CE, Fiori C, Lifshin E. In:
Scanning Electron Microscopy and X-Ray Microanalysis. 2nd ed. New York: Plenum Press,
1992a, pp 6263.
Goldstein JI, Newbury DE, Echlin P, Joy DC, Romig AD Jr, Lyman CE, Fiori C, Lifshin E. In:
Scanning Electron Microscopy and X-Ray Microanalysis. 2nd ed. New York: Plenum Press,
1992b, p 21.
Goldstein JI, Newbury DE, Echlin P, Joy DC, Romig AD Jr, Lyman CE, Fiori C, Lifshin E. In:
Scanning Electron Microscopy and X-Ray Microanalysis. 2nd ed. New York: Plenum Press,
1992c, p 341.
Goldstein JI, Newbury DE, Echlin P, Joy DC, Romig AD Jr, Lyman CE, Lifshin E. In: Scanning
Electron Microscopy and X-Ray Microanalysis. 2nd ed. New York: Plenum Press, 1992d,
p 395.
Goldstein JI, Newbury DE, Echlin P, Joy DC, Romig AD Jr, Lyman CE, Fiori C, Lifshin E. In:
Scanning Electron Microscopy and X-Ray Microanalysis. 2nd ed. New York: Plenum Press,
1992e, p 503.
Green M, PhD thesis, University of Cambridge, 1962.
Green M, In: Pattee HH, Cosslett VE, Engstrom A, eds. 3rd International Congress on X-Ray
Optics and Microanalysis. New York: Academic Press, 1963, p 361.
Hall TA, In: Heinrich KFJ, ed. Quantitative Electron Probe Microanalysis. NBS Special Publication
298, Washington, DC: Department of Commerce. 1968, p 269.
Heinrich KFJ. In: Castaing R, Deschamps P, Philibert J, eds. 4th International Congress on X-Ray
Optics and Microanalysis. Paris: Hermann, 1966, p 1509.
Heinrich KFJ. NBS Technical Note 521. Present State of the Classical Theory of Quantitative
Electron Probe Microanalysis. Washington, DC: Department of Commerce, 1970.
Heinrich KFJ. In: Electron Beam X-ray Microanalysis. New York: Van Nostrand Reinhold, 1981a,
p 45.
Heinrich KFJ. In: Electron Beam X-ray Microanalysis. New York: Van Nostrand Reinhold, 1981b,
p 231.
Heinrich KFJ. In: Electron Beam X-ray Microanalysis. New York: Van Nostrand Reinhold, 1981c,
p 250.
Heinrich KFJ. In: Electron Beam X-ray Microanalysis. New York: Van Nostrand Reinhold, 1981d,
p 331.
Heinrich KFJ. In: Electron Beam X-ray Microanalysis. New York: Van Nostrand Reinhold, 1981e,
p 521.
Heinrich KFJ. In: Geiss RH, ed. Microbeam Analysis1987. San Francisco: San Francisco Press,
1987, p 24.
Heinrich KFJ, Yakowitz H. Mikrochim Acta 1:123, 1970.
Heinrich KFJ, Newbury DE, Myklebust RL. In: Newbury DE, ed. Microbeam Analysis1988.
San Francisco: San Francisco Press, 1988, p 273.
Heinrich KFJ, Yakowitz H, Vieth DL. The correction for absorption of primary x-rays. Proceedings
7th Nat Conference Electron Probe Analysis Society, San Francisco, 1972, paper 3.
Henoc J. In: Heinrich KFJ, ed. Quantitative Electron-Probe Microanalysis. National Bureau of
Standards Special Publication 298, Washington, DC: Department of Commerce, 1968, p 197.
Copyright 2002 Marcel Dekker, Inc.
Hovington P, Drouin D, Gauvin R. Scanning 19:1, 1997a.
Hovington P, Drouin D, Gauvin R, Joy DC, Evans N. Scanning 19:29, 1997b.
Hutchins GA. In: McKinley TD, Heinrich KFJ, Wittry DB, eds. The Electron Microprobe.
New York: Wiley, 1966, p 390.
Joy DC. Monte Carlo Modeling for Electron Microscopy and Microanalysis. New York: Oxford
University Press, 1995, p 216.
Kanaya K, Okayama S. J Phys D: Appl Phys 5:43, 1972.
Karduck P. Mikrochim Acta (suppl 15) 109, 1998.
Karduck P, Rehbach W. In: Newbury DE, ed. Microbeam Analysis1988. San Francisco,
San Francisco Press, 1988, p 227.
Karduck P, Renbach W. In: Heinrich KFJ, Newbury DE, eds. Electron Probe Quantitation.
New York: Plenum Press, 1991, p 191.
Konig R, In: Electron Microscopy in Mineralogy. Berlin: Springer-Verlag, 1976, p 526.
Kramers HA. Philos Mag 46:836, 1923.
Kyser DF, Murata K. IBM J Res Dev 18:352, 1974.
Kyser DF, Murata K. In: Heinrich KFJ, Newbury DE, Yakowitz H, eds. Use of Monte Carlo
Calculations in Electron Probe Microanalysis and Scanning Electron Microscopy. NBS
Special Publication 460. Washington, DC: Department of Commerce, 1976, p 129.
Laguitton D, Rousseau R, Claisse F. Anal Chem 47:2174, 1975.
Lenard P. Ann Phys (Leipzig) 56:255, 1895.
Lorimer GW, Cli G, Clark JN. In: Venables JA, ed. Developments in Electron Microscopy and
Analysis. London: Academic Press, 1976, p 153.
Lorimer GW, Al-Salman SA, Cli G. In: Missen DL, ed. Developments in Electron Microscopy and
Analysis. Institute of Physics Conference Series 36. Bristol: Institute of Physics, 1977, p 369.
Love G. In: Scott VD, Love G, eds. Quantitative Electron-Probe Microanalysis. Chichester: Wiley,
1983, p 175.
Love G, Scott VD. J Phys D 11:1369, 1978.
Love G, Scott VD. In: Newbury DE, ed. Microbeam Analysis1988. San Francisco: San Francisco
Press, 1988, p 247.
Love G, Cox MGC, Scott VD. J Phys D 9:7, 1976.
Love G, Cox MG, Scott VD. J Phys D 11:7, 1978.
Love G, Sewell DA, Scott AD. J Phys C 2(suppl 2):21, 1984.
Marinenko RB, Myklebust RL, Bright DS, Newbury DE. J Micros 145(pt 2):207, 1987.
Marinenko RB, Myklebust RL, Bright DS, Newbury DE. J Micros 155(pt 2):183, 1988.
Mayr M, Angeli A. X-ray Spectrom 14:89, 1985.
McGill RL, Hubbard FH. In: Quantitative Microanalysis with High Spatial Resolution. London:
The Metals Society of London, 1981, p 30.
Merlet C. Mikrochim Acta 114=115:363, 1994.
Mott RB, Waldman CG, Batcheler R, Friel JJ. In: Bailey GW, Ellisman MH, Henningar RA,
Zaluzec NJ, eds. Microscopy and Microanalysis 1995. New York: Jones and Begell, 1995,
p 592.
Murata K, Kotera H, Nagami K. J Appl Phys 54:1110, 1983.
Murata K, Matsukawa T, Shimizu R. Jpn J Appl Phys 10:67, 1971.
Myklebust RL. J Phys C 2(suppl 2):41, 1984.
Myklebust RL, Newbury DE. In: Newbury DE, ed. Microbeam Analysis1988. San Francisco:
San Francisco Press, 1988, p 261.
Myklebust RL, Newbury DE. In: Heinrich KFJ, Newbury DE, eds. Electron Probe Quantitation.
New York: Plenum Press, 1991, p 177.
Myklebust RL, Fiori CE, Heinrich KFJ. FRAME C, NBS Technical Note 1106, 1979.
Myklebust RL, Newbury DE, Yakowitz H. In: Heinrich KFJ, Newbury DE, Yakowitz H, eds. Use of
Monte Carlo Calculations in Electron Probe Microanalysis and Scanning Electron Micro-
scopy. NBS Special Publication 460. Washington, DC: Department of Commerce, 1975, p 105.
Nasir MJ. J Microsc 108:79, 1976.
Copyright 2002 Marcel Dekker, Inc.
Newbury DE. In: Bailey GW, Alexander KB, Jerome WG, Bond MG, McCarthy JJ, eds.
Microscopy and Microanalysis
0
98, Proceedings. New York: Springer-Verlag, 1998, p 194.
Newbury DE, Swyt CR, Myklebust RL. Anal Chem 67:1866, 1995.
Nockolds C, Nasir MJ, Cli G, Lorimer GW. In: Mulvey T, ed. Electron Microscopy and Analysis.
Bristol: Institute of Physics, 1980, p 417.
Oda Y, Nakajima K. Jpn Inst. Metals 37:673, 1973.
Packwood RH. In: Heinrich KFJ, Newbury DE, eds. Electron Probe Quantitation. New York:
Plenum Press, 1991, p 83.
Packwood RH, Brown JD. X-ray Spectrom 10:138, 1981.
Packwood RH, Milliken KS. CANMET Report PMRL=85-25(TR), 1985.
Packwood RH, Remond G. Scanning Microsc 6:367, 1992.
Packwood RH, Remond G, Brown JD. In: Brown JD, Packwood R, eds. 11th International
Congress on X-ray Optics and Microanalysis. London, Ontario: University of Western
Ontario Press, 1987, p 274.
Philibert J. In: Pattee HH, Coslett VE, Engstrom A. eds. X-ray Optics and X-ray Microanalysis.
New York: Academic Press, 1963, p 379.
Philibert J, Tixier R. In: Heinrich KFJ, ed. Quantitative Electron Probe Microanalysis. NBS Special
Publication 298. Washington, DC: Department of Commerce, 1968, p 53.
Piccone TJ, Butrymowicz DB, Newbury DE, Manning JR, Cahn JW. Scripta Met 16:839, 1982.
Pouchou J-L. Anal Chim Acta 283:81, 1993.
Pouchou J-L, Pichoir FJ. Microsc Spectrosc Electron 10:279, 1985.
Pouchou J-L, Pichoir FJ. In: Brown JD, Packwood R, eds. 11th International Congress on X-ray
Optics and Microanalysis. London, Ontario: University of Western Ontario Press, 1987, p 249.
Pouchou J-L, Pichoir F. In: Newbury DE, ed. Microbeam Analysis1988. San Francisco:
San Francisco Press, 1988, p 315.
Pouchou J-L, Pichoir FJ. In: Heinrich KFJ, Newbury DE, eds. Electron Probe Quantitation.
New York: Plenum Press, 1991a, p 31.
Pouchou J-L, Pichoir F. In: Heinrich KFJ, Newbury DE, eds. Electron Probe Quantitation.
New York: Plenum Press, 1991b, p 37.
Prutton M, El Gomati MM, Walker CG. In: Joy DC, ed. Analytical Electron Microscopy.
San Francisco: San Francisco Press, 1987, p 304.
Reed SJB. Br J Appl Phys 16:913, 1965.
Reed SJB. In: Electron Microprobe Analysis. Cambridge: Cambridge University Press. 1997, p 194.
Reuter W. In: Shinoda G, Kohra K, Ichinokawa T, eds. X-ray Optics and Microanalysis. Tokyo:
University of Tokyo Press, 1972, p 121.
Roberson S, Marinenko RB, Small JA, Blackburn D, Kauman D, Leigh S. In: Etz E, ed.
Microbeam Analysis1995. New York: VCH. 1995, p 225.
Russ JC. Quantitative analysis of sulphides and sulphosalts using an energy dispersive spectrometer.
Proceedings of the 9th Annual Conference of the Microbeam Analysis Society, 1974, paper 22.
Schrieber TP, Wims AM. In: Geiss RH, ed. Microbeam Analysis1981. San Francisco:
San Francisco Press, 1981, p 317.
Scott VD, Love G. In: Heinrich KFJ, Newbury DE, eds. Electron Probe Quantitation. New York:
Plenum Press. 1991, p 19.
Scott VD, Love G, Reed SJB. Quantitative Electron-Probe Microanalysis. 2nd ed. New York: Ellis
Horwood, 1995, p 234.
Sewell DA, Love G, Scott VD. J Phys D 18:1233, 1985a.
Sewell DA, Love G, Scott VD. J Phys D 18:1245, 1985b.
Sheridan PJ. J Electron Microprobe Technol 11:41, 1989.
Small JA, SEM 1:447, 1981.
Small JA, Newburry DE, Myklebust RL. In: Newbury DE, ed. Microbeam Analysis1979.
San Francisco: San Francisco Press, 1979b, p 243.
Small JA, Heinrich KFJ, Newbury DE, Myklebust RL. SEM 2:807, 1979a.
Small JA, Heinrich KFJ, Friori CE, Myklebust RL, Newbury DE, Dilmore MF. SEM 1:445, 1978.
Copyright 2002 Marcel Dekker, Inc.
Small JA, Newbury DE, Myklebust RL, Fiori CE, Bell AA, Heinrich KFJ. In: Heinrich KFJ,
Newbury DE, eds. Electron Probe Quantitation. New York: Plenum Press. 1991, p 317.
Springer G. Mikrochim Acta 3:587 1966.
Springer G. Neues Jahrb Mineral Monatshefte 9=10:304 1967.
Statham PJ, Pawley JB. SEM 1:469 1978.
Sweatman TR, Long JVP. J Petrol 10:33, 1969.
Sweeney WE, Seebold RE, Birks LS. J Appl Phys 31:1061, 1960.
Swyt CR, Fiori CE. In: Romig AD, Chambers WF, eds. Microbeam Analysis1986. San Francisco:
San Francisco Press, 1986, p 482.
Tixier R, Philibert J. In: Maurice F, Meny L, Tixier R, eds. Proceedings 5th International Congress
on X-ray Optics and Microanalysis. Berlin: Springer-Verlag, 1969, p 180.
Valamontes E, Nassiopoulos AG. Mikrochim Acta 13(suppl):605, 1996.
Waldo, RA. In: Newbury DE, ed. Microbeam Analysis1988. San Francisco: San Francisco Press,
1988, p 310.
Waldo RA. In: Howitt DG, ed. Microbeam Analysis1991. San Francisco: San Francisco Press,
1991, p 45.
Warner RR, Coleman JR. Micron 4:61, 1973.
White EW, Denny PJ, Irving SM. In: The Electron Microprobe. New York: Wiley, 1966, p 791.
Widdington R. Proc Roy Soc A86:360, 1912.
Williams DB, Goldstein JI. In: Heinrich KFJ, Newbury DE, eds. Electron Probe Quantitation.
New York: Plenum Press, 1991, p 371.
Willich P, Overtop D. Surface Interf Anal 13:20, 1988.
Willich P. Mikrochim Acta 12(suppl):1, 1992.
Willich P, Schimann K. Mikrochim Acta 12(suppl):221, 1992.
Wilson RR. Phys Rev 60:749, 1941.
Wollman DA, Irwin KD, Hilton GC, Dulcie LL, Newbury DE, Martinis JM. J Micros 188:196,
1998.
Wood JE, Williams DB, Goldstein JI. J Microse 133:255, 1984.
Wrigt FW, Hodge PW, Langway CG. J Geophys Res 68:5575, 1963.
Yakowitz H, Myklebust RL, Heinrich KFJ. NBS Technical Note 796. FRAMB: An On-line
Correction Procedure for Quantitative Electron Probe Microanalysis. Washington, DC:
Department of Commerce, 1973.
Zeller C. Cited by Ruste J, Gantois M. J Phys D Appl Phys 8:872, 1975.
Ziebold TO, Ogilvie RE. Anal Chem 36:322, 1964.
Ziebold TO, Anal Chem 39:858, 1967.
Copyright 2002 Marcel Dekker, Inc.
14
Sample Preparation for X-ray Fluorescence
Martina Schmeling
Loyola University Chicago, Chicago, Illinois
Rene E. Van Grieken
University of Antwerp, Antwerp, Belgium
I. INTRODUCTION
The preparation of samples presents, for any analytical method, one of the most important
steps for a reliable analysis. It is critical with regard to errors, but better precision as well
as lower detection limits can also be achieved by suitable sample pretreatment. The X-ray
uorescence (XRF) techniques [energy-dispersive XRF (EDXRF) and wavelength dis-
persive XRF (WDXRF)] allow the analysis of almost all types of sample with usually less
eort of sample preparation. However, for some materials, the sample preparation step is
even more time-consuming than the analysis itself; therefore, suitable procedures should
be developed and introduced, leading to reliable results.
Usually, the specimens are presented to the XRF spectrometer as solids, powders,
briquettes, fusion products, liquids, or lms, and for most of them, the pretreatment can
be kept to a minimum. Especially for metallic specimen, liquids and atmospheric particles
need only small eort. If required, possible sample preparation steps may include simple
procedures like cutting, grinding, milling, and mixing, or a combination of them. The
preparation step may also involve a conversion from the present state of the sample into
another one in order to receive a homogenous sample. It is important that the treatment
must be practicable for both samples and standards.
Metals and metallic alloys often need only to be polished to obtain a suitable surface
for analysis. Liquid samples, containing high enough concentrations for a direct analysis,
can be transferred into special cups or pipetted onto a surface target and dried to obtain a
thin lm. Atmospheric particles, collected onto lters, need practically no sample pre-
treatment. However, if sample preparation is required, as for geological and/or in-
homogeneous samples, the procedure should be fast and reproducible, require small eort,
and avoid contaminations. A consultation of the literature is very important before the
application of special sample preparation procedures because usually the problem oc-
curred already earlier and some proposals or possible solutions can be found. Extensive
sample preparation should always be avoided because with each additional step, the risk
of losses and contaminations increases.
Copyright 2002 Marcel Dekker, Inc.
A more extensive sample preparation is necessary when the sample is very in-
homogeneous, the surface layer is not representative for the whole sample, the average
particle size or the particle size distribution varies among the samples, and the amount of
porosity or pore size varies among samples.
As a condition for a reliable analysis by XRF, a at and even surface of the specimen is
required. Particularly for the determination of light elements, the surface should be prepared
mirrorlike, as scattering eects due to a rough texture inuence the results. An appropriate
surface can be normally obtained by polishing, melting, or fusion of the samples.
If absorption-enhancement eects are severe, an internal standard or a masking
agent must be added. Sometimes, it is also necessary to dilute the sample for the reduction
of such eects. For liquid samples with a low trace element amount, preconcentration or
tracematrix separation might yield satisfying results.
A rare application is still the analysis of gaseous samples. Kno chel et al. (1983) used
synchrotron-radiation XRF (SRXRF) for the determination of nitrogen in the presence of
small amounts of xenon. Special cups were designed for this purpose and the results were
quite promising. Therefore, it is surprising that this technique is not more frequently used
for the determination of elements in gases.
In the following section, the most common techniques of sample preparation are
discussed, with emphasis on more recent applications and developments. Sample pre-
paration for special techniques, like SRXRF, total-reection XRF (TXRF), photon-in-
duced x-ray emission (PIXE), and electron-probe microanalysis (EPMA), is only briey
mentioned in some cases, because in earlier chapters of this book, extensive relevant
discussions have already been given. It should be mentioned that several classical books
about XRF include chapters about conventional sample preparation procedures (Bertin,
1970; Bertin 1798; Jenkins et al., 1981; Tertian and Claisse, 1982; Jenkins, 1988; Injuk and
Van Grieken, 2000) and, recently, an entire book was dedicated to practical sample pre-
paration in XRF and x-ray diraction (Buhrke et al., 1998).
II. SOLIDSAMPLES
X-ray uorescence is a traditional analytical method for the analysis of solid samples.
They may be presented to the spectrometer as such, as powders, as briquettes, or as
fusion products. In principle, it is possible to determine the bulk composition of solid
samples directly if the element distribution inside the solid specimen is homogeneous in
all three dimensions and the specimen shows a satisfying at surface without scratches
and striations. Often such samples need only to be polished at the surface with suitable
abrasive materials and to be cut into the appropriate size for the sample holder of the
spectrometer. If the sample is too small for the sample holder, the positioning might be
critical, and it is useful to prepare a wax mold that ts into the holder. Such a mold
permits a series of samples and standards and can be replaced reproducibly. This tech-
nique especially allows the analysis of small manufactured parts (Bertin, 1970). Often
solid samples to be analyzed show surface roughness and=or damages, which cause
problems due to absorption eects. Such specimens must be nished at the surface before
analysis in order to reduce these damages. Several dierent techniques are available for
nishing a sample surface. The most common and eective one is mechanical polishing
with special abrasives like Al
2
O
3
or SiC, or with coarse and ne abrasive paper. A nal
nish of $100 mm surface roughness for short x-ray wavelengths and down to 10 mm for
long wavelengths may be satisfactory for a reliable analysis. Other polishing techniques
Copyright 2002 Marcel Dekker, Inc.
are electropolishing or etching. However, these are not as universal as mechanical pol-
ishing because the sample might be destroyed or the surface composition changed.
A. Metallic Specimens
X-ray uorescence is one of the most common techniques for the analysis of samples in
the metal production and, therefore, the preparation of such a specimen should be fast,
simple, and reproducible. Usually, the metallic specimens are prepared as solid disks by
conventional methods like cutting and milling. Hard alloys can be ground and pelletized
prior to analysis. However, for most metallic samples, it is required to polish the surface
in order to avoid striations, which give rise to the so-called shielding eects (Jenkins and
Hurley, 1965; Mu ller, 1972). For a reliable analysis, the surface roughness should not
exceed the path length at which 10% absorption on the radiation will occur (Jenkins
et al. 1981). In Table 1, some 10% absorption path lengths (presented as x
90%
in the
table) for common analytical situations of metallic specimens are indicated. For mea-
surements at short wavelength, a nal nish of $100 mm might be satisfactory, which can
easily be reached with abrasive paper or diamond paste. If light elements are determined,
a 1050-mm surface roughness should be claimed and ne abrasives are required to
obtain an almost mirrorlike surface. Striations, even after polishing, cause a reduction of
the uorescence intensities, which is more serious for long wavelengths than for short
ones (Bertin, 1978; Tertian and Claisse, 1982).
Another aspect which should be taken into account is the orientation of the speci-
men inside the spectrometer. By simple rotating of the sample with 45
or 90
in the
spectrometer and repetitive measurements, the grooves decrease the uorescence intensity
more when oriented in the perpendicular direction to the incident beam than in the parallel
direction. In Figure 1, a plot of the uorescent intensity versus the groove size for Al, Fe,
Cu, Pb, and Mo is presented (Mu ller, 1972). Also, here it is obvious that the reduction of
the uorescence intensity is much more severe for elements with low Z numbers. However,
today, most spectrometers are provided with a spinning mechanism, which averages this
eect. For a spectrometer without a spinning mechanism, the sample should be placed in
such a way that the grooves appear parallel to the incident radiation in order to minimize
the eect. Several mathematical models have been developed to determine the relationship
between surface roughness and emergent analyte line intensity (Berry et al., 1969; Mu ller,
1972).
However, simple mechanical polishing is not always possible or even advisable.
Particularly, when such elements as Si, Al, or Fe are to be determined, it is sometimes
dicult to nd a suitable abrasive which does not contain one of these elements and does
not introduce contaminations into the sample. SiC and Al
2
O
3
, for instance, are very
Table1 Mass Absorption Coefcients and 90% Transmission Path Lengths for Some Elements
Measured
radiation
Matrix
composition
m-Matrix
(cm
2
=g)
r
(g=cm
3
)
x
90%
(mm)
MgKa Fe 5430 7.87 0.02
TiKa Cr 144 7.19 1.02
CuKa Al 50 2.70 7.7
CuKa U 306 18.7 0.18
ZrKa U 57 18.7 0.98
Copyright 2002 Marcel Dekker, Inc.
common and eective abrasives, but both contain elements of high analytical interest. In
such cases, electrolytic polishing or etching might be more useful. When the surface of the
specimen is too rough (e.g., in the case of turnings) and suitable polishing and milling is
not practicable, the samples can be prepared properly by pressing them into pellets with a
hydraulic press at several hundreds of mega Pascals. With such a procedure, most of the
turning are compacted into satisfactory smooth surfaces.
If the sample is composed of soft, malleable multiphase alloys, smearing eects of
softer components like Pb or Al cause problems. The soft elements will be enriched onto
the surface and the uorescent intensities of the softer phases increase, whereas those of
the harder ones decrease (Tertian and Claisse, 1982). In such cases, the sample should be
nished by electrolytical polishing or etching, instead of mechanical polishing.
In the case of metallic specimens with irregular shape, it is possible to embed the
piece in a special wax resin (e.g., acrylic resin and methyl acrylic resin). The resulting block
can be polished and ground to appropriate surface smoothness with abrasives like SiC and
diamond paste before analysis (Wybenga, 1979).
If the surface of the sample is not representative for the bulk composition, due to
heterogeneity, corrosion, smearing, and so forth, it is necessary to treat the sample in
dierent ways. Bronze alloys can be prepared, for example, by etching with a solution of
alcoholic ferric chloride (120 mL C
2
H
5
OH, 30 mL HCl, and 10 g FeCl
3
), and brass alloys
by a solution of aqueous ammonium persulfate [100 mL H
2
O and 10 g (NH
4
)
2
S
2
O
8
]. For
archeological objects, however, it is not recommended to treat them by etching because
many inclusions could be dissolved and lost.
Another possibility is the complete decomposition or dissolution of the sample.
However, such a radical method of sample preparation should only be carried out when
other, less aggressive preparation techniques are not suitable for obtaining satisfactory
results. It should always be taken into account that a digestion or dissolution step con-
tributes an additional source of contaminations or losses and is quite time-consuming
(Tscho pel et al., 1980). The procedure and the necessary reagents must be carefully se-
lected in order to get the highest benet. An eective and common method is the digestion
Figure1 Change in relative fluorescent intensity of massive samples as a function of groove size a
for different elements, calculated for pure elements. (From Ref. 8. Reprinted by permission of
A. Hilger, Ltd.)
Copyright 2002 Marcel Dekker, Inc.
of the sample with aqua regia (HCl:HNO
3
, 3:1). Most of the alloys and metals, even Pd
and Pt, can be attacked by this reagent. The strength of this combination is based on the
formation of chloro complexes during the reaction and to the catalytic eect of Cl
2
and
NOCl. The eectiveness can be improved additionally when the solution is allowed to
stand for 1020 min before heating to 60
C was obtained.
Another, relatively new technique, is the remelting of the sample (Coedo and Dor-
ado, 1994; Coedo et al., 1994; Cobo et al., 1995). This method is particularly useful for the
determination of Cr in Mo- and Nb-ferroalloys. Such alloys are not suitable for direct
analysis by XRF, due to their inhomogenity and brittleness. With application of the re-
melting technique, the original ferroalloy is ground to grain sizes of 2 mm with a steel disk
mill, followed by a W mill, after previously crushing with a jaw beaker. Subsequently the
grained particles are mixed fairly with a portion of pure Fe (15 g FeMo alloy 25 g pure
Fe, and 12 g FeNb alloy 28 g pure Fe) and this nal mixture is melted inside an induction
furnace. Afterward, the melted metal is centrifuged and cast in an appropriate mold prior
to analysis. Calibration samples can easily be obtained in the same way.
The analysis of semiconductors is mainly carried out by TXRF or SRXRF; however,
there is also an interesting way to apply XRF for the determination of Ni in Si wafers.
Bubert et al. (1991) measured the Ni implantation of the wafers directly by XRF, but the
calibration standards were prepared in an unconventional way by spiking the gelatin-
containing Ni standard on parts of the Si wafer. The XRF analysis was performed after
drying of the sample.
B. Powdered Specimens
A common technique for geological, industrial, and biological materials is the preparation
of powders and pellets. Powdered specimens are prepared when the original sample is too
heterogeneous for a direct analysis or too brittle to form self-supporting disks or when a
suitable surface nishing is not possible.
Powders can be derived from dierent materials: solids like minerals, rocks, ores,
slags, and so forth, metal llings, metal oxides, precipitates, and residues from solutions,
ground fusion products, ion-exchange resins, and ashed or freeze-dried biological mate-
rials. Some of the materials are already available as powders, but most of them must be
pulverized by crushing, grinding, and milling. The powders are presented to the spectro-
meter directly as loose powders, packed in cells or spread out on lm materials like
Mylar
1
or other foils and as pellets or briquettes. Sometimes, fusions (see Sec. III) of the
powdered material may be a more suitable alternative, especially when particle size eects
are severe and need to be overcome. However, the preparation of a powdered sample
requires much less time than the preparation by using a fusion technique.
The analysis of a powdered specimen shows several advantages over bulk materials:
Powdered samples allow the addition of a standard and low- and high-absorption dilution.
The heterogeneity of the material will be reduced to a certain level, and standards can be
prepared easily in the same way. However, if the sample consists of elements with dierent
grinding properties, which is often the case for mineralogical and geological samples,
additives must be added to overcome such a problem. In general, 210% additive to aid in
the grinding, blending, and briquetting process is sucient for most of the materials.
To fulll the condition of a homogeneous sample surface with innite thickness, on
which most of the correction algorithms are based (see Chapter 5), the sample must be
converted into a very ne powder, otherwise particle size eects can occur and interfere
with the quantitative analysis. For very ne particles, the uorescent intensity is constant,
but it decreases with increasing particle size. This eect is more signicant for long
wavelengths than for short or intermediate ones. Furthermore, the degree of packing of
Copyright 2002 Marcel Dekker, Inc.
the specimen inuences the uorescence intensity additionally. However, this can be re-
duced by the preparation of pellets and by briquetting the sample; a higher pressure and a
longer pressing time for the pellets give an additional improvement.
Segregation and, hence, heterogeneity eects can arise when the particles are not of
the same size, shape, or gravity. This problem occurs particularly when the sample is
analyzed in the form of a loose powder. Claisse (Claisse, 1957; Claisse and Samson, 1972)
proposed a qualitative interpretation for the origin of such heterogeneity eects in some
publications. The eects can be a major source of errors because large variations in
uorescence intensity may occur and the inuence becomes more and more detrimental
with increasing wavelength; sometimes, errors of about 50% are possible (Berry et al.,
1969; Rhodes and Hunter, 1972; Madlem, 1976; Holynska and Markowicz, 1982). To
avoid severe errors, it is always recommended to tap the powdered sample for better
mixing of the powder prior to analysis.
Dierent mathematical methods for the correction of particle size eects have been
developed. However, with only few exceptions (see Chapter 6), they are not very useful, as
most of them require an a prior knowledge of the particle size distribution in the sample,
which is usually unknown.
1. Grinding
For the preparation of ne powders, the original sample should usually be split and milled
into smaller pieces, which are then suitable for further grinding. Splitting in smaller
subsamples can be carried out by various equipments, like jaw mills or crushers. The main
objective at this stage is to obtain a representative subsample from the original material
and to avoid sampling errors due to wrong handling or treatment. For further grinding of
the sample to appropriate small sizes, various methods and devices are available. The most
common equipments are disk mills or shatterboxes, in which samples of smaller sizes than
6 mm can be ground further.
However, both steps present a source of contaminations, as most of the grinding
tools contain elements, which might be of interest in the sample. These elements can be
introduced easily as blank values during the grinding process and, therefore, the material
of the grinding tools should be selected carefully for each particular application. Agate,
zircon, steel, alundum, and WC are the most common materials for such vessels. All of
them can introduce various blank values. Agate, for instance, introduces traces of SiO
2
,
Mg, and Ca to the specimen. This might be of less interest for geological materials, which
often contain these elements as main compounds, but for biological material, this re-
presents a major source of errors (Van Grieken et al., 1980). However, agate and zircon
equipments insert still lower blank values to the sample in comparison to other materials.
Unfortunately, agate is very expensive and brittle, and for this reason, it is mainly used for
the preparation of small sample amounts by manual grinding in a mortar. With zircon
devices, only blanks of Zr and Hf occur, which are negligible for most of the applications.
Despite its good mechanical properties, the use of WC vessels is only partially re-
commended. During the sample preparation, substantial amounts of W, Nb, Ta, and Co
can be introduced to the sample (Hergt and Sims, 1994). Especially the W contamination
can give serious problems because the L and M lines of this element interfere with the
elements of Z numbers between 28 and 35. However, despite this drawback, WC is the
most favorable material for the preparation of hard alloys due to its material properties.
Steel is also a frequently used material, and the preparation of rock samples with bakelite
phenolic resin in a steel mill showed good results (Longerich, 1995). However, a con-
tamination with Fe, Cr, Ni, and Mn is regular and can cause severe errors.
Copyright 2002 Marcel Dekker, Inc.
When grease and oil-forming materials are analyzed, the selection of a suitable vessel
material should be done carefully, as well. The preparation of these samples should be
carried out in vessels with smooth, nonporous surfaces, like agate. Otherwise, parts of the
sample can cling to the surface of the device and losses will occur. Sieber (1993) obtained
homogenized samples of greases by mixing them with a chemplex binder in a high-speed
rotar. The mixture was heated up during the procedure and was analyzed directly after
cooling down as a cake. Oil samples can be ground with MgO and baked afterward at a
temperature of 270
C. The received solid cake is suitable for direct analysis (Liu et al., 1992).
Reagents or additives are used during the preparation procedure for dierent rea-
sons. Fine abrasives like Al
2
O
3
or SiC are often added as grinding aids to accelerate the
process (Adler and Axelrod, 1995). The addition of wet grinding aids like ethyl or iso-
propyl alcohol leads to a better homogeneity and higher eciency. Calcination in an
airstream at 800
C results in a target
containing about 2 mg=cm
2
of sample material. Microscopic photographs showed that
the sample is quite homogenous with a grain size below 10 mm (Sauer et al., 1979; Van
Grieken et al., 1979).
Figure 2 (a) Grinding time versus intensity. (b) Pelletizing pressure versus intensity. (From Ref.
30. Reprinted by permission of Plenum Press.)
Copyright 2002 Marcel Dekker, Inc.
3. Pelletizing
Pelletizing of the powdered samples is often required for the reduction of surface ef-
fects and to yield better precision than with loose powders. The process can be carried
out very fast and easily. The required tools are only a hydraulic press and a suitable
die set, including a die body, base, a plunger, and two polished metal disks. When the
disks are made of WC, they also allow the pressing of hard alloys and abrasive ma-
terials. For pelletizing, the powdered sample is transferred into such die and pressed
under pressures of mega Pascals for several seconds up to 1 or 2 min. For most of the
materials, this will be sucient to reach a mirrorlike surface and to smooth scratches
and turnings.
The eect of the pelletizing pressure on the uorescent intensities for a coal sample is
displayed in Figure 2b. The coal powders were pelletized with a boric acid backing at
250 MPa. With uorescence measurements, the optimum pelletizing pressure was de-
termined. The light elements such as Si and Ca show a much higher dependence on the
pressure than, for example, Fe. However, other light elements like K and S seem to have
much less dependence on pressure than Si and Ca. This might be an indication that sig-
nicant dierences in the particle size of these elements are present in the coal sample
(Wheeler, 1983).
Binders are often added to the mixture to help form stable pellets and prevent caking
of the sample at the die surface. However, in the selection of a suitable binder, some
regulations should be carefully considered. First, binders should show low contamination,
particularly for the elements of interest, have low absorption for all wavelengths of in-
terest, and be stable under air, vacuum, and x-ray irradiation. The necessary amount of
binding material must be chosen carefully as well, because binders are normally composed
of light elements or organic materials and scatterred x-rays increase the background of the
spectrum. Binders also dilute the sample, which might be a serious problem for the de-
termination of trace elements. A recommended mixture is the addition of 210% binder to
the sample. In some special cases, higher concentrations of the binder can be accepted.
Bassari and Kumru (1994), for example, used a combination of 50% soil sample and 50%
cellulose binder in their work.
Dierent types of binders are available; most of them also act as grinding aids and
can be added during the grinding procedure. Typical binders are cellulose, starch, lucite,
urea, boric acid, graphite, KCl, and so forth. (Bertin, 1978; Frechette et al., 1979; Tertian
and Claisse, 1982). Liquid binders like ethyl alcohol or diethyl ether can be mixed with the
sample manually, whereas it is recommended that binders be mixed mechanically for the
formation of a homogenous mixture.
Liquid binders have the advantage of evaporation after the mixing process. Finally,
a stable pellet is formed and only minor amounts of binding material are still present in
the specimen, which can almost be neglected. To accelerate the evaporation process,
careful heating by an infrared (IR) lamp can be performed. If the heating process is
carried out too fast, cracks may develop inside the pellet and this can weaken the whole
structure.
In the last years, several new binders, mainly polymers, have been recommended; for
example, Pb-containing pulp samples were mixed with poly(vinyl acetate) binder in order
to form a stable disk (Volodin and Uranov, 1990). Other polymers are poly(vinyl alcohol),
poly(vinyl pyrrolidone) (PVP), or methyl methacrylate (Bettinelli and Taina, 1990; Domi,
1992; Wilson et al., 1995; Watson, 1996). Watson (1996) developed a simple and fast
method for pellet preparation by mixing PVP and microcrystalline cellulose binder in a
Copyright 2002 Marcel Dekker, Inc.
new polyethylene bag. The obtained moist mixture is then pressed to a pellet and dried.
For the determination of traces in geochemical samples, elvacite (a copolymer of methyl
n-butyl methacrylate) was selected as the binder (Ingham and Vrebos, 1994). In one recent
application, provenance studies were performed of coastal and inland archeological pot-
tery in South Africa. Here, the ground samples were mixed with movariol as binder and
analyzed (Punyadeera et al., 1997).
For pressing a pellet, the powderbinder mixture is poured into a metal support,
which has only the analysis surface open to prevent cross-contamination and to extend
the life time of a pellet. The support is usually made of Al and painted on the sides.
The exterior paint acts as an antifriction substance in the press mold, reducing scoring
of the die. Before pressing, the container is tapped several times in order to reduce air
pockets.
Another alternative for obtaining a stable pellet is pressing after the addition of boric
acid. In this case, the powder mixture is transferred into the die and attened by a plunger.
The die has been lled earlier with a small cylinder, which acts as a spacer and prevents
backing of the powder at the rim. The used plunger is made of Plexiglas to allow the
observation of the attening process. The inset cylinder is then removed carefully, which
leaves a gap between sample and die. Boric acid crystals are poured in this gap and on top
of the sample, and the whole assembly is pressed.
In some cases, heating of the sample during the mixing procedure with binders can
occur and volatile components may evaporate. Special mixing machines with cooling
mechanisms or slower milling procedures are in use instead of ring and puck mills for
prevention of such losses (Domi, 1992).
Another procedure can be applied when only small amounts of the sample are
available. The powdered sample is placed as a thin layer on top of a backing material layer
and the pellet can be subsequently pressed in an easy way.
To check whether a specimen is innitely thick, the sample is covered by a small disk
made of Cu or other metals, depending on the elements to be analyzed, and then pressed.
The presence of Cu lines in the spectrum indicates that the sample is not innitely thick
and the pressing procedure should be repeated with more material.
For pressing of the nal pellet, some precautions need to be considered. Normally,
the used pressure is about 15 tons, which is sucient for the formation of stable pellets for
a long time. The pressure should be released carefully to avoid breaking or backing of the
pellet, and the removal of the pressed pellet from the press should be carried out with
caution for the same reason. If the pellet is too brittle, it may be broken or pulverized
again during rough handling.
However, if oily or waxy samples are pressed, they may ow under the pressure and
form coatings at the surface of the pressing tools. This can be avoided by the addition of
components, which stabilize such pellets. Sometimes, the sample structure itself might be
destroyed by unsuitable pressures or too long exposure. This is often not visible im-
mediately after pressing, but after storage of the pellet for longer times, problems might
occur. Also, conical fractures can occur inside the pressed pellet and weaken it. Un-
fortunately, such fractures are also not always visible and breaking out of the pellet inside
the instrument during the evacuation process might be the result. For the prevention of
these unwanted eects, the pellet can be exposed to vacuum in a desiccator before
transferring it into the XRF unit.
In their study, Novosel-Radovic and Malijkovic (1993) monitored the deterioration
of a pressed sample as a function of time. They found that briquets age with time,
resulting in the formation of surface cracks.
Copyright 2002 Marcel Dekker, Inc.
III. FUSEDSPECIMEN
In 1956, Claisse proposed preparing fused samples for analyses by XRF in order to over-
come the problems arising with powders and pellets (Claisse, 1956, 1957). This technique
has become one of the most successful methods for the preparation of specimens in x-ray
uorescence and can be applied to almost all types of samples (Feret, 1990, 1993). With a
fusion, all particle and mineralogical eects, which cause serious problems for analysis of
powdered and pelletized specimen, become negligible and the analysis shows a better ac-
curacy (Banajee and Olson, 1978; Frechette et al., 1979; Tertian and Claisse, 1982; Metz and
Davey, 1992). The basic procedure consists of heating a powdered sample together with the
ux to such high temperatures that the ux melts, dissolves the sample, and reacts with it by
the formation of a one-phase disk. Agitation during the fusion process is essential for fu-
sions with low uidity, in order to avoid the formation of bubbles and to receive a homo-
genous sample (LeHouillier and Turmel, 1974). The reaction time varies between 5 and
30 min depending on the sample composition. For a 30-mm disk 67 g material (ux and
sample) is required. For most applications, the ux=sample ratio varies between 2:1 and 10:1
(Bower and Valentine, 1986; Eastell and Willis, 1990; Kvaratskheli and Kornienko, 1992;
Ossaka et al., 1994; Spangenberg et al., 1994). In some special applications this ratio may be
dierent (Haukka and Thomas, 1977; Fink et al., 1993; Sato and Kohori, 1995).
One of the major advantages of the fusion technique is the possibility of adding a
heavy absorber for the reduction of matrix eects, especially for the determination of light
elements. Other advantages are the relatively simple preparation of synthetical standards
and the wider calibration range (Feret, 1990, 1993). The main disadvantage of this tech-
nique presents the dilution eect of the sample, which renders the determination of traces,
especially of light elements, more dicult, as their uorescence intensity is decreased.
However, in the last years, several attempts were carried out to overcome this dis-
advantage [e.g., by application of the low dilution fusion technique (LDF) (Haukka and
Thomas, 1977; Thomas and Haukka, 1978; Lee and McConchie, 1982; Eastell and Willis,
1990; Maruta et al., 1992; Yamamoto et al., 1995) or a semilow dilution technique
(Malmqvist, 1998)]. Further disadvantages include the higher costs in comparison to the
powder methods and the more complicated preparation procedure of the fusion. Some
experience is required and most of the analysts have their special tricks to obtain the
best results. Today, several automated ux preparers are commercially available with
more or less sophisticated technology (Stephenson, 1969; Schroeder et al., 1980; Kvar-
atskheli and Kornienko, 1992).
For some elements like S, however, the briquetting technique is preferred and oers
better accuracy, because high losses due to volatilization during the fusion procedure occur.
A. Fluxes and Additives
Several uxes, with dierent properties, are in use for the dissolution of the samples and
the formation of glass disks. The uxes are mainly composed of light elements, which do
not interfere strongly with the elements to be determined and form stable glass beads
(Muia and Van Grieken, 1991). Borate uxes, containing metaborate or tetraborate,
dissolve almost all materials except metallics, suldes, and organic materials. The choice
for each application is dependent on the sample composition. Common uxes are lithium
tetraborate (Li
2
B
4
O
7
), lithium metaborate (LiBO
2
), sodium tetraborate (Na
2
B
4
O
7
), and
sodium phosphate (NaPO
3
)
6
. Low-temperature uxes are potassium pyrosulfate (K
2
S
2
O
7
)
and sodium carbonate (Na
2
CO
3
).
Copyright 2002 Marcel Dekker, Inc.
Usually, the Li uxes are preferred to the Na ones. They produce more uid uxes
and have lower mass absorption coecients; this inuences the uorescence intensity less
and is essential for light elements. Their disadvantages are the higher reaction tem-
peratures and higher hygroscopicity. Lithium tetraborate is an eective reagent for at-
tacking basic oxides, like carbonates or Al oxide. However, its glass beads are cracking
easily and the disks should be handled with care. Lithium metaborate forms glass beads
with excellent mechanical properties and attacks specimen containing high concentra-
tions of oxidic acids (e.g., silica). Often, mixtures of the lithium uxes are used, which
can be more eective, as they combine the properties of the single components. The
addition of Li
2
CO3 or LiF increases the acidity or basicity as well as the reaction speed
and uidity.
With a mixture of sodium metaborate and sodium nitrate (NaNO
3
), samples which
are dicult to attack by lithium reagents (e.g., chrome-containing ores and refractories)
can be dissolved (Sear, 1997). NaNO
3
and boric acid react with geological materials like
casserite (SnO
2
) and chromite (FeO Cr
2
O
3
) (Couture et al., 1993). However, Na uxes
tend to wet and stick to the crucible, which makes it necessary to clean the crucibles
regularly. In Table 3, some examples for ux compositions and ux=sample ratios are
given.
Table 3 Some Examples for Flux Compositions and Typical Applications
Flux base
Flux
composition Properties Applications
LiBO
2
LiBO
2
=Li
2
B
4
O
7
4:1 Good mechanical
properties
Silica or alumina major components
(Istone et al., 1991), CaO (Alvarez, 1990)
LiBO
2
LiBO
2
=silica 22:3 Ceramics with lead zirconate (Sato and
Kohori, 1995)
LiBO
2
LiBO
2
High fluidity Coal ash, rock samples (Sweileh and
Van Peteghem, 1995)
Li
2
B
4
O
7
LiB
4
O
7
=LiBO
2
12.5:10 Aggressive Cr-containing refractories (Giles et al.,
1995)
Li
2
B
4
O
7
Bead cracks easily Metal oxides, basic oxides, basic metal
sulfides, rocks, carbonates, silicates,
zeolithes (Luke, 1963; Le Houillier and
Turmel, 1974; Eastell and Willis, 1990;
Oishi et al., 1995, Rutherford, 1995)
Li
2
B
4
O
7
Li
2
B
4
O
7
=Li
2
CO
3
Ferro-alloys, paper pigments
(LeHouillier et al., 1976; Alvarez, 1990;
Muia and Van Grieken, 1991)
Li
2
CO
3
Li
2
CO
3
=NH
4
NO
3
20:1 Feores (Sato and Kohori, 1995)
Na
2
B
4
O
7
Metal oxides, rocks
volcanic ash, MgO, chrome-containing
ores (Dow, 1981; Kvaratskheli and
Konienko, 1992)
Na
2
B
4
O
7
Na
2
B
4
O
7
=Na
2
CO
3
3:1 Slags, sinters, metal oxides (Muia and
Van Grieken, 1991)
Copyright 2002 Marcel Dekker, Inc.
To overcome the problem of attacking the crucible material, especially by suldes
and partially oxidized organic compounds, the sample can be preignited or an oxidant
added (Martin and Richardson, 1992). Slags from Cu-smelting processes can be analyzed
after dissolution in sodium tetraborate combined with lanthanum oxide (La
2
O
3
) as a
heavy absorber and NaNO
3
as an oxidant to prevent the attack of the crucible by Cu (Le
Houillier and Turmel, 1974). Other examples for oxidants are BaO
2
, CeO
2
, KNO
3
, and
LiNO
3
.
Sulfur-, chlorine-, and uorine-containing samples tend to stick in the crucible and
shorten their lifetimes. Preliminary roasting of the samples minimizes this problem (Dow,
1981). To avoid losses of elements, like sulfur and organic compounds, the sample can be
calcinated for several hours in a mue oven at 750800
C.
The fusion procedure can be carried out in high-frequency induction furnaces, gas
burners, or mue ovens. The rst two show the advantages of higher temperatures and
faster heating. The sample is previously heated until the glass-forming agent is melted;
with a further increase of the temperature to 10501100
C is necessary
to volatize all CO
2
. Otherwise, a weight loss will occur during the fusion, which cannot be
compensated for by mathematical calculation.
Evaporation of moisture or crystal water can also introduce some weight loss. If this
loss is similar for both the sample and ux, it can be calculated using the method of
Hutton and Elliot (1980) or compensated for by the addition of SiO
2
, if less than 20%
(Harvey et al., 1973). Any additional mathematical correction procedure is then super-
uous. The losses of volatiles and=or ux usually increase with increasing temperature.
King and Vivit (1988) developed the following simple method for the prevention of losses
in geological samples: The powdered sample was dried for 2 h at 105
C and 2 g was
Copyright 2002 Marcel Dekker, Inc.
weighed exactly and calcinated in a mue furnace at 925
and
SO
2
4
) of drinking water were preconcentrated onto polypropylene foil. The addition of a
polysaccharide (GELERITE) reduced the hydrophobic eect of the foil and, thus, de-
tection limits from 0.12 mg=L for Ca to 1.78 mg=L for SO
2
4
were claimed (Warner et al.,
1995). The advantage of this technique is the possibility of a simultaneous determination
of cations and anions. In another application in this direction, the sample is transferred as
a droplet onto a polycarbonate lm, which is cemented on a 35-mm plastic photographic
slide mount. After evaporation, the resulting salt deposit is more or less circular. Detection
limits in the microgram per liter region could be reached (Meltzer and King, 1991).
Particularly, thin-lm supports (like the type A2) have recently been developed by
Process Analytics (Orem, UT). The lm can accumulate a small amount of solution in a
small spot because the lm is treated to have a hydrophilic center in a hydrophobic eld or
the lm has a dimple in its center. These lms are clean and resistant to high temperatures.
Wastewater samples with low salinity can be preconcentrated by a special vapor
ltration procedure (Luke, 1963; Rickey et al., 1977) in which the sample is deposited into
a container, with a cellophane foil bottom. The foil is permeable for water vapor, but not
for liquid water or dissolved material. Subsequent evacuation of the container leads to
deposition of the dissolved solids onto the membrane or foil, which is then presented for
analysis. In combination with particle-induced x-ray emission (PIXE), detection limits in
Copyright 2002 Marcel Dekker, Inc.
the range of 0.13 mg=L were claimed. However, this method shows several disadvantages:
It is time-consuming and samples with high salinity cannot be applied successfully. Crystal
formation can occur, even with very low salinity, which has to be corrected by special
matrix correction procedures (Van Grieken, 1982).
Sometimes, metal surfaces can also be used for the deposition of extracted solutions.
A thin-lm technique was applied for the determination of SO
2
4
and Cl
in small volumes
of saline samples, deposited onto suitable machined Cu disks (Dhir et al., 1995).
Other procedures are a nebulization technique (Giauque et al., 1977) and a multi-
droplet spotting technique (Camp et al., 1974), which result in a fairly homogeneous
distribution of the sample onto the surface.
The evaporation of larger samples can be carried out by impregnation of the lter
material (Whatman 41) with a sample solution. A wax ring 29 mm in diameter keeps the
spotted sample solution inside the area to be irradiated when the water is evaporated by
passing an airstream from underneath. The reported detection limits were below 50 mg=L
for 2000 s analyzing time, with optimum secondary target excitation (Giauque et al., 1977).
Water samples can be properly analyzed by TXRF without any preconcentration
step. Some microliters of the sample are pipetted directly onto a highly polished quartz
glass carrier and evaporated. The detection limits achieved with this techniques are one to
sometimes two orders of magnitude better than for conventional XRF. However, if
samples with higher salinity are analyzed, a separation step is also required.
Probably the best solution, when high precision and low detection limits are ne-
cessary, is a sample preparation involving a chemical preconcentration or separation step.
B. Chemical Preconcentration
Chemical preconcentration may be grouped into three main classes:
1. (Co)precipitation
2. Ion exchange
3. Chelation and sorption immobilization
Numerous applications to dierent water samples have been established in these
three groups (Camp et al., 1974; Rickey et al., 1977; Bruninx and Van Ebergen, 1979;
Andrew et al., 1982; Chakravorty and Van Grieken, 1982; Van Grieken, 1982; Becker
et al., 1985; Hirayama and Leyden, 1986; Shan et al., 1988) and most of them include a
nal step in which the preconcentrated sample is deposited onto a surface (lter or foil) by
sorption or ltration. This surface serves as sample carrier and is subsequently exposed to
the x-radiation.
1. (Co)precipitation Methods
Precipitation methods are quite popular for the tracematrix separation and they have
been used in analytical chemistry for a long time. They can be carried out very quickly and
with little eort. Selective precipitation of one or two species is usually the main task and
much of the time is dedicated to nding suitable reagents for this purpose. However, in the
case of XRF as a multielement technique, it is much more important to obtain a pre-
cipitate containing all elements of interest. Many attempts were carried out in this di-
rection and numerous reagents with dierent properties are practicable for this. A simple
and traditional method is the application of inorganic reagents like Fe or Al hydroxide,
which coprecipitate many elements at selected pH values. Suldes are less common,
Copyright 2002 Marcel Dekker, Inc.
because of the higher solubility of their precipitates. However, most of the inorganic re-
agents show an incomplete precipitation and a strong dependence on accurate pH values.
More common and eective are organic components, which form complexes of very
high stability. Often a spike element is added, acting as a carrier and an internal standard
at the same time. The most suitable organic reagents are the carbamates (Van Grieken,
1982). They form very stable and strong metal chelates. Luke (1968), for example, in-
vestigated, using his coprex technique, the coprecipitation of trace elements by sodium
diethyldithiocarbamate (NaDDTC) in the presence of a suitable metal ion spike. In a
further work, Kessler and Vincent (1972) improved the detection limits, obtained by Luke
(1968), by ltration of the received NaDDTC=hydroxide precipitate and analyzing of the
residue with a highly collimated XRF setup.
Toxic elements were successfully determined in plastic compounds for toys by ap-
plying NaDDTC. The samples were rst mineralized with molten NaOH. Then Cd, Ba,
Sb(III), Cr(III), Hg, Pb, and As(III) were coprecipitated at pH 8.5 with NaDDTC and
sodium rhodizonate and Fe
3
ions as the carrier. The results achieved by XRF were
comparable with the AAS results (Gimeno-Adelantado, 1993). When a high sample
throughput is required and all samples are of the same kind, automation might be pos-
sible. For such purposes, Tanoue et al. (1979) developed an automated device for the
preconcentration of trace metals in wastewater samples with DDTC and subsequent XRS
measurement.
Ammonium pyrrolidine dithiocarbamate (APDC) acts more eciently, especially for
concentrations below 10 mg=L at a pH of 4. APDC is the most successsful nonspecic
reagent for the preconcentration of traces at this pH in comparison to other reagents.
Quantitative recoveries are obtained for a dozen elements in water, independent of the
alkaline earth element content. Various elements can be also determined in seawater, using
APDC as reagent. A small amount of Mo as carrier ion is added to the ltered seawater
sample, then the pH is adjusted at 4, and a portion of the 1% APDC solution is added.
After 1520 min the precipitate is ltered o, dried, and analyzed. Ulrich and Hopke
(1977) achieved detection limits between 4 ng=cm
3
for Br and 80 ng=cm
3
for Ca by ap-
plication of this method. In another approach, traces of soluble metals were separated
from estuarine waters. In the rst step, the water was ltered and acidied to pH 3. Fresh
APDC solution (2%) was added and then the complexes ltered. The obtained detection
limits were between 2 and 5 m=L (Mazo-Gray et al., 1997).
Quantitative recoveries are also received by application of dibenzyl dithiocarbamate
(DBDTC) for Mn, Fe, Co, Ni, Cu, Zn, Se, Sb, Hg, Tl, Ag, Cd, and Pb with detection
limits around 1 mg=mL for 100-mL samples (Watanabe et al., 1986). The very low solu-
bility of this reagent also eliminates the need for a metal carrier (Lindner et al., 1978; Ellis
et al., 1982). A detection limit of 10 mg=L could be reached for Mo in water samples by
preconcentration with DBDTC at pH 3 (Saitoh et al., 1984).
1-(2-Pyridylazo)-2-naphthol (PAN) represents another attractive reagent for several
water samples. This organic compound is soluble in hot water and also in ethanol, but not
in cold water. This behavior can be utilized for the separation of special elements. In one
study, 2 L of water were heated to 70
4
for pre-
concentration and activation of the lter surface. The sample, consisting of snow, rain-
water, or even seawater of high salinity, was dissolved in HCl solution before presenting to
the lter surface. A monitoring of trace elements in underground and wastewater might be
also possible with this technique. The use of a cascade of lters under dynamic ow
conditions renders it possible to preconcentrate suspended substances and trace elements
from the solution in only one step.
Earlier studies, dealing with the enrichment of trace elements from environmental
samples onto lter surfaces, were mainly focused on ion-collecting papers (Van Grieken,
1982). However, most of these applications suer from unselectivity and low capacity for
the preconcentration of transient elements in large sample amounts, because alkaline and
alkaline earth elements are also trapped. Moreover, their capacity of enrichment for small
samples is too low to receive relevant detection limits for environmental samples.
Hyphan prepared by pelletizing of cellulose ber material as thin layers is of more
interest for large volumes. Good recoveries for trace metals are found from large volumes
of waters at pH 7. The detection limits usually lay in the range of 1 mg=L (Burba and
Lieser, 1979). Selective analysis for Co, for example, can be performed, after washing the
Hyphan-loaded lter with 3 M HCl. The washing procedure removes all other metals,
except Co. The amount of 0.4 mg=L Co can be determined, even in the presence of 1 mg=L
Fe (James and Lin, 1982; Coetzee and Leiser, 1986).
3. Chelationand Sorption Immobilization Methods
For most of the analytical methods, the extraction techniques are popular, as they supply
solutions which can be often transferred directly into the sample chamber of the instru-
ment. For XRF, however, this is normally not possible because most of the element
concentrations in the extracted solutions are still too low for a reliable analysis.
Reversed-phase extractions, in which organic chelation agents are xed to a solid
phase, are much more convenient for the measurement by XRF. For such a method, the
adsorbent and the sample solution are mixed together in a vessel. Agitation of the solution
during the reaction process ensures faster adsorption. Subsequently, the sorbent is sepa-
rated from the solution by ltration or decantation. For the removal of undesired traces,
the sorbent can be washed with appropriate solutions several times and then dried or
pelletized. The dried lter or pellet is then directly presented to the spectrometer.
Activated charcoal is well-known and often used as adsorbent. This material is a
good adsorbent for organic and colloidal materials and for the trace metal species bound
Copyright 2002 Marcel Dekker, Inc.
to such materials. With the addition of a chelating compound to charcoal suspensions, free
ions can be adsorbed quantitatively as well. Vanderborght and Van Grieken (1977), for
example, applied 8-quinoline as a chelating agent. The optimized preconcentration pro-
cedure consists of the addition of 10 mg 8-quinolinol per liter of water sample at pH 8 and
the application of 100 mg precleaned activated carbon to this solution. After ltration of
the suspension, quantitative recoveries were achieved for $20 ions from dierent media,
with enrichment factors near 10,000, more or less independently of the alkali and alkaline
earth element content.
Pera niemi et al. (1994) investigated zirconium-loaded activated charcoal as an ad-
sorbent for As, Se, and Hg. The activated coal was added to an aqueous solution, con-
taining these three elements; then the solution was stirred, allowed to stand, and, nally,
ltered through membranes. The lters were carefully dried at room temperature in order
to avoid losses, placed between Mylar sheets, and irradiated.
A selective determination of Se(IV) in various water samples can be carried out after
the reduction of Se(IV) with ascorbic acid to elemental Se and subsequent adsorption onto
activated carbon. After ltration, the lter is analyzed by EDXRS. Detection limits of
0.05 mg=L and a precision of 6% were found by Robberecht and Van Grieken (1980).
Foams represent another attractive extraction medium for the separation of traces
from dierent bulk materials. Carvalho et al. (1995) determined Ga from bauxite or Al-
based alloys as GaCl
4
, with polyurethane foam serving as sorption medium. The material
was rst dissolved with HFHCl, then evaporated to dryness and dissolved again in
6 mol=L HCl mixed with TiCl
3
. Finally, polyurethane foam, pretreated with HCl, was
added and the solution ltered after some minutes of reaction time. In combination with
organic chelation agents, polyurethane foam is even more ecient. Mercury was de-
termined in dierent matrices and chemical states by Braun et al. (1984) after pre-
concentration on DDTC-loaded polyurethane foam disks. The preparation of the disks
was carried out by the addition of a 4% solution DDTC in chloroform (assuring a 20%
concentration of the regent on the disk) and the addition of 0.2 mL dinonyl phthalate
plasticizer. After adjustment to pH 5, 25 mL of the sample solution were shaken with the
foam disk for 1 h to separate Hg. Extraction eciencies between 88% and 100% were
reported for 0.2 and 2 mg=L concentrations.
Gold can be extracted with tributyl phosphate as solid extractant from ore samples.
Therefore, the element is enriched from pulps after aqua regia leaching. For the de-
termination by XRF, a back-extraction with thiourea solutions at 100
C is required; to
obtain nally a thin sample, the thiourea solution can be placed inside a polyethylene
cylinder, supplied with a lter paper disk covering the bottom, and evaporated. For a 25-g
sample, the detection limits were at 10 ng=g (Dmitriev et al., 1991).
The extraction of dierent elements from coal samples is carried out in three steps,
and each step is based on the dierent solubilities of the dierent trace elements. In the
rst step, the ground coal sample was treated with deionized water 24 h, the residue re-
moved and vacuum dried. In the second step, ammonium acetate is added to the sample
and this suspension is heated to 70
C and 600
C in order to destroy the protein material. Subsequently, the material was pulverized
and pressed into a pellet, which was presented to the spectrometer (Maddox et al., 1990).
Sometimes only partial ashing is recommended. Havranek et al. (1986) performed
this for the analysis of hair. An 1-g aliquot of hair is partially ashed in a mue furnace at
200
C, the cool ash was homogenized, and a 0.3-g portion was treated with a few drops of
Copyright 2002 Marcel Dekker, Inc.
polystyrene solution in chloroform and pressed into a pellet. Urine samples are analyzed
as pellets as well after evaporation and partial decomposition at 220450
C (Nielson and
Kalkwarf, 1978).
However, treatment of samples at high temperatures suers from several drawbacks
and is therefore rarely employed. The major drawback is the loss of volatile components of
the sample. The addition of an ashing aid may minimize it, but the elements cannot be
retained completely (Nielson and Kalkwarf, 1978; Hang and Ha, 1996).
Another more sophisticated but even more expensive technique is the ashing of the
biological material with an oxygen plasma. For this procedure, the nely pulverized
sample is placed onto the bottom of a specially designed plasma tube made of quartz. The
tube is then connected to the plasma unit and evacuated. The sample is ashed by passing
oxygen through the tube, and continous cooling by a cooling nger assures that most of
the volatile components remain in the sample. Selenium and arsenic can be kept almost
quantitatively, but mercury shows high losses. The remaining residue is then taken up by
HNO
3
or a combination of HNO
3
and HCL. Amounts of 50100 mL of the solution are
pipetted on a foil target of Mylar or Nuclepore and dried prior to analysis. Unfortunately,
organic materials often contain high concentrations of Ca and K salts, which crystallize on
the foil and cause a high-background signal. Sometimes, it is necessary to separate such
salts by suitable extraction methods or to select an alternative way of sample preparation.
In general, ashing in an oxygen plasma is rather fast and easy to handle. However, for
complete ashing in an aordable time, the sample should be pulverized to small grain sizes
in order to obtain a suciently large surface for the reaction with the oxygen gas. Fur-
thermore, the reaction time depends on the material to be digested and might vary between
several minutes and several hours for materials that are dicult to digest (Raptis et al.,
1983; Knapp, 1984).
A traditional and relatively fast method for sample preparation is wet ashing with
HNO
3
or a mixture of HNO
3
with H
2
SO
4
, or HClO
4
. With these mixtures, most of the
organic material can be attacked (Bock, 1979). Cu, Fe, Rb, Se, and Zn in serum samples
were determined after digestion with HNO
3
. For quantication, Ga was added as an in-
ternal standard and a 20-mL aliquot was pipetted and dried onto a sample carrier for
analysis by TXRF (Raptis et al., 1983). Even more eective is a combination of HNO
3
and
HF, as it dissolves also silicates in organic materials.
A microdigestion method was developed for the trace element determination in
biological materials and human serum. For this, 500 mL of human serum were treated with
500 mL HF and dropped onto a Si-wafer chip. After drying, the chip was analyzed by XRF
(Sayama, 1995).
In general, a digestion procedure should be carried out in a closed vessel, as working
under open conditions might cause contaminations and losses which inuence the results
(Tscho pel, 1983). Using HClO
4
requires special care. This acid tends to explode when
organic material is not completely oxidized. A better and safer method is the application
of closed PTFE vessels or quartz tubes, which can also be used for the digestion under
high pressure. However, the selection of quartz tubes is highly recommended, because the
smooth surface of this material avoids adsorption eects of the elements. Moreover, it can
be cleaned easily by exposure to nitric acid fumes in order to avoid the introduction of
contaminations. PTFE or Teon might be easier to handle, but it shows higher con-
taminations and is more dicult to clean compared to quartz because of rough surface.
Zinc for example, is a well-known contaminant of Teon vessels. Another problem that
might occur with Teon is diusion from and into the walls. This is particularly serious for
volatile elements and can lead to signicant mistakes in the nal data.
Copyright 2002 Marcel Dekker, Inc.
For the determination of trace elements in lichen material, 200 mg of the collected
samples are digested with 3 mL HNO
3
, 0.2 mL, H
2
O
2
, and 0.05 mL HF in a PTFE vessel.
From the digested sample, targets can be prepared by pipetting 300 mL of the solution onto
a Mylar lm and drying (Calliari et al., 1995).
A very eective method for the digestion of organic materials is the high-pressure
digestion in specially designed bombs. Usually, a small amount of the pulverized sample is
placed in a quartz or Teon vessel and some milliliters of acid are added. The tube is then
placed inside a high-pressure chamber and exposed to pressures of about 200 bar under
nitrogen atmosphere, at temperatures of about 260