Escolar Documentos
Profissional Documentos
Cultura Documentos
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.
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.)
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 I0 emrt 1
where
X
n
m Wi mi 2
i1
I0 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 mi and Wi 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.
Figure 3 Loss in energy of x-rays in Compton scattering at angles 90 , 120 , and 150 .
The I s are the x-ray intensities of the elements measured, kij are correction coecients
determined by multivariable linear least-squares tting, and Wi 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.
lnI0 =I=rt mM
Wi 11
mi mM
where mi is the mass absorption coecient of the x-rays in the analyte and mM is the ab-
sorption coecient of the matrix, given by
X
n
mM mj Wj 12
j6i
and the subscript j refers to the matrix elements and S Wj 1 Wi . 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 Wi of the desired analyte is
m0M Rm00M
Wi 13
m0i m0M Rm00i m00M
where
lnI0 =I0
R 14
lnI0 =I00
and
m0M m0i m0M Wi m0
R 15
m00M m00i m00M Wi m00
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 m0i
m0M and when rt is large. The analysis is
accurate when the ratio m0M =m00M 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
s
2 2
R dI=I0 dI=I00
dR 16
rt m0 m00
where
0 2 0 2 0 2
dI dI dI
00 0 17
I I0 I
and
00 2 00 2 00 2
dI dI0 dI
00 00 18
I I0 I
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 R dR 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
mM new mM old mk ml 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.
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 few radioisotope sources that are used frequently 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 3.761010 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 N0 to N after an elapsed time t being given by
N N0 e0:693t=T1=2 20
where T1=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 109Cd (Amersham, 1986).
Cylindrical sources used in portable analyzers are usually 8-mm-diameter by 5-mm-thick
Copyright 2002 Marcel Dekker, Inc.
Table 1 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
Dose at 1 m
Radio- Half-life X- or g-ray Photons per from 1 GBq Analytical
isotope (years) energy (keV) disintegration (27 mCi) (mSv=h) 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.
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 55Fe 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.
Copyright 2002 Marcel Dekker, Inc.
109
Figure 4 Encapsulation of disk and annular Cd g-ray sources. Dimensions are in millimeters.
(From Amersham, 1986.)
Class
Test for
resistance to 1 2 3 4 5 6
Temperature No test 740 C (20 min.), 740 C (20 min.) 740 C(20 min.), 740 C (20 min.), 740 C (20 min.),
80 C (1 h) 180 C (1 h)
400 C (1 h) and
600 C (1 h), 800 C (1 h)
thermal shock thermal shock from thermal shock from
from 400 C to 20 C 600 C to 20 C 800 C to 20 C
External No test 25 kPa absolute to 25 kPa absolute to 25 kPa absolute 25 kPa absolute 25 kPa absolute to
pressureb atmospheric pressure 2 MPa absolute to 7 MPa absolute to 70 MPa absolute 170 MPa absolute
Impactc 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 30 min, 2550 Hz at 90 min, 2580 Hz
5gn peak amplitude 5gn peak amplitude; at 1.5 mm amplitude
5090 Hz at 0.635 mm peak-to-peak;
amplitude peak-to-peak; 802000 Hz at 20gn
90500 Hz at 10gn
Punctured 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.
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 Z elements.
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 such limited 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.
Copyright 2002 Marcel Dekker, Inc.
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 mm263 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).
The background intensity, Ibgd , 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 E1 (where E1 FWHM), it will generate in its FWHM-
wide window a certain intensity, I1. If the detector resolution now degrades to, say, E2
(i.e., the analytes x-ray peak is now wider), the intensity I2 will still be equal to I1 (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, E2, than in the case of E1, because the integration window for the back-
ground intensity is wider in the case E2 than in the case E1. Then, it follows from
Eq. (19) that the DL for the E2 case will be larger (worse) than in the case of E1
energy resolution. Therefore, for the best DLs, it is of paramount importance to use
the detector with the best energy resolution.
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
Figure 8 The energy difference between the Ka x-rays of adjacent atomic number elements.
Figure 10 Calculated efficiencies of NaI scintillation detector and silicon and germanium solid-
state detectors used in XRF analysis.
C. Electronics
The electronics used with the various detectors are discussed in Chapters 2 (Sec. III.F) and
in 3 (Sec. IV) and are also covered in detail by Jenkins et al. (1981a). The limits to accuracy
and sensitivity of XRF analysis are usually determined by the limitations of the detector in
energy resolution, eciency of detection, and maximum count rate, for example, rather
than of the electronics. With the excellent gain stabilization electronics now available and
the reliability of low-power, surface-mount components, it is rare that the electronics
system is a signicant limiting factor even in the harsh environmental conditions of
industrial plants.
Copyright 2002 Marcel Dekker, Inc.
IV. X-RAYAND g-RAY TECHNIQUES
The range of radioisotope x- and g-ray techniques used for analysis is far more extensive
than the range based on x-ray tube techniques. Almost all x-ray tube systems are based on
the high-energy resolution of wavelength-dispersive (the crystal spectrometer) or energy-
dispersive (the solid-state detector) devices. With this high resolving power, there is less
need to tailor a technique to the specic application. Radioisotope x-ray systems, espe-
cially those involving scintillation or proportional detectors, usually must be carefully
matched to the specic application. This disadvantage is more than compensated for by
such attributes as mechanical ruggedness, simplicity, and portability, which are so im-
portant in industrial and even more so in eld applications.
The selection of the radioisotope source to analyze dierent elements depends on
many factors, including whether the energy of the radioisotope x- or g-rays is sucient to
excite the element, the energies of the x-rays scattered by the sample, and the energy re-
solution of the detector. Figure 12 is an approximate guide and, although prepared for
proportional detectors (courtesy of Metorex Int., formerly Outokumpu Oy), can be used
for solid-state detectors and to a more limited extent for scintillation detectors. The iso-
topes of 238Pu and 244Cm, emitting x-rays similar in energy, can be used interchangeably
although curium is preferred for safety reasons. A 57Co isotope can be used with scintil-
lation detectors or germanium solid-state detectors for the K shell XRF analysis for high-
Z elements, such as uranium, gold, and lead.
This section reviews the radioisotope XRF techniques used with solid-state, pro-
portional, and scintillation detectors, x-ray preferential absorption techniques that are
normally based on the use of scintillation detectors, and x-ray scattering techniques that
are often based on use of scintillation detectors. An example of the application of each
technique is also given.
Figure13 The spectrum of x-rays from a copper, lead, and zinc ore sample excited by 238Pu x-rays
and detected by a Si(Li) solid-state detector.
proportional counter lled with a neonargon Penning mixture gas to a pressure of 7 bar
(Jarvinen and Sipila, 1984b). The detection limits are comparable to those obtained with a
silicon solid-state detector with x-ray tube excitation (Jarvinen and Sipila, 1984b). If other
elements in the atomic number range 2630 had been present, however, there would have
been incomplete resolution of the K x-rays emitted and, hence, poorer sensitivity of
analysis.
Figure 17 shows detection limits for the elements in water (Jarvinen and Sipila,
1984a; other data courtesy of Asoma Instruments), normalized to a common counting
time of 100 s. The low detection limits were obtained by careful choice of lling gas and
pressure in relation to the exciting x-ray energy used. The issue of proper match of the
detector with the energy of exciting radiation is much more important for gas-lled pro-
portional detectors than for the solid-state ones. This is because of the much greater
uniform detection eciency of the solid-state detector for x-rays than that of the gas-lled
one. For most, except low-atomic-number elements, the detection limits were less than
10 mg=g, which is similar to the results for solid-state detectors (Fig. 15); however, they
were achieved in a time six times shorter than before. These low detection limits were
achieved using dierent proportional-counter gas llings, and, as a consequence, the
technique is less exible than that based on the solid-state detector. If adjacent Z elements
had been present, the detection limits for the proportional counter would have been
considerably worse, but for the solid-state detector, they would have been much less
changed.
Copyright 2002 Marcel Dekker, Inc.
Figure 16 Spectrum of x-rays from a pure water sample and from water spiked with 100 mg/g of
iron and zinc as determined using a (Ne Ar)-filled proportional detector and excited by x-rays from
a 244Cm source. (From Jarvinen and Sipila, 1984b.)
1. Filters
Filters placed between the sample and detector (Fig. 18) increase the sensitivity of analysis
by ltering out a higher proportion of uorescent x-rays of matrix elements than those of
the analyte. Zinc, for example, may be the analyte in samples also containing iron. Cal-
culations based on Eq. (1) and mass absorption coecients [see Eq. (76) in Chapter 1]
show that a 27-mg=cm2 aluminum lter transmits 27% of the Ka x-rays of 8.6 keV but
only 4.5% of the iron Ka x-rays of 6.4 keV (Fig. 19); that is, it reduces the intensity of iron
K x-rays relative to the zinc K x-rays by a factor of 6. If the sample also contains lead,
however, about 55% of the lead La x-rays (10.5 keV) would be transmitted, twice that of
the zinc Ka x-rays. In this case (Fig. 19), an absorption-edge-type lter of copper (22.4 mg/
cm2) could be used to reduce the lead La x-ray transmission to only 1 % and also to reduce
the iron K x-ray peak. This selective enhancement of the zinc compared with the iron and
lead x-ray components partly compensates for the limitation of the poor resolution of the
scintillation detector. Although some iron and lead uorescent x-rays will still be detected
within the pulse-height channel set about zinc Ka x-ray peak, in many applications this
component will have been suciently reduced to make the analysis possible.
If the measurement with one absorption-edge lter does not give sucient selectivity
to the desired elements uorescent x-rays, balanced lters are used (Fig. 20). The in-
tensities of x-rays in the uorescent x-ray channel are measured separately, rst with one
lter and then with the other (with up and down lter). The atomic numbers of the
two lters are chosen so that their K shell absorption-edge energies are just above and
below the energy of the Ka x-rays of the analyte. The masses per unit area of the two lters
are chosen so that the product mrt [Eq. (1)] is practically identical for both lters, except
Figure 20 X-ray transmission through balanced filters of copper (6) and nickel (solid circle).
Count-rate measurements made first with one and then with the other filter are subtracted to give a
count rate proportional to the intensity of zinc Ka x-rays from a sample.
2. DirectExcitation Assemblies
The directexcitation technique (Fig. 18) is the most widely used of the three scintillation
detector assemblies. The energy of the radioisotope x-ray is usually chosen so that the
uorescent and backscattered x-rays are resolved by energy analysis (Fig. 21). The in-
tensity of the backscattered x-rays is used to correct for the absorption of the analytes
uorescent x-rays by the sample matrix [Eq. (8) or (9)]. The lter enhances the sensitivity
and selectivity of analysis.
Direct-excitation assemblies are used extensively in industry (e.g., in laboratory and
portable elemental analyzers (Rhodes, 1971) and in on-line (in-stream) analysis of mineral
slurries (Watt, 1983).
3. DetectorRadiator Assemblies
The detectorradiator assembly (Fig. 18) discriminated well against interfering x-rays of
energy just below that of the uorescent x-rays of the measured element (Watt, 1972). The
Copyright 2002 Marcel Dekker, Inc.
basis of this discrimination is that the atomic number of the radiator element can be
chosen so that, of the two x-ray components with nearly similar energies, only the higher
of the two has sucient energy to excite the K x-rays of the radiator element. The detector
is shielded from the sample and, hence, sees only the x-rays emitted by the radiator. Ba-
lanced-radiator techniques, analogous to balanced-lter techniques, can also be used to
improve selectivity to the analyte if there is another matrix element, in this case emitting
uorescent x-rays of energy higher than those of the analyte. The count rates obtained
with detectorradiator assemblies are about 5% of those obtained with direct-excitation
assemblies, using a source of the same activity, because of the additional excitation stage
of the radiator.
The intensity of higher-energy x-rays scattered by the sample can be measured si-
multaneously in the one assembly by use of a second radiator element of atomic number
considerably higher than that of the rst radiator. The x-ray energies of the two com-
ponents are well resolved and similar to that shown in Figure 21.
Detectorradiator systems are much less widely used than balanced-lter techni-
ques. They are less versatile than balanced-lter systems. Applications include the de-
termination of lead in zinc concentrates, in which the zinc Ka x-rays (8.6 keV) from the
high concentration of zinc (e.g., 50 wt%) swamp the lead L x-rays (10.514.8 keV) from
the low concentration of lead (e.g., 0.5 wt%). A radiator of zinc (absorption-edge energy
of 9.66 keV) is excited by the lead L but not by the zinc K x-rays. This radiator technique
238
Figure 21 Typical spectrum of x-rays from copper ore slurry excited by a Pu source and
measured using a scintillation detector.
The choice of a radioisotope x-ray analysis technique for a specic application depends on
several interacting factors: the overall accuracy of the sampling and analysis required, the
time available to achieve this accuracy, technique available to obtain a suciently
representative sample of the material being analyzed, and the sample preparation
Copyright 2002 Marcel Dekker, Inc.
requirements, such as grinding, pressing, and so forth. The inuence of each of these
factors on the accuracy of XRF analysis is discussed here.
where
sinstr Instrumental component such as thermal drift
sstat Uncertainty due to statistics of counting
shet Uncertainty reecting sample heterogeneity
scal Reects uncertainties contributed by reference analysis and calibra-
tion t
ssamp:plac: Uncertainty due to sample placement
smatrx Uncertainty associated with matrix eects
spart: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 stot can
be regarded as a good measure of the accuracy of analysis.
The rst two uncertainty components, sinstr and sstat , 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:
q
stot s2instr:precis: s2applic: 23
where
q
sinstr:precis: s2instr s2stat 24
and
q
sapplic: s2het s2cal s2samp:plac: s2matrx s2part:size s2?? 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, sstat , can be easily reduced by extending
Copyright 2002 Marcel Dekker, Inc.
the measurement time, whereas the uncertainty associated with calibration, scal , can be
reduced by using more accurately analyzed calibration samples and by selection of a better
calibration equation.
Sample placement uncertainties, ssamp: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, smatrx , and particle size, spart: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.
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.
Table 4 Some Typical Applications of Radioisotope Based Laboratory and Portable
Elemental Analyzers
Alloy sorting and Low-alloy steels; stainless steels; nickel alloys; high-temperature alloys;
identification 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 Copper, zinc, tin, gold, silver and chromium plating thicknesses; metals
coatings in plating solutions; silver in photographic film; manganese coating
thickness on magnetic tape; titanium on glass; ruthenium on electrodes
Chemicals and Lead, titanium, and zinc in paint; sulfur, iron, alumina, silica, and calcium
process control 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; TiO2 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
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) 109Cd 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
Copyright 2002 Marcel Dekker, Inc.
Table 5 Manufacturers and Suppliers of Radioisotope-Based Laboratory and Portable
Elemental Analyzers
Figure 23 The Metorex Int. portable alloy analyzer based on high-resolution silicon p-i-n diode
detector, MetalMasterTM 2000. (Courtesy Metorex Int.)
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
Copyright 2002 Marcel Dekker, Inc.
Table 6 On-line Analysis Systems Based on Radioisotope X-ray Sourcesa
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.
Copyright 2002 Marcel Dekker, Inc.
Table 7 Performance of a Portable X-ray Analyzer in Alloy Identication
Identication results
Alloy group Measured elements (% 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 Ti, Cr, Mn, Fe, Co, Ni, Cu, Nb, Mo 90100; 100
high-temperature alloys
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.
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) 109Cd and 1.48-GBq (40 mCi) 55Fe sources.
Source: From Piorek et al., 1997.
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 SiO2 and Al2O3 with smaller con-
centrations of Fe2O3) 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 238Pu (or 244Cm) 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 Fe2O3 compared with Al2O3 and SiO2. 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 Fe2O3 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 241Am (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
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=cm2 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).
routinely by one of two radioisotope x-ray techniques. One technique combines XRF
(55Fe 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) 55Fe 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).
Copyright 2002 Marcel Dekker, Inc.
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.)
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
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.)
VIII. CONCLUSIONS
Figure 32 Niton II, XRF Alloy Analyzer based on a silicon p-i-n diode. (Courtesy Niton Corp.)
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