Surface Characterizatioti: A User3 Sourcebook

Edited by D Brune, R Hellborg, H J Whitlow & 0 Hunderi

copyright&)WILEY-VCH Vcrlag Grnh11.19Y7

9 Particle-induced X-ray emission and particle-

induced gamma ray emission
K.G. Malmqvist

9.1 Introduction
The development of high resolution, semiconductor detectors in the late 1960s
meant a giant step forward in both X-ray and gamma-ray spectroscopy. The signifi-
cantly improved spectral resolution in both X-ray and gamma-ray spectra facilitated the
development of both Particle-Induced X-ray Emission (PIXE) and Particle-Induced
Gamma Emission (PIGE). Today, PIXE and PIGE analysis are often carried out
simultaneously and the set-ups are designed to accommodate the relevant detectors and
other equipment required for this combination. The accelerators used to produce the
suitable ion species and energies are normally of the electrostatic type. For more details
about suitable accelerators for ion-beam analysis techniques refer to the chapter by
Whitlow and Ostling. To understand both PIXE and PIGE the details of the interaction
between the impinging ions and matter are crucial (refer to the chapter by Whitlow and
Ostling). When a beam of ions penetrates matter, the ions gradualiy lose their energy
with depth, until they finally stop within a well-defined range. The cross-sections for
atomic interactions (PIXE) vary slowly with ion energy while nuclear reac-
tions/excitation (PIGE) may vary rapidly (resonant) and in a less predictable manner.
The genuinely complementary characteristics of the two techniques are based on the
overlapping elemental ranges of analysis. The use of nuclear reactions is facilitated by
a low Coulomb barrier, i.e., a light ion in combination with a light (2120) target
nucleus. The soft X-rays from low-Z elements makes PIXE suitable for analysis mainly
of elements with 2212. If employed in a common experimental set-up, it is thus possi-
ble to measure both light and heavy elements and to compare independent results from
both techniques for quality assurance.
Each of the two techniques or a combination of the two have been used extensively
in many fields of application, for instance, geology, medicine, environmental studies,
materials science, archaeology, etc.

9.2 Particle-induced X-ray emission

9.2.1 Basic principles

In 1970 it was demonstrated (Johansson et al., 1970) that the bombardment of a
specimen with protons with an energy of a few MeV producing characteristic X-rays
 9 PIXE and PIGE 155

can form the basis for a highly sensitive elemental analysis. This pioneering paper
formed the starting point of the X-ray emission analysis technique, which became
known as particle-induced X-ray emission analysis. As in X-ray fluorescence (XRF)
and electron probe microanalysis (EPMA), the characteristic X-rays produced in the
de-excitation of the atoms can be measured by either a wavelength-dispersive or an
energy-dispersive detection system. In PIXE, an energy-dispersive spectrometer is used
almost exclusively.

a)

-1
W
Z
z - Ep 10 keV
1
4
I
u F i I 1 I 1 I I I
oc
w 100 200 300 400

P
b)
lo3
PROTON- MICROPROBE

1 o2

10

1
Fig. 9-1. X-ray spectra from
analysis of the same organic
100 200 300 40G specimen using electrons (a) and
__+ CHANNEL N U M B E R protons (b) (Bosch et al.,1980).

In contrast to X-rays, ions and electrons can be focused by electrostatic or magnetic

lenses and may hence be transported over large distances without loss of beam inten-
sity. As a result, incident beam densities are generally much higher in PIXE than in
ordinary tube-excited XRF.Moreover, focusing of particle beams down to micrometer
dimensions is possible, so that PIXE analysis can be carried out with high lateral reso-
lution. This variant of PIXE, micro-PIXE, is used in a nuclear microprobe (see Section
9.2.3). With electrons, focusing to nanometer dimensions is possible and has given rise
to the widespread EPMA technique (see chapter by Pettersson and Selin-Lindgren), but
heavy charged particles have the clear advantage that they give rise to much lower
156 Part 2: Elemental composition

continuous background intensity in the X-ray spectrum. As a result, the relative detec-
tion limits (micrograms per gram) are typically two orders of magnitude better in
micro-PIXE than in EPMA (see Fig. 9-1).
After the initial experiments the favourable characteristics of PIXE were rapidly re-
alized by many researchers. Besides the traditional bombardment in vacuum, external
beam approaches were also attempted and were found to be useful, particularly in ex-
amining delicate and/or large objects. The progress of PIXE over the years is demon-
strated by the proceedings of seven international conferences exclusively dedicated to
the PIXE technique and its applications. Today, PIXE has evolved into a mature tech-
nique, as demonstrated by the increasing numbers of research papers in which PIXE
provided the analytical results and by two textbooks on the technique (Johansson and
Campbell, 1988; Johansson et al., 1995).

Ionization and X-ray production cross-sections

Protons or heavier ions of a few MeV/u energy have a high probability of causing
the ejection of inner shell electrons. Various theoretical approaches have been used to
calculate the cross-sections for inner-shell vacancy creation. The plane wave Born
approximation (PWBA) model has been most elaborated. Since then, a series of modi-
fications in the model to correct for its inherent approximations have been incorpo-
rated, resulting in the ECPSSR treatment (Brandt and Lapicki, 1981) of K- and L-shell
ionization cross-sections. For K-shell ionisation with protons, the cross-sections as
predicted by the ECPSSR and other theories have been thoroughly compared with
experimental data. Although no theory has emerged that will predict the experimental
data within a few percent for all target elements and energies, it is generally agreed that
existing data tables are adequate for most K-shell proton PIXE work. In the case of the
L ionization, the situation is less favourable, however, as discussed in detail by
Campbell, 1988.
The probability of X-ray production is also dependent on the fluorescence yield w
which expresses the ratio between X-rays and Auger electrons resulting from a shell
vacancy. The fluorescence yield varies strongly with atomic number of the target atom.
For the K-shell, w varies from below 0.1 at Z<20 to above 0.9 for Z>50.

Continuous background production

The characteristic X-ray lines in a PIXE spectrum are superimposed on a continuous
background (see Fig. 9-1). It has been shown that the electron bremsstrahlung in PIXE
originates essentially from three processes: quasi-free-electron bremsstrahlung
(QFEB), secondary-electron bremsstrahlung (SEB), and atomic bremsstrahlung (AB).
SEB is formed by a two-step process: the incident particle first ejects an electron from
a target atom, and the secondary electron is subsequently scattered in the Coulomb
field of a target nucleus, thus producing the bremsstrahlung. The photon spectrum of
SEB is characterized by an end-point energy:
T, = 4m,E,N, (9-1)
 9 PIXE and PICE 157

where me and M, are the electron and projectile masses and E, the projectile energy.
Above T, the intensity of SEB decreases rapidly. QFEB is emitted when an electron of
a target atom is scattered by the Coulomb field of the projectile. The QFEB end-point
energy T, is equal to T,/4. The process AB occurs when a bound target electron is ex-
cited to a continuum state by the projectile and, returning to its original state, emits a
photon. AB predominates in the high-energy part of the spectrum whereas QFEB be-
comes the prevailing component at low photon energies (below Tr). Besides E , the
matrix composition of the target also plays a critical role in both the shape and inten-
sity of the electron bremsstrahlung background. The background becomes more intense
with increasing average target Z. Furthermore, the electron bremsstrahlung is emitted
anisotropically with a maximum intensity at an angle of 90". For this reason, in a typi-
cal experimental PIXE chamber, the X-ray detector is positioned at an angle of 135"
relative to the beam direction. For a general overview of background production in
PIXE spectra refer to Ishii and Morita (1 990).

Detection limits and basic formalism

In many cases, intense characteristic X-ray peaks are present in the PIXE spectrum.
This is particularly true when the matrix consists of element(s) with atomic number
above Z 2 1 1. These intense peaks and their escape peaks and pile-up peaks seriously
hamper or preclude the detection of other elements with characteristic X-rays of similar
energies. Furthermore, incomplete charge collection in the detector and other processes
have the effect that each X-ray peak exhibits a low-energy tail. Consequently, tails as-
sociated with very intense peaks form a substantial component of the total background.
The limit of detection (LOD) is determined by the relationship

N, 2 3 & (9-2)
where N, is the number of counts in a characteristic X-ray peak and NR is the number
of background counts in the corresponding interval ( & x one standard deviation).
For a given proton-induced, characteristic X-ray emitted by an element, its yield can
be calculated from the formula:

where the photon attenuation:

and Y,=X-ray yield of element p, C = relative concentration by weight of element p,

No= Avogadro's number, R = solid angle subtended by the detector, Q = total beam
charge hitting the target, E = detector efficiency, including filter effects, W = atomic
158 Part 2: Elemental composition

weight of the element, e = ion charge, Eo = incident proton energy, Ef final proton en-
ergy, 0;( E ) = X-ray production cross-section for a peak p, p = mass attenuation coef-
ficient, S(E) = matrix stopping power calculated using Bragg's rule, and 4 and
lo= angle of the incoming beam and the outgoing X-ray with respect to the normal to
the surface of the target.

Target thickness
Eq. 9-3 is a general expression, allowing for calculation of the yield of a charac-
teristic X-ray emitted by an element of interest in all situations, i.e. for thin, inter-
mediate or thick targets. Targets of intermediate thickness are those for which beam
particles pass through the sample, but their energy loss cannot be neglected. For two
extreme cases, simplifications are possible. For very thin targets, projectile energy
losses in the sample are negligible and the energy-dependent terms in the integral over
dE in eq. 9-3 are constant. The X-ray absorption for the emerging X-ray may also be
neglected. Hence, the dependence of yield of a characteristic X-ray is a linear function
of the element concentration and no matrix information is necessary. However, prepa-
ration of thin samples is often difficult and in many cases impossible and many appar-
ently thin samples violate the thin specimen criterion for protons and even more so for
heavier ions.
In the second extreme case of the X-ray yield equation, the particle beam comes to a
complete stop in the sample, i.e. the final projectile energy in eq. 9-3 becomes zero.
Thus it is easier to calculate correctly the integral in eq. 9-3. The possibility of analys-
ing infinitely thick samples in an essentially virgin state is one of the major advantages
of PIXE. The only two requirements for thick target preparation are a smooth surface
and sample uniformity. The required smoothness depends on the X-rays to be meas-
ured, the degree of their absorption in the matrix and the accuracy to be achieved
(Campbell and Cookson, 1984).

9.2.2 Quantification
In PIXE, elements with Z up to about 50 are generally determined by their K X-rays
and the heavier elements are measured by their L X-rays. The basis for a quantitative
analysis is that there is a relationship between the net area of an element's characteristic
K or L X-ray line in the PIXE spectrum and the amount of element present in the sam-
ple as given by eq. 9-3 above. To derive the concentration from its X-ray yield, several
approaches are possible. One can solve eq. 9-3 and thus employ the absolute or funda-
mental parameter method. This requires accurate knowledge of all the parameters
involved. Because of the difficulties with the absolute quantification method, many
PIXE workers prefer to rely on a relative approach, and they calibrate their experi-
mental PIXE set up using thin-film standards.
In the analysis of infinitely 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 (Johansson et al., 1981). The thick-target cali-
 9 PIXE and PIGE 159

bration factors incorporate the integral of eq. 9-3 and are usually expressed in X-ray
counts per pC and per pg/g. They are commonly derived from PIXE measurements on
standard samples with known matrix and trace element composition. In a strict sense,
the thick-target factors are valid only for the analysis of unknown samples with a com-
position identical to the standards, but in practice, some variability in composition can
be tolerated or corrected for.

9.2.3 Precision and accuracy

As in any other analytical technique, high precision and accuracy should be aimed
for in PIXE. It is therefore essential that careful attention is given to all stages of the
analysis. These include sample and specimen preparation, specimen bombardment,
spectral data processing, quantification and correction for matrix effects. During the
specimen bombardment stage, great care has to be taken to minimize radiation- or
heat-induced losses. Such losses are particularly critical for volatile trace elements and,
in the case of organic or biological specimens, also for certain matrix elements (mainly
H and 0).The current density applied during analysis plays an important role but the
dose deposited is even more crucial. The danger or losses is more severe in micro-
PIXE than in macro-PIXE. More information on this subject can be found in a tutorial
paper by Maenhaut (1990). In routine PIXE analysis it is possible that the accuracy and
precision can drop to the order of 90-95%.

9.2.4 Instrumentation

PlXE in vacuum
Fig. 9-2 shows the design and principal components of a hypothetical, typical set-
up for PIXE analysis in vacuum combined with other IBA methods. In the following,
the details of such a system are outlined.
A typical experimental facility for PIXE analysis employs a system for producing an
ion beam which irradiates the specimen homogeneously. This is required for quantita-
tive analysis of heterogeneous samples and can be achieved, for instance, by focusing
the beam on to a scattering foil. The homogenized beam is defined by a pair of colli-
mators, normally circular, with diameters between 1 and 10 mm.
The spectral shape in PIXE can be advantageously modulated by placing an X-ray
absorber between sample and detector. When measuring light elements (Z111) the
X-rays are soft and only a very thin absorber can be used between the target and the
detector. Charged particles back-scattered from the target can then strike the Si(Li)
detector and cause problems.
The X-ray detectors used for PIXE typically have an effective area from 10 to
80 mm2.The amplifiers and pulse processors in energy-dispersive X-ray spectroscopy
require long time constants for optimum energy resolution. This implies that pulse pile-
up can become a serious problem at relatively low count rates.
160 Part 2: Elemental composition

When analysing thin specimens, the ions pass through and are dumped in a Faraday
cup for charge integration. When the samples are thick enough to stop the ions, the
beam current must be measured either on the whole irradiation chamber or through
some indirect approach.
As will be discussed below PIXE target chambers often also include gamma-ray de-
tectors for measuring prompt y-rays necessary for complementary P E E (see Fig. 9-2).

1 Faradaycup c ion beam

collimators

gamma ray detector

Fig. 9-2. Schematic experimental arrangement in a typical system for PIXE and complementary IBA
methods, including data acquisition system.

External beam PIXE

It is sometimes advantageous to use atmospheric pressure or moderate vacuum in-
stead of high vacuum in ion-beam analysis. The heat conductivity is increased, the tar-
get temperature is decreased, the charge is conduced from the target, and the vacuum
requirements are less, thus making it easier to design a low-cost chamber or to irradiate
very large specimens. Because the accelerator requires a high vacuum, the ion beam
must be extracted into the region of moderate vacuum or atmospheric pressure either
through a thin exit foil or through a narrow orifice combined with differential pump-
ing. The beam eventually causes deteriorates of the foil, and therefore its material must
be carefully selected. A good choice is the commercially available foil Kapton@,which
withstands high intensities and a high radiation dose before breakdown. Supporting the
foil by a carbon grid and direct flow of liquid nitrogen-cooled helium gas allows the
 9 PIXE and PIGE 16 1

use of high beam intensities over extended time intervals (Hyvonen-Dabek et al.,
1982). The chamber gas is normally helium or nitrogen. In addition to its better cooling
properties, helium produces less bremsstrahlung and causes less X-ray attenuation than
nitrogen.
An external beam in air is needed when large objects are to be analysed without sec-
tioning or sub-sampling. This is very important for such objects as those of interest in
archaeology and history of art, for example. The strong argon K X-ray line produced in
air can be used to monitor the beam charge; this is otherwise difficult to do in non-
vacuum PIXE.

Micro PIXE - the nuclear microprobe

In the instrument called a nuclear microprobe (Watt and Grime, 1995) PIXE analy-
sis is the most important analytical technique. It is often called micro-PIXE. With the
equipment used in regular, accelerator-based ion beam analysis, beam sizes down to
typically a few tenths of a mm in diameter are easily attained. In the nuclear micro-
probe the ion beam is collimated and/or focused down to dimensions in the range
1-50 pm.
The simplest way of producing such a microprobe is to employ a pinhole collimator.
For very small collimator sizes, however, the beam intensity obtainable is much too
low for practical use. In addition, a substantial fraction of the ions is scattered at the
edge of the collimator, and this gives rise to a halo around the central beam. In most
nuclear microprobe systems, fine collimation is therefore combined with an electro-
static or magnetic demagnification system, such as doublets, triplets, or quadruplets of
quadruples. The best systems currently available reach a spatial resolution of about
0.5 pm at the specimen while maintaining an ion current that is useful for PIXE
analysis (>lo0 PA). For more details refer to the chapter by Lindh.

9.2.5 Data Acquisition

The data acquisition systems used in PIXE have much in common with the data
acquisition systems used in energy-dispersive X-ray analysis in EPMA. The signals
from the X-ray detector are converted in analogue-to-digital converters and then stored
in computer (PC)-based multi-channel analysers (see Fig. 9-2). The X-ray spectra are
then evaluated by sophisticated, spectrum-fitting-computer codes, normally yielding
quantitative results (Maxwell et al., 1989; Ryan et al., 1990).
Once a PIXE spectrum has been acquired, the first step in the quantification is the
extraction of the net peak intensities for the elements of interest. By far the most com-
mon approach to spectrum analysis in PIXE is to model the spectrum by an analytical
function. This function includes modified Gaussians to describe the characteristic
X-ray peaks and a polynomial or exponential polynomial to represent the underlying
continuum background. Because of the low continuum background in PIXE, the range
of peak heights in a PIXE spectrum can be up to five to six orders of magnitude. This
leads to PIXE spectra that often exhibit some details, such as escape and sum peaks,
162 Part 2: Elemental composition

and low-energy tailing for intense peaks. Despite the many fine details, accurate mod-
elling of PIXE spectra is quite feasible, as was demonstrated in an inter-comparison
exercise of five different PIXE spectrum analysis programs (Campbell et al., 1986).

9.3 Particle-induced gamma ray emission spectrometry

9.3.1 Basic principle

In the definition of particle-induced gamma-ray emission, it is rather difficult to
completely separate it from nuclear reaction analysis in general. For details, refer to the
chapter by Whitlow and Hellborg. In this context, PIGE is limited to the detection of
prompt gamma rays produced by ion bombardment with particles at MeV energies of a
specimen homogeneous in depth, i.e., the resonant reactions are not used for depth pro-
filing. In principle, any ion with appropriate bombarding energy can be used to induce
gamma-ray emission. For the reasons discussed above, protons and alpha particles are
mostly used, but a few experiments have been done using other light ions such as deu-
terons, tritons, 3He, N and F. Because of the interaction with the atomic nucleus, the
cross sections are lower than for PIXE, and PIGE typically represents a less sensitive
analytical technique than PIXE. However, the gamma-ray peaks are generally well
isolated and the energy of the emitted gamma-ray photons is high enough that correc-
tion for absorption is not necessary. Hence PIGE may in some situations be more
favourable than PIXE for elements such as calcium and potassium (Uzunov et al.,
1995). The high penetrability of gamma rays also simplifies the experimental arrange-
ments and the nuclear interaction facilitates also the gaining of isotopic information.
The main experimental components for PIGE analysis are illustrated in Fig. 9-2,
showing the combined experimental facility for various IBA techniques. External
beams are now also a standard in conjunction with PIGE (Raisinen, 1989). The exter-
nal beams offer the same advantages as discussed above for the case of PIXE. It is
rather unusual to include PIGE as an option for nuclear microprobes. However, occa-
sionally PIGE has been applied with nuclear reactions of large cross sections, for
instance. those of fluorine.

9.3.2 Thick target gamma-ray yields

More or less systematic measurements of gamma-ray yield vs bombarding energy in
proton or alpha bombardment have been made at several laboratories. The main source
of error in determining these absolute yields is the stopping power. The uncertainty can
be as much as 20% in the case of compounds. On the other hand, these absolute yields
are normally not used for high accuracy analysis since comparison with proper stan-
dards with a similar composition to that of the sample is generally possible.
The gamma-ray peaks from certain reactions are significantly broader than the reso-
lution of a semiconductor detector. This broadening is due mainly to the short lifetimes
 9 PIXE and PIGE 163

of de-excitation states and the high recoil of the light nucleus (Doppler effect) or due to
the large resonance width. The broadening of peaks is useful in identifying the origin
of the gamma rays, especially for elements which have only one peak.

9.3.3 Quantification
Homogeneous concentrations of elements in thick samples, detected by prompt
gamma-ray emission, can be obtained by comparison with standards. Many multi-
component, elemental standards exist for the analysis of geological, biological and
medical samples.
The concentration of the element (isotope) C, can be calculated from the gamma-ray
yield for a certain gamma-ray peak in sample Y, and the yield from a standard YSt
using this simple formula

where the stopping powers Si (Ziegler, 1992) have to be calculated at the energy E1,2,
where the thick target yield has fallen to half of its value at the bombarding energy Eb.
If Eb is used instead of E1/2, the error is less than a few percent. If the excitation curve
vs bombarding energy is known, a more accurate calculation can be made using a sim-
ple computer program, although this is not generally necessary.
Even if the relative method described above is used for quantification of PIGE
measurements the same precautions as described above for PIXE are crucial. The
losses of light elements occurring during intense and large dose deposition, in particu-
lar in organic matrices, has to be accounted for.

9.3.4 Sensitivity
Examples of sensitivities obtainable in PIGE using protons and alpha particles are
given in Table 9-1 a and b. These values should be taken as order of magnitude values,
because the actual sensitivities are better than those given in the tables. This is true es-
pecially when experimental arrangements and the bombarding energy are optimized for
the detection of a specific element. On the other hand, sensitivities can be much worse
if the sample contains major amounts of other elements with high gamma-ray yield or
if the sample can withstand only low beam currents. Therefore, possible interference
from both radioactive and prompt background, as well as from other sample
components, should be investigated.
The sensitivity of protons and helium ions for elements of medium mass number is
somewhat better with higher bombarding energies. This fact, and problems with spec-
tral interference in PIXE measurements, make the PIGE method preferable for the
analysis of samples with several components typical of steel. At higher bombarding
164 Part 2: Elemental composition

energies, 4He+ions can be used also for the detection of heavier elements, with detec-
tion limits down to 0.01-0.1% (Giles and Peisach, 1979).

Table 9-la. Detection limits (atomic fraction) for PICE using protons (E,<9 MeV) (Antilla et al., 1981)
>I% 0.1 - I % < 0.1 Yo
Pd, Sm. Gd, Hf, W, Au, Pb S, K, Sc, Ti, Co, Cu, Ge, Y, Zr, Mo, Li, Be, B, C, N,Na, Mg, Al, Si, C1,
Ru, Ag, Sn, I, Ta, Pt Ca, V, Mn, Fe, Ni, Zn, Nb, Cd, In,

Table 9-lb. Detection limits (atomic fraction) for PIGE using alpha particles ( 5 MeV) (Gilles and Pei-
sach, 1979).
> 1 Yo 0.1 - 1 % <O.l Yo
Sc, Cr, Cu, Zn, Rb, Zr, Er, Hf, 0, Mg, Si, C1, K, Ti, Fe, Br, Mo, Li, B, N, F, Na, Al, P, V, Mn, Rh
Hg Ru, Pd, Ag, Cd, Ta, W, Re, Ir, Pt,

9.4 Sample collection and preparation

9.4.1 Macrobeam analysis

PIXE and PIGE analysis can in principle be applied to any type of sample. How-
ever, it is clear that IBA techniques are more suitable for analysing solids than liquids
since most irradiation takes place in vacuum. Ion beam analysis of liquids normally
involves some pre-concentration by drying or some other physical or chemical separa-
tion of the relevant elements from the liquid phase. As far as the analysis of infinitely
thick samples is concerned, it should be kept in mind that for both PIXE and PIGE the
mass actually probed by the beam is at most a few mg. Determination of the bulk com-
position of a solid sample without preliminary sapple preparation is therefore possible
only for samples that are homogeneous in all three dimensions. However, there are
numerous analytical problems for which PIXE, PIGE or a combination of them is the
most suitable technique or at least among the most suitable. Examples are the multi-
elemental analysis of milligram-sized samples consisting of a light-element matrix, the
non-destructive analysis of micro- to millimetre-sized areas on a large sample or of
thin, superficial layers on a bulk sample, and various problems that require sensitive
analysis with high lateral resolution.
Once PIXE and/or PIGE has been selected, full use should be made of the inherent
characteristics of the techniques, particularly of their non-destructive and instrumental
character. Therefore, if possible, sampling should be done in such a way that subse-
quent sample preparation can be avoided or kept to a strict minimum. For unique sam-
ples or samples of high commercial or historical value, sample preparation or sub-
sampling may not even be allowed because generally the sample must be returned