You are on page 1of 6

SPECTROCHIMICA

ACTA
PART B

ELSEVIER SpectrochimicaActa Part B 53 (1998) 43-48

A comparison of the performance of a fundamental parameter method


for analysis of total reflection X-ray fluorescence spectra and
determination of trace elements, versus an empirical
quantification procedure
Dariusz W~grzynek, Barbara Hotyfiska*, Beata Ostachowicz
Faculty of Physics and Nuclear Techniques, University of Mining and Metallurgy, AI. Mickiewicza 30, 30-059 Krakdw, Poland
Received 31 July 1997; revised 11 November 1997; accepted 20 November 1997

Abstract

The performance has been compared of two different quantification methods - - namely, the commonly used empirical
quantification procedure and a fundamental parameter approach - - for determination of the mass fractions of elements in
particulate-like sample residues on a quartz reflector measured in the total reflection geometry. In the empirical quantification
procedure, the spectrometer system needs to be calibrated with the use of samples containing known concentrations of the
elements. On the basis of intensities of the X-ray peaks and the known concentration or mass fraction of an internal standard
element, by using relative sensitivities of the spectrometer system the concentrations or mass fractions of the elements are
calculated. The fundamental parameter approach does not require any calibration of the spectrometer system to be carried out.
However, in order to account for an unknown mass per unit area of a sample and sample nonuniformity, an internal standard
element is added. The concentrations/mass fractions of the elements to be determined are calculated during fitting a modelled
X-ray spectrum to the measured one. The two quantification methods were applied to determine the mass fractions of elements
in the cross-sections of a peat core, biological standard reference materials and to determine the concentrations of elements in
samples prepared from an aqueous multi-element standard solution. © 1998 Elsevier Science B.V.

Keywords: Total reflection X-ray fluorescence; Trace element determination; Spectra fitting; Fundamental parameter method

I. Introduction spectrometry (FAAS), electro-thermal atomic absorp-


tion spectrometry (ET-AAS) and inductively coupled
Nowadays, total reflection X-ray fluorescence plasma atomic emission spectrometry (ICP-AES) [1].
(TXRF) spectrometry is applied extensively for the For most o f the elements, because o f very low mass
determination o f trace elements in various kinds o f per unit area o f the sample, the absorption and
sample. The capabilities o f the TXRF method can enhancement effects can be neglected, making the
compete with other advanced instrumental analytical quantification procedure fast and simple. In spite o f
techniques such as flame atomic absorption these advantages, still there is room for improvements
in spectrometer design as well as in quantification
* Corresponding author. procedures. In this paper we present a fundamental

0584-8547/98/$19.00 © 1998 Elsevier Science B.V. All rights reserved


Pll S0584-8547(97)00129-8
44 D. Wqgrzyneket aL/Spectrochimica Acta Part B 53 (1998) 43-48

parameter approach applied to the determination of measurements, the concentrations/mass fractions of


trace elements by the TXRF technique. A detailed the elements in unknown samples are determined.
description of this approach has been given Before applying the procedure described above one
elsewhere [2]. In the present paper, a comparison of has to deconvolute the measured X-ray spectra of both
the fundamental parameter (FP) method and the the investigated and standard samples in order to
empirical quantification (EQ) procedure is given. obtain the net intensities of the characteristic peaks.
Both methods were applied for the determination of In this study, the Quantitative X-Ray Analysis System
elements in samples prepared from peat core layers, (QXAS) software package [4] including the AXIL
biological standard reference materials, and an program [5] has been used to perform all of the
aqueous solution. The comparison of both quantifi- calculations necessary in the empirical quantification
cation methods was performed on samples selected procedure.
in such a way they contained elements giving strong
interferences in the X-ray spectra collected, namely 2.2. Fundamental parameter method
titanium-barium and arsenic-lead, and were also
abundant in low-Z elements such as potassium and The fundamental parameter approach used applies
calcium. a theoretical model of the acquired X-ray fluorescence
spectrum given as a function of the concentrations/
mass fractions of elements present in the analysed
2. Theoretical sample [2]. The concentrations/mass fractions of
the elements to be determined are obtained by fitting
2.1. Empirical quantification method the model to the acquired X-ray fluorescence spec-
trum of a sample. The modelling function takes
A commonly used empirical quantification proce- into account measurement conditions, modification
dure requires the spectrometer set-up to be properly of the excitation radiation on its way to the sample,
calibrated with the use of samples prepared from detector efficiency, low-energy tailing of X-ray
standard solutions [3]. In the process of calibration, peaks, natural widths of the characteristic lines, the
the ratios of elemental sensitivities to that of an presence of escape peaks and the presence of peaks
internal standard element are calculated and relative resulting from the most intense radiative Auger
sensitivity curves are obtained. transitions [6-10]. Since the sample analysed is
It should be noted that these relative sensitivity assumed to form a thin, particle-like residue on a
curves have to be established from measurements quartz reflector, absorption and enhancement correc-
of standard samples containing two groups of ele- tion factors are not included in the model. The avail-
ments: i.e., elements detected by their K-series and able fundamental parameter data allow the procedure
elements detected by their L-series characteristic to be used for the determination of elements with
radiation [3]. Both groups of elements have to be atomic number from Z = 11 to Z = 92 [11-14]. The
present in calibration standards. Owing to different concentrations or the mass fractions of the elements
values of fluorescence yields, photoelectric cross- are determined during a spectra-fitting procedure
sections and other parameters for the K- and L-series, taking into account all of the intensities of K-series
the relative sensitivities of the elements determined and/or L-series characteristic peaks of elements
with the use of K-series peaks and L-series peaks present in the acquired spectrum. In this approach
form two distinct calibration curves. For example, an internal standard element has still to be added
the analysis of mercury, an element detected by to the sample before measurement, to account for
its L-series radiation, could not be performed if the unknown mass per unit area of the sample and
the set of the calibration standards used to obtain irreproducible measurement geometry. Determina-
the relative elemental sensitivities contained only tion of the concentrations/mass fractions of ele-
elements giving K-series characteristic peaks. ments by the fundamental parameter method was
Knowing the concentration or mass fraction of an performed with the use of a home-made computer
internal standard, added to the sample prior to the program.
D. Wggrzynek et al./Spectrochimica Acta Part B 53 (1998) 43-48 45

3. Experimental Bamsted/Thermolyne). The volumes of samples


pipetted onto quartz reflectors were optimised to
3.1. Instrumentation and experimental arrangement keep the total count rate at moderate levels and to
prevent distortion of the X-ray spectra collected.
All measurements were performed with a modified Three groups of samples were prepared: i.e., cross-
TXRF module, constructed in the Atominstitut der sections taken from a peat core, biological standard
Ostereichischen Universit~iten [15,16], attached to a reference materials and aqueous multi-element
molybdenum-anode long fine-focus X-ray tube standard solutions.
(Philips PW 2275/20). A 50/zm molybdenum filter Two peat core layers, previously freeze-dried and
was placed in the front of the X-ray tube window. powdered, were taken for analysis. The samples,
The tube was operated at 55 kV and 40 mA. The weighing 300 mg each, were digested in a Teflon
X-ray fluorescence spectra were collected by an bomb with a mixture of nitric and hydrofluoric acids
Si(Li) detector of energy resolution equal to with the addition of hydrogen peroxide. The digested
170keV at 5.9keV and a Canberra SI00 multi- samples were spiked with selenium as internal
channel analyser. standard element. Then, three 5/~1 sub-samples were
taken from each peat solution and pippetted onto
3.2. Preparation of samples siliconised quartz reflectors, dried in a vacuum
desiccator and measured for 2000 s.
All of the reagents used in the process of sample The second group of samples to be analysed was
preparation were of the highest analytical grade prepared from biological standard reference materials,
(Fluka, Merck). Suprapure nitric acid was addi- IAEA 350 Tuna Fish Flesh and a Polish standard
tionally sub-boiled (Ktirner Analysentechnik) and reference material Cabbage Leaves CL-1. A mass of
water was distilled twice and deionised (EASYpure, 400 mg of the Tuna Fish Flesh standard was digested

Table 1
The mass fractions o f the elements in two peat core sections obtained b y T X R F using the empirical quantification method and the fundamental
parameter algorithm

Element Peat core layer sample 1 Peat core layer sample I1

Empirical quantification Fundamental parameter Empirical quantification F u n d a m e n t a l parameter


method method method method

[#g/g]

K 289 -+ 24 320 -+ 37 1754 + 165 1845 ± 174


Ca 1028 ± 34 1065 _+ 30 655 ± 30 675 ± 70
Ti 89.7 _+ 1.4 99.2 _+ 1.3 583 ± 18 644 _+ 17
V 1.8 + 0.3 3.1 ± 0.7 10.0 +_ 0.4 I 1.4 ± 0.2
Cr 3.3 _+ 0.3 4.4 _+ 0.5 9.8 ± 1.1 12.6 ± 1.8
Mn 4.8 ± 0.3 6.4 -+ 0.2 12.0 ± 0.3 13.3 ± 0.4
Fe 588 -+ 3 644 -+ 3 871 ± 33 958 ± 39
Ni 1.4 +- 0.2 1.8 -+ 0.2 6.6 ± 0.8 8.6 -+ 1.0
Cu 5.0 -+ 0.3 5.5 ± 0.4 10.2 ± 0.5 12.4 -+ 0.6
Zn 88.0 + 1.0 95.8 ± 0.9 85.6 ± 5.0 91.7 ± 5.4
As 1.7 + 0.3 1.6 ± 0.2 5.5 + 0.3 6.1 ± 0.2
Rb 1.3 ± 0.1 1.2 +_ 0.1 7.4 + 0.7 8.0 ± 0.8
Sr 7.7 + 0. I 8.5 ± 0.1 12.0 +_ 0.7 12.9 ± 0.8
Zr n.d. < LOD n.d. 24.2 + 2.0
Ba 8.3 ± 1.5 8.0 -+ 0.8 27.3 + 8 331. + 4.7
Pb 42.97 ± 1.51 38.7 ± 1.1 103 ± 9 92 + 7

n.d. - - not determined, because the relative sensitivity value for zirconium was not established
L O D - - limit o f detection.
46 D. Wegrzynek et al./Spectrochimica Acta Part B 53 (1998) 43-48

Table 2
The mass fractions of the elements in IAEA-350 Tuna Fish Flesh standard reference material obtained by TXRF using the empirical
quantification method and the fundamental parameter algorithm
Element 1AEA-350Tuna Fish Flesh
Empirical quantification method Fundamentalparameter method Certifiedconfidence interval
[/~g/g]
Ca 160 _+ 10 140 _+ 10 (72-191)
Fe 68.3 -+ 0.4 73.2 -+ 0.3 (66.7 77.3)
Ni 1.4 _+ 0.5 1.5 _+ 0.4 (0.32 1.13)
Cu 3.4 _+ 0.6 3.7 _+ 0.6 (2.55 3.10)
Zn 18.1 _+ 1.2 18.1 _+ 1.2 (16.6 18.5)
As 4.38 + 0.01 4.69 _+ 0.02 (2.50-5.71)
Se 4.6 + 0.2 4.6 _+ 0.2 (4.40 5.95)
Rb 2.3 + 0.1 2.3 + 0.1 (2.41 3.40)
Hg 2.9 _+ 0.1 3.5 + 0.2 (3.31 4.42)

in a Teflon bomb with nitric acid. The digested sample by the fundamental parameter algorithm. Small
was spiked with yttrium as internal standard. Three systematic differences can be observed in the case
3 #1 sub-samples were pipetted onto siliconised o f the results shown in Table 1 and Table 3, where
quartz reflectors. To prevent the losses o f mercury most o f the mass fractions obtained by the funda-
which was present in the Tuna Fish standard, mental parameter method are slightly higher than
2/zl volume o f an 0.1% ammonium-pyrrolidine- those obtained by the empirical quantification proce-
dithiocarbamate (APDC) water solution was addition- dure. However, both sets of results agree well taking
ally pipetted onto the quartz reflectors. The quartz into account the values o f uncertainties, except for
reflectors with samples were left to dry on a clean elements giving strongly interfering characteristic
bench under atmospheric pressure and then measured lines such as T i - B a and A s - P b . For the other results
for 2000 s. For analysis o f the Cabbage Leaves shown in Tables 2, and 4 the systematic differences
standard, a 230 mg sample was taken and digested do not appear, except for iron for which the mass
under the same conditions as before. After digestion fraction shown in Table 2 obtained by the FP algo-
the sample was spiked with selenium and volumes rithm is higher than that obtained by EQ. The results
o f 2/zl were pipetted on to three siliconised quartz of the analyses o f standard reference materials
reflectors, dried in a vacuum desiccator and measured presented in Table 2 and in Table 3 indicate that
for 2000 s. the mass fractions o f the elements obtained by the
The multi-element aqueous solution was prepared FP algorithm match the certified confidence intervals
from Merck single-element standard solutions o f slightly better than the mass fractions calculated with
barium, titanium, arsenic, lead and selenium. The use o f the EQ method. From the 17 determinations
1000/xg/ml standard solutions were diluted in such a produced by the FP algorithm 11 fit the certified
way as to obtain a solution of the following concen- confidence intervals, whereas for the results obtained
trations o f elements: Ti - - 10/~g/ml, Ba - - 2/zg/ml, by the EQ procedure eight fall within the range o f
Pb - - 10/~g/ml, As - - 2/zg/ml and Se - - 2 #g/ml. the certified confidence intervals. These small, not
very distinctive or decisive, systematic differences
between the results obtained by both methods may
4. Results and discussion have their source in the different algorithms used
for deconvolution of the spectra. In the case of the
The results of the analyses are presented in EQ method applied by us, the net intensities of
Tables 1-4. As can be seen, the concentrations/ K~ or L~ characteristic peaks were used to calculate
mass fractions determined by the empirical quantifi- the concentrations or mass fractions o f the elements,
cation procedure agree well with those determined whereas the FP method uses an advanced home-made
D. Wggrzynek et al./Spectrochimica Acta Part B 53 (1998) 43-48 47

Table 3
The mass fractions of the elements in CL-1 Cabbage Leaves standard reference material obtained by TXRF using the empirical quantification
method and the fundamental parameter algorithm

Element CL-1 Cabbage Leaves

Empirical quantification method Fundamental parameter method Certified confidence interval

[mg/g]
K 20.6 ÷ 0.4 21.0 _+ 0.5 (24.4-28.0)
Ca 6.0 +_ 0.1 6.0 + 0.1 (5.3 7.2)
[~g/g]
Mn 51.4 _+ 0.8 59.0 _+ 0.9 (55.5-59.7)
Ni 1.9 -+ 0.4 2.1 _+ 0.5 (2.7-3.3)
Cu 3.4 _+ 0.3 3.9 _+ 0.4 (3.1 3.6)
Zn 35.8 -+ 1.2 38.5 +- 1.3 (36.6-40.4)
Rb 37.2 -+ 0.4 40.9 -+ 0.5 (38.0-44.8)
Sr 21.8 + 0.4 24.0 +- 0.4 (14.9-26.3)

program taking into account all K-series or L-series sections shown in Table 1. These systematic differ-
X-ray peaks to calculate the concentrations/mass ences in titanium content could originate either from
fractions of the elements. However, it should be a shift between the empirical sensitivity value used
noted that in other implementations of the EQ algo- in the EQ procedure and that calculated in the FP
rithm, bundled with commercial TXRF spectrometers algorithm, or from the fact that different spectra-
such as the EXTRA II (Rich. Seifert and Co., fitting algorithms were used to obtain net peak areas
Ahrensburg, Germany), all the K-series or L-series in both methods. As can be seen from Table 3, both
X-ray peaks are used in the course of the analytical algorithms do not give good values for the mass
procedure. fraction of potassium, the lightest element analysed.
As can be seen from Table 4 in the case of the This is probably caused by absorption of the charac-
analysis of an aqueous solution containing two pairs teristic radiation of potassium in the analysed sample,
of elements giving strongly interfering X-ray peaks which was not corrected by either of the methods
(Ti-Ba and As-Pb), the concentrations of the ele- studied.
ments obtained by both methods do not differ signifi-
cantly. The exception is titanium, for which the result
obtained by the EQ method is about 10% lower than 5. Conclusions
the true value and the value obtained by the FP algo-
rithm. A similar tendency can be observed for the As can be seen from the results presented, both
analysis results of real samples such as the peat core methods of quantitative analysis give concentrations

Table 4
The concentrations of the elements in an aqueous multi-element solution containing known amounts ofTi, Ba, As, Pb and Se obtained by TXRF
using the empirical quantification method and the fundamental parameter algorithm

Element Multi-element solution

Empirical quantification method Fundamental parameter method Known concentration

[mg/1]

Ti 9.1 _+ 0.1 10.2 -+ 0.1 10.0


Ba 2.0 +_ 0.2 2.0 -+ 0.1 2.0
As 1.89 + 0.01 1.87 + 0.01 2.0
Pb 10.2 _+ 0.2 9.6 _+ 0.2 10.0
Se internal standard 2.0
48 D. Wggrzynek et al./Spectrochimica Acta Part B 53 (1998) 43-48

or mass fractions o f the elements that agree well References


within the values o f uncertainties. In the case of
samples with a simple matrix the results are almost [1] G. Trig, R. Klockenk~imper, Spectrochim. Acta 48B (1993)
identical (aqueous solution), whereas for samples 111.
[2] D. WCgrzynek,B. Hot)'fiska, Spectrochim. Acta 52B (1997)
with complex matrix or samples containing higher 915.
abundances o f low-Z elements the FP algorithm [3] R. Klockenk/imper, Total-reflection X-ray fluorescence
seems to perform slightly better than the empirical analysis, in: J.D. Winefordner (Ed.), Chemical Analysis - -
quantification procedure. On the other hand, the A Series of Monographs on Analytical Chemistry and Its
differences in the performance o f these quantifica- Applications, vol. 40, John Wiley and Sons, Inc., New York,
1997, p. 154.
tion procedures are not significant compared with
[4] IAEA Manual: QXAS Quantitative X-Ray Analysis System,
the complexity o f the FP approach. The FP algorithm IAEA, Vienna, 1996.
is more complex than the empirical quantification [5] P. Van Espen, H. Nullens, F. Adams,Nucl. Instrum. Meth. 142
procedure, but when developed it can easily be (1977) 163.
applied. The advantage o f the FP algorithm is that it [6] M.O. Krause, J.H. Olivier, J. Phys. Chem. Ref. Data 8 (1979)
329.
produces results o f the same accuracy and quality as [7] K. Debertin, W. Pessara, Nucl. Instrum. Meth. 184 (1981)
the EQ method but without the need for any calibra- 497.
tion procedure. The FP algorithm does not even need [8] G.W. Philips, K.W. Marlowe,Nucl. lnstrum. Meth. 137 (1976)
sensitivity values which have to be determined for EQ 525.
methods. Also, the advantage o f the FP method is that [9] J.L. Campbell, B.M. Millman, J.A. Maxwell, A. Perujo, W.J.
Teesdale, Nucl. Instrum. Meth. B9 (1985) 71.
the FP algorithm utilises the intensities o f all K-series [10] E. Marageter, W. Wegscheider, K. Miiller, X-Ray Spectrom.
or L-series characteristic peaks to calculate the con- 13 (1984) 78.
centrations/mass fractions of elements, whereas our [11] W.H. McMaster, N.K. Delgrand, J.H. Hubbel, Compilationof
EQ method takes into account only the net intensities X-rays cross sections, UCRL-50174, Lawrence Radiation
of K~ or L~ peaks. This makes the FP algorithm more Laboratory, CA, 1968.
[12] M.O. Krause, C.W. Nestor Jr, C.J. Sparks Jr, E. Ricci, X-ray
suitable for the analysis of samples with a complex
fluorescencecross sections for K and L X-raysof the elements,
matrix producing X-ray spectra with strongly interfer- ORNL-5399,Oak RidgeNational Laboratory,Oak Ridge, TN,
ing characteristic peaks. 1978.
[13] J. Scofield,, Phys. Rev. 179 (1969) 9.
[14] E. Storm, H.I. Israel, Nuclear Data Tables A7 (1970) 561.
[15] B. Hotyfiska, B. Ostachowicz,J. Ostachowicz,A. Ostrowski,
Acknowledgements J. Ptasifiski, D. WCgrzynek,J. Trace Microprobe Techniques
13 (1995) 163.
This work was partially supported by the State [16] P. Wobrauschek, P. Kregsamer, Spectrochim. Acta 44B
Committee for Research, grant no. 11.220.01. (1989) 453.

You might also like