JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY MARCH 1994 VOL. 9 311 Particle-induced X-ray Emission Thick-target Analysis of Inorganic Materials in the Determination of Light Elements* J. Perez-Arantegui Departamento de Quimica Analitica Facultad de Ciencias Universidad de Zaragoza 50009 Zaragoza Spain G. Querre Laboratoire de recherche des musees de France Palais du Louvre 75007 Paris France J. R. Castillo Departamento de Quimica Analitica Facultad de Ciencias Universidad de Zaragoza 50009 Zaragoza Spain Particle-induced X-ray emission (PIXE) has been applied to the analysis of inorganic materials to determine some elements with Z<27 Na Mg Al Si K Ca Ti Mn and Fe in thick-target analysis. A PIXE method has been developed for the analysis of geological materials ceramics and pottery.Work has been carried out with an ion beam analytical system using a low particle beam energy. Relative sensitivity detection limits reproducibility and accuracy of the method were calculated based on the analysis of geological standard materials (river sediments argillaceous limestone basalt diorite and granite). Analysis using PIXE offers a number of advantages such as short analysis time multi-elemental and non-destructive determinations and the results are similar to those obtained with other instrumental techniques of analysis. Keywords X-ray specfromefry; particle-induced X-ray emission; thick-target analysis; geological and archaeological materials Although it was first introduced in the 1970s particle-induced X-ray emission (PIXE)' has recently reached a high level of importance as is shown by the number of papers and reviews that have been published in the last few Interesting comparisons with other analytical methods can be found in these publications.One of the main features of PIXE is that it is an X-ray method where excitation is performed by charged heavy particles usually protons. As well as the already established advantages of PIXE for the determination of trace elements' (high sensitivity multi- element method versatility and simple sample preparation good spatial resolution rapid and non-destructive analysis) this method can also give excellent results when quantifying major and minor elements. This is the case even for materials whose matrices consist mainly of light elements (the K X-ray peaks for the light elements are just superimposed on the Bremsstrahlung background produced in the sample and on the background produced by secondary electrons) and for the analysis of thick targets (thicker than the energetic range of the particles). When working with geological samples ceramics or pottery it is precisely the light elements (with 2<27) such as Na Mg Al Si K Ca Ti Mn and Fe which are the major and minor components and therefore the ones that have to be considered when characterizing the matrix of such materials.The principal objective of the present work was to establish a method of analysis that enables the determination of the chemical composition for the major elements (Na Mg Si Al K Ca and Fe) and some minor ones (Ti and Mn) in inorganic materials of the following type geologcal ceramic and archae- ological.Hence geology new materials archaeology and the are important areas of application for PIXE in order to characterize such materials. Experimental Apparatus These studies were carried out with an analytical system of ion beams AGLAE a 6SDH-2 2 MV tandem electrostatic * Presented at the XXVIII Colloquium Spectroscopicurn Inter- nationale (CSI) York UK June 29-July 4 1993. NEC Pelletron accelerator in the Laboratoire de recherche des musees de France (Louvre Paris France).'6917 The ion source was a classical Alphatross r.f. source (National Electrostatics Company). The beam of particles provided by a 45" line reached the sample inside an irradiation chamber under vacuum [l x Torr (1 Torr= 133.322 Pa)] and the detection of X-rays was carried out by an Ortec Si(Li) detector with an Ortec 7800 multichannel analyser.A VME (Versa Module Eurocard) processor system was used for the required automatic operation. Instrumental Parameters and Procedure As the first element to be determined was Na ( Z = l l ) for the determination of the other elements (Na Mg Al Si K Ca Ti Mn and Fe) no filter was placed in front of the detector. Since the bombarded material was an insulator a filament of tungsten was used as an electron gun in order to prevent a background owing to self-charging of the sample. Because of the heterogeneity of the material analysed and of the reduced dimensions of the beam (approximately 1 mm') a sample preparation stage was used in order to give more reliable results.However one of the advantages of PIXE is that it requires only very simple sample preparation. By taking about 100mg of material in a powdered form small pellets were prepared of the standards and samples mixed with an organic binder [Moviol a poly(viny1 alcohol)] and compressed in a steel die under 3 tonne of pressure. These pellets 8 mm diameter and 1 mm thickness were placed in Plexiglass sample holders 4.5 x 4.5 cm (Fig. 1). The target was bombarded with a beam of low energy protons 0.7 MeV with a current of 3 nA and with an integrated charge of 0.4 pC. The analysis of each sample lasted approxi- mately 150 s and the spectrum was registered between 0.8 and 10 keV. The detection angle is 150" and the distance to the detector is 40 mm.Both fitting of the spectra to reference spectra and the calculation of integration of the yield of X-rays over the whole particle range were made using a computer programmed with PIXGRAF and THICK (PIXAN package).I8 The first program takes into account the calibration the parameters of the detector and a background model of iterative peak filing. The312 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY MARCH 1994 VOL. 9 l o o 0 1 I Fig. 1 Plexiglass sample holders; dimension of the holder 0.7 cm second program considers the geometric parameters the characteristics of the incident beam and the composition of the sample under analysis. Analysis of Geological Standards A study of different parameters of analytical interest was made by analysing various geological standards.Five standards were chosen 82GOV1 DR-N (diorite) 84GOV2 GS-N (granite) and 80GOV1 BE-N (basalt) three geostandards from the Centre de Recherches Pktrographiques et Gkochimiques; and two Standard Reference Materials (SRMs) 2704 Buffalo River Sediment (RS) and l c Argillaceous Limestone (AL) from the National Institute of Standards and Technology (NIST). The elements studied were Na Mg Al Si K Ca Ti Mn and Fe. As an example the spectrum given by the standard DR-N can be seen in the Fig. 2. Results and Discussion The relative sensitivity of each element was first calculated in photons ng-’ cm2 pC-’ as shown in the Table 1 together with the minimum detection limit of the signal measured in photons. However it is the calculation of the detection limits which is of greater analytical interest expressed as the concentration of an element in the samples studied.This gives a clearer idea of the sensitivity of the method. If it is considered that if the peak area in the spectrum corresponding to element i is related to its concentration by eqn. (1) and if the instrumental param- eters are maintained constant then there exists a linear relation - 104 u 103 5 102 C C L a) P fn 4- 0 Y- g 10 1 0 2 4 6 8 Energy of X-ray/keV Fig. 2 Spectrum of the DR-N standard (diorite) between the area of the peak and the product of the concen- tration and the integral for element i [eqn. (2)] where Si=area of peak i; Ci=concentration of element i; Q/e = number of incident protons; R = detection solid angle; E = detector efficacity; N = Avogadro’s number; A = relative atomic mass of element i; o= yield of X-ray fluorescence; E= particle energy in the sample; E,=particle energy on reaching the target; o(E) = cross-section probability of ionization; T(E) = transmission of photons from successive depths in the matrix; S ( E ) = stopping power; a = intercept of the calibration line; and I =integral term of eqn. ( 1).si = a + b(CJ) (2) The model proposed by Clayton et a1.I9 was then applied in order to calculate the detection limits from a linear cali- bration with unknown parameters (3) where xd(P,q) =detection limit; oo = parameter fixed by the calibration design for r (number of determinations made from a ‘similar’ source matrix) and n (number of observations); A(P,q) = tabulated values for specified p q and v (degrees of freedom n - 2); p =false positive rate; q = false negative rate; c2 =variance; and b =slope of the line.This approach is used to define the detection limits so that protection against both false positives (reporting an analyte as present when it is not) and false negatives (reporting an analyte as not present when it is) is assured. The method is based upon classical (Neyman-Pearson) statistical theory for hypoth- esis testing it assumes that standard calibration procedures are used to provide estimates of all unknown parameters. The detection limits calculated in this manner for this type of material are 0.91 0.40 1.07 2.18 0.25 0.60 0.11 0.109 and 0.91 (as YO of the element) for Na Mg Al Si K Ca Ti Mn and Fe respectively for r = 1 (number of determinations made from a ‘similar’ source matrix) n = 12 (number of observations) p = 0.1 and q = 0.1 with ten degrees of freedom (n - 2).Although these detection limits would appear high it has to be recognized that they are calculated for working conditions and for a matrix for which it is precisely the major elements that are to be studied and which are the centre of interest for the determination so this is no real disadvantage. In order to check the precision and reliability of the method various determinations were carried out for each of the stan- dards. The results obtained for the diorite (DR-N) for seven determinations can be seen in Table 2. The value proposed for the concentration of each element and the average value obtained together with the standard deviation and the relative standard deviation for each of the elements are all given.The results obtained for the four other standards with respect to the proposed values are also given in Table 2. The chemical compositions are calculated from the linear cali- bration made by determining the product of the concentration and the integral and by taking either a hypothetical integral as would be done in the case of samples of an unknown composition but with a similar matrix (column I) or the known integral for this standard (column 11). From this variation it can be concluded that it is important to analyse various standards together with the unknown samples but Table 1 Parameter Na Mg A1 Si K Ca Ti Mn Fe Relative ~ensitivity/lO-~ 1.2-1.6 0.7-0.8 0.20 0.18 MDL/photons Relative sensitivity and minimum detection limit (MDL) of the s.igna1 photons ng-’ cm2 pC-‘ 0.4-0.5 1.8-2.1 5.6-6.0 7.8-8.3 2.4-2.6 100-150 180-225 300-350 300-375 100-130 100-140 25-40 9-20 8-15JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY MARCH 1994 VOL.9 313 Table 2 Results (as % of the element) for the analysis of several standards (n=4) DR-N* GS-N BE-N RS AL Element Na Mg Al Si K Ca Ti Mn Fe Cert.7 AV§ sm RSDll Cert. AV SD RSD Cert. AV SD RSD Cert. AV SD RSD Cert. AV SD RSD Cert. AV SD RSD Cert. AV SD RSD Cert. AV SD RSD Cert. AV SD RSD I 2.22 2.40 0.109 4.56 2.65 2.71 0.084 3.10 9.27 9.33 0.362 3.88 24.70 24.62 0.8 13 3.30 1.41 1.30 0.085 6.52 5.04 4.69 0.149 3.18 0.654 0.65 0.030 4.69 0.172 0.179 0.017 9.71 6.78 6.84 0.321 4.69 I 2.80 3.28 0.142 4.33 1.39 1.53 0.156 10.20 7.76 8.22 0.3 18 3.87 30.75 32.04 0.766 2.39 3.84 3.69 0.112 3.03 1.79 1.81 0.058 3.19 0.408 0.43 0.032 7.31 0.0431 - - - 2.64 2.50 0.247 9.88 I 2.36 3.26 0.198 6.06 7.93 6.93 0.103 1.49 5.33 4.81 0.22 1 4.59 17.85 16.69 0.176 1.05 1.15 1.20 0.014 1.19 9.91 10.12 0.147 1.45 1.565 1.61 0.03 1 1.92 0.155 0.138 0.01 3 9.49 8.98 9.05 0.200 2.21 I 0.55f 0.59 0.030 5.06 1.20 1.34 0.106 7.86 6.11 7.26 0.329 4.53 29.08 28.45 0.378 1.33 2.00 2.12 0.023 1.10 2.60 2.82 0.239 8.48 0.457 0.49 0.025 5.02 0.0551 - - - 4.1 1 4.54 0.310 6.83 I 0.015$ - - - 0.25$ 0.25 0.010 4.05 0.69$ 0.25 0.016 6.28 3.20 2.79 0.096 3.44 0.23$ 0.22 0.017 7.76 35.90 37.52 0.03 1 0.08 0.0401 - - - 0.019$ - - - 0.381 0.19 0.026 13.59 * n = 7 .t Cert. = certified value.$Below the detection limit. 9 AV =average value. ISD =standard deviation. )I RSD = relative standard deviation. these should have as similar a chemical composition as is possible. In Table 2 it can also be observed that sometimes the practical detection limits for some elements (Na Mg A1 and K) seem to be better than predicted. It should be remembered that the concentrations at the detection limit have been calcu- lated at a confidence level of 90% protected against both false positives and false negatives. Hence sometimes lower concen- trations of these elements could be measured but with a poorer accuracy for the low-energy peaks just where the sensitivity and reproducibility are also lower. Conclusions Particle-induced X-ray emission as an analytical method for the determination of major and minor elements (with 2<27) offers a series of advantages simple preparation of the sample; speed the number of samples analysed daily can be very high; multi-elemental character by determining all of the elements between Na and Fe at the same time; a non-destructive method of analysis samples are taken which can be recovered after the analysis; its sensitivity and detection limits under the concentration ranges which are going to be determined; and its reliability accuracy and reproducibility with coefficients of variation similar to those obtained from other instrumental methods of analysis.Therefore PIXE can be successfully applied to the charac- terization of geological materials and indeed other similar inorganic materials such as pottery ancient ceramics or glass.We appreciate the collaboration with the Laboratoire de recherche des musCes de France and especially the cooperation of T. Calligaro J. Salomon F. Saltron and A. Bouquillon. 1 2 3 4 5 6 7 8 9 10 References Johansson S. A. E. and Campbell J. L. PIXE A Novel Technique for Elemental Analysis Wiley Chichester 1988. Proceedings of the International Workshop in Ion Beam Analysis in the Arts and Archaeology Nucl. Instrum. Methods Phys. Res. Sect. B 1986 14 1. Bird J. R. Duerden P. and Wilson D. J. Nucl. Sci. Appl. Sect B 1983 1 357. Houdayer A. Lessard L. and Brissaud I. Nucl. Instrum. Methods Phys. Res. Sect. B 1984 3 412. Maenhaut W. Nucl. Instrum. Methods Phys. Res. Sect. B 1990 49 518. Demortier G. Nucl. Instrum. Methods Phys.Res. Sect. B 1991 54 334. Johansson S. A. E. Analyst 1992 117 259. David D. Surf. Sci. Rep. 1992 16 333. Kristiansson K. and Malmqvist L. Geoexploration 1987 24 517. Burnett D. S. Woolum D. S. Benjamin T. M. Rogers P. S. Z.,314 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY MARCH 1994 VOL. 9 Duffy C. J. and Maggiore C. Nucl. Instrum. Methods Phys. Res. Sect. B 1988 35 67. 11 Katsanos A. A. Nucl. Instrum. Methods Phys. Res. Sect. B 1986 14 82. 12 Malmqvist K. G. Nucl. lnstrum. Methods Phys. Res. Sect. B 1986 14 86. 13 Peisach M. Nucl. Instrum. Methods Phys. Res. Sect. B 1986 14 99. 14 Bird J . R. Nucl. Instrum. Methods Phys. Res. Sect. B 1986 14 156. 15 Rye 0. S. and Duerden P. Archaeometry 1982 24 59. 16 Amsel G. Menu M. Moulin J. and Salomon J. Nucl. Instrum. Methods Phys. Res. Sect. B 1990 45 296. 17 Menu M. Calligaro T. Salomon J. Amsel G. and Moulin J. Nucl. Instrum. Methods Phys. Res. Sect. B 1990 45 610. 18 Clayton E. Duerden P. and Cohen D. D. Nucl. Instrum. Methods Phys. Res. Sect. B 1987 22 64. 19 Clayton C. A. Hines J. W. and Elkins P. D. Anal. Chem. 1987 59 2506. Paper 3f03948J Received July 7 1993 Accepted October 12 1993