Fischer Application report vr118 HELM UT FISCHER GMBH + CO. KG Institut für Elektronik und Messtechnik Industriestrasse 21-7169 Sindelfingen, Germany Tel.: (+49) 731 33- - Fax: (+49) 731 33-79 E-Mail: mail@helmut-fischer.de Internet: w w w.helmut-fischer.com XUV 773: X-Ray Fluorescence Analysis of Gemstones The energy dispersive X-ray fluorescence analysis (ED-XRFA) is a well-established tool for multielement composition analysis in different fields/1/. It works non-conducting and it is non-destructive. No particular sample preparation is necessary if the specimens have an almost planar and smooth surface. For the characterization of precious gemstone material it is also important that the deposited X-ray radiation dose is definitely too weak for inducing colour alterations, in contrast to WD-XRFA /2/. This is essential for gemstone testing. The concentrations of the major and minor and the trace elements give important hints for the type and/or the authenticity of the gemstone. In this way ED-XRFA becomes essential for the identification of the mineral in addition to structural analysis (Raman, XRD) and classic optical methods. 1. XRFA in vacuum The ED-XRFA of gemstones and mineral samples has to cover both the analysis of light and heavy elements. The evaluation of the light matrix elements Na, Mg, Al enforces the evacuation of the measuring head is necessary for the detection of light elements (Z < 14). Also for the Z-range 15 2 (P Ca) the detection limits are improved. The vacuum instead of normal atmosphere eliminates the air path absorption between detector and sample (cf. tab. 1). May be that for Si-K and Al-K the absorption can be counterbalanced by longer measuring times. The measurement of the fluorescence of Mg-K and in particular Na-K requires vacuum absolutely. A combination of different X-ray excitation conditions is applied in order to reach optimal measuring sensitivities for a wide range of atomic numbers Z. The WinFTM /3/ software includes the evaluation of the respective spectra in a single run which does not require particular efforts of the user. This so-called multiple excitation mode offers also possibilities to suppress the influence of X-ray diffraction. This quantification consequently utilizes a complete computer simulation of the measured spectra which is based on a fundamental parameter treatment described in /4/. This report describes the application of an XRF vacuum measuring system for the analysis of several gemstone reference samples in order to qualify the methodology. 1
Tab. 1 Transmission (%) of X-Rays for different detector to sample air path lengths. 1.5 cm is the minimum distance for the standard geometry of Fischer XRF systems with several apertures and video optics. air path Na Kα 1.4keV Mg Kα 1.25keV Al Kα 1.49keV Si Kα 1.74keV P Kα 2.2keV S Kα 2.31keV 1.5cm.26 2.8 1.9 25 4 57 1.cm 1.9 9.3 23 39 54 69.5cm 13.8 3 48 63 74 83.2cm 45 62 74 83 88 93 Fig. 1 A sample carrier with 64 gemstone specimen is placed on the motorized stage of the Fischerscope XUV 773 (top). p symbolizes the primary X-ray beam which excites the sample for the emission of fluorescence radiation (f). The sample has to be fixed over a radiation trap. It avoids that the backscattered radiation (s) can be detected. The fig. 1 shows a sample fixing support for 64 gemstone samples. They are placed over a hole to prevent scattering background radiation from the backing. Also the measuring geometry is shown schematically. 2
2. Multi-excitation and Braggs Fig. 2 XRF spectra of a mineral sample (andesine) for different excitations: a) 5 kv, 1 mm Al primary filter (green) b) 2 kv,,1 mm Al filter (blue) c) 8 kv,,1 mm Mylar filter (ochre). Fig. 3 The effect of different primary absorption filters to a continuous primary radiation. The thin filter has maximum intensity. The energy range without potential elastic scattering inclusive Bragg is small (origin - 1). The filter reduces the low energy part mainly. A thick filter causes a large region without elastic scattering (origin 3) but it is counterbalanced by s low primary intensity. Optimal detection limits for light elements are achieved by a soft excitation or a low high voltage. The optimal HV (X-ray tube high voltage) should be about 3 times the ionization threshold of the respective inner shell, as a guideline. So 5 kv should be optimal for Al (K-shell), and 15 kv for Ti, and 3 kv for Ga, etc. 3
Different primary excitations have to be applied in order to cover the full element range of interest /5/. The component mode has been used because it is known that the mineral consists of oxides. The list of defined and analysed components is given in tab. 1 below. It is important to note that the software itself decides which excitation mode is used for the determination of any element. These decisions depend on the measuring sensitivities of a given element for each excitation mode, which is calculated in any case. This internal weighing is influenced by possible diffraction peaks. The samples have a crystalline structure which is the origin of diffraction peaks in the spectrum. These so-called Bragg-peaks appear in the spectrum by random due to the current position of the reflecting lattice plane which is normally unknown. They can disturb the evaluation considerably and are a source of non-reproducibility. They can be avoided or suppressed by using filtered primary radiation (fig. 3). The optimum excitation condition for a certain element without potential Bragg interferences predicts a certain high voltage filter combination. A compromise has to be found in order to analyse several elements with a large Z-range (Na to U). It has to take into account - the application of only a few filters (thickness & material) - in combination with only a few high voltage settings. WinFTM restricts the number of different excitations (high voltage and filter settings) to maximum three. This so-called multiple excitation mode automatically takes the optimum information from all spectra recorded with different excitation conditions (fig. 2). Their definition has to consider the mentioned criteria (avoid Braggs and optimal sensitivities for the elements of interest). The fig. 4 illustrates the compromise situation with respect to Na analysis. The softest excitation (no filtering) yields the best Na-K excitation (yellow spectrum). Unfortunately, there are some strong Bragg peaks labelled with B. The Na-K fluorescence peak is overlapped by an intense diffraction peak. Hence, Na cannot be determined from this spectrum. The selection of high voltage filter combinations for the single excitations has to prevent such cases. The blue spectrum of fig. 3 belongs to the same sample. The soft primary radiation is absorbed (filtered) by a thin plastic foil. Scattered primary radiation which could overlap by diffraction in the 1 kev-region is avoided. However, this filtering reduces the intensities of the low energy fluorescence peaks. The unfiltered soft primary radiation can be applied in case of a non-crystalline sample (e.g. glass). Fig. 4 The effect of a thin mylar primary filter to the low energy part of the spectrum. The yellow spectrum of the unfiltered primary radiation (tube operation voltage = 8 kv) causes some Bragg reflection labelled with a B. The Na-K peak is completely overlapped by a diffraction peak and Na cannot be analysed from this measurement. A 1 µm thick plastic (mylar) filter reduces the probability of such interferences (blue spectrum). 4
3. Quantification and sample geometry The WinFTM FP treatment calculates a best-fit of the theoretical spectra with the measured ones. The free parameters of this iterative fit calculation are the unknown concentrations. The mathematical best fit is the theoretical or standardless result. The treatment assumes a sample which consists of planparallel layers. In case of a mineral sample we have only one layer. And the interaction length within the sample is small compared with the detector-sample-distance d int eraction << d det sample, cf. fig. 5. For several mineral and gemstone samples both preconditions do not hold true in any case (in particular for high fluorescence energies). WinFTM calculates the absorption of a fluorescence line for a plane surface. A shaped one may influence the result. Fig. 5 Geometry of gemstone analysis The internal algorithm of version 6.23 has been modified with respect to get more tolerance with respect to (small) changes of the intensity of a spectrum. This will not correct the result for such undefined variations but it can avoid a complete failure of the iteration. Optimal DefMA setup setup - Exitation with 3 different modes (5 kv & 1mm Al filter, 2 kv and.1 mm Al filter, 8 kv and.1 mm mylar filter). - The concentrations of oxides have to be analysed: the component modus is used. - The balance component (or balance element) is the last one defined. No trace element at the last position. - Start parameters for the single concentrations are necessary for the component mode and helpful for the element mode. - The soft 8 kv excitation is defined with ratio method. - A ROI for the lower limit of the spectrum is helpful, e.g. ROI1=25. 5
4. Test results This section reports some comparisons of standardless XRF results with data from other analytical techniques. Comparing the respective results we have to take into account both the statistical uncertainties (cf. tab. 2) and systematic differences. These may have different reasons such as Local variations of element concentrations (examples see figs.6-8), Different information depth, Systematic calibration errors (of the other methods). Tab. 2 Measuring results of a mineral (andesine). Measuring time is 1 s per excitation mode. Excitation conditions acc. to fig. 2 Component Mean (wt.-%) Std.dev. (wt.-%) Estimated detection limit (wt.-%) Na 2 O 7.5.9 2 MgO.8.2 1 Al 2 O 3 24.6.9 1 SiO 2 57..9.5 K 2 O.5.2.1 CaO 9.1.5.1 TiO 2.4.5.2 V 2 O 5 -.2.1 MnO -.1.3 Fe 2 O 3,32.16.3 Co 3 O 4 -.1.3 Ni 2 O 3 -.6.2 CuO.57.3.2 ZnO -.3.1 Ga 2 O 3 -.4.1 SrO.14.9.1 PbO -.2.6 Bi 2 O 3 -.2.6 Video image of an andesine. The points drawn in mark the line scan shown in the follow- Fig.6 ing figs 7-8. 6
Fig. 7 Linescan of the Na 2 O concentration for an andesine sample (fig. 6). The scattering of the measuring points reflects the statistical uncertainty. Fig. 8 Linescan of the Al 2 O 3 concentration for an andesine sample (fig. 6). The scattering of the measuring points reflects the statistical uncertainty. Normally, the mineral samples are more homogeneous then the andesine sample of figs. 6-8. Many gem specimens are used as reference standards /6/ which are analyzed by different techniques. Also some mineral certified reference material is available. The figs. 9-12 depict a comparison of XRF measurements with data from other methods. The correlation is sufficient or good. Remaining deviations are expected to be caused by different reasons which cannot be discussed in detail here. 7
Mg y =,998x +,7296 % M g O ( o t h e r m e th o d s ) 1 5 2 4 6 8 1 12 14 %MgO (XRF) Fig. 9 Comparison with MgO analysis in different minerals and gemstones. The single points refer to SEM measurements of some selected gemstones /6/ and to certified standard materials % Al2O 3 (other methods) Al y =,953x -,7114 1 8 6 4 2 2 4 6 8 1 %Al2O3 (XRF) Fig. 1 Comparison with Al 2 O 3 analysis in different minerals and gemstones. The single points refer to SEM measurements of some selected gemstones /6/ and to certified standard materials 8
Si 8 y =,8753x + 2,5867 7 %SiO2 (other methods) 6 5 4 3 2 1, 2, 4, 6, 8, 1, %SiO2 (XRF) Fig. 11 Comparison with SiO 2 analysis in different minerals and gemstones. The single points refer to SEM measurements of some selected gemstones /6/ and to certified standard materials %CaO (other methods) Ca y = 1,591x -,1376 15 1 5, 5, 1, 15, %CaO (XRF) Fig. 12 Comparison with CaO analysis in different minerals and gemstones. The single points refer to SEM measurements of some selected gemstones /6/ and to certified standard materials 9
%TiO2,1,9,8,7,6 XRF,5,4,3,2,1,1,2,3,4,5,6,7,8,9,1 LA a) %V2O5 X R F,8,7,6,5,4,3,2,1,1,2,3,4,5,6,7,8 LA b) 1
%Cr2O3 1,4 1,2 1,8 XRF,6,4,2,2,4,6,8 1 1,2 1,4 LA c) MnO,1,9,8,7,6 XRF,5,4,3,2,1,1,2,3,4,5,6,7,8,9,1 LA d) 11
%Fe2O3 3 2,5 2 XR F 1,5 1,5,5 1 1,5 2 2,5 3 LA e) %Ga2O3,1,9,8,7,6,5,4,3,2,1,1,2,3,4,5,6,7,8,9,1 LA f) Fig. 13a-f Comparison of LA-ICP-MS with standard free XRFA results 12
The trace element and minor element concentrations of a number of reference standards of the GRS Lab /6/ have been determined by means of LA-ICP-MS which is a well accepted method in gemstone characterization. The comparisons (figs. 13a-f) prove that XRFA reflects the trace element concentrations quite reliable. In contrast to the laser beam technique 1 it works non-destructive and so XRFA has a big potential in this field. References /1/ T. Kouichi, J. Injuk, and R. van Grieken (eds.), X-ray spectrometry: recent technological advances, Wiley 24. /2/ A. Burkhardt, priv. communication. /3/ Helmut Fischer GmbH, WinFTM softwareversion 6.22 ff. /4/ V. Rößiger and B. Nensel, in Handbook of practical X-Ray fluorescence analysis, Springer 26, p. 554. /5/ Fischer application note vr99, Multi excitation. /6/ A. Peretti, priv. communication 1 The sophisticated equipment is very expensive also. 13