Quantitative Determination of Platinum by XRF Techniques
Transcription
Quantitative Determination of Platinum by XRF Techniques
Platinum Manufacturing Process Platinum Day Symposium Volume IX N5 Quantitative Determination of Platinum by XRF Techniques Greg Normandeau & David Ueno Imperial Smelting & Refining Co. of Canada Ltd. • • • The quantitative determination of platinum by x-ray fluorescence (XRF) methods involves analysis of a solid sample to determine content in commercial alloys at a repeatable absolute accuracy level of 0.10%. This level is considered quantitative while 1-2% relative accuracy is considered qualitative. The goal is to supplement classical wet chemical methods (A.A & I.C.P) with XRF techniques as an accepted method for demonstrating compliance to international precious metals marking requirements. Coverage of the topic includes the following sections: • A general definition of XRF concepts. • Explanation of our analytical methods related to x-rays. (secondary target, filters, high total counts, long count times, triplicate analysis). • Explanation of the principles of our analytical methods re- • • lated to regression models, correlation coefficients and the number of standards required. Discussion of current alloy applications and future potential. An outline of our general techniques for standard preparation, assay via A.A & I.C.P. and polished surface preparation. Explanation, with graphics, of the routine for running a sample, printing results, storing data, examining and confirming quality of results Explain results of commercial applications in 1999, 2000, 2001. We’ll emphasize the number of tests, show range, standard deviation and accuracy with statistical terms. Conclusions and Further Research. DEFINITION OF TERMINOLOGY: Consideration of this topic must be based on understanding some basic terms and technical concepts of the trade. The following are considered significant: X-Ray fluorescence spectroscopy: (XRF) The detection and analysis of x-rays emitted from a sample after impingement by a controlled x-ray energy source. X-Ray: unique and characteristic energy emitted from a solid sample when it is irradiated with a high energy electron or x-ray beam. Accuracy: the uncertainty limits on an analytical result for a specific platinum content. e.g.: 95.20% Platinum ±0.10% Precision: the repeatability of an analytical result through the application of a technique. These terms will be referred to throughout the document. Figure 1 outlines the concept of x-ray generation. It is beyond the scope of the article to present the atomic physics that underlie the photoelectric effect phenomenon. Entire texts have been written on the subject. Energy from an x-ray source hits the electrons in an atom’s orbital and displaces a particular electron. A nearby electron in a neighboring orbital shell Figure 1 © Platinum Guild International USA 2002 All Rights Reserved 1 Platinum Manufacturing Process Platinum Day Symposium Volume IX N5 will replace the missing electron to preserve energy stability. The change to a lower energy level when filling the void created by the missing electron causes emission of a specific characteristic xray for that element. Electrons coming from different shells create the Kα, Kβ, Lα and Lβ, type x-rays that all have distinctive energy levels unique for each element. Figure 2 shows a typical xray spectrum with labeled energy peaks at various points along the spectrum. Note that the signals are very tall, distinct energy peaks while the general background noise is minimal. This high signal to noise ratio is desirable for accurate analysis. Figure 2 Figure 3 illustrates the general schematic of a typical x-ray fluorescence unit. Figure 4 shows the general size and shape of a commercial XRF unit. This particular type deploys a secondary target and filter system. The primary x-rays from a Rh tube source impinge on a tin secondary target. That target emits x-rays that are filtered before hitting the sample. This technique produces clean monochromatic x-rays that enhance accuracy. The x-rays emitted by the sample are captured by a liquid nitrogen cooled detector that translates them into electrical pulses. The specimen emits an x-ray spectrum characteristic of its atomic constituents from an area of approximately 8mm round. The depth of about x-ray penetration is dependent on the matrix and power used. This point is critical when considering the validity of results. Surface treatments and some history of the sample must be known. Figure 3 2 © Platinum Guild International USA 2002 All Rights Reserved Platinum Manufacturing Process Platinum Day Symposium Volume IX N5 Solid-state pulse processing equipment and a PC manipulate the emitted counts through statistical means to compute the quantity of metal present in the sample. Specific computational routines must be developed to establish a high degree of accuracy for a specific set of x-ray generation and acquisition conditions. These routines are usually developed in consultation with the xray supplier. SCOPE OF APPLICATION: We used XRF technology to quantify platinum content of wrought prepared samples representative of the following alloys and melt lot sizes: 95Pt-5Co 2000-6000g lots 95Pt-5Ru 12000-21000g lots 90Pt-10Ir 12000-21000g lots XRF methods were not used on 95Pt-5Pd, 95Pt-5Ir or 95Pt-5 heat treatable alloys. In all cases, samples were extracted from wrought (cold worked and annealed) sections of a billet. The pieces were approximately 1” long, ½” wide and 3/16” thick. They were always rolled flat, ground on a metallographic lapping wheel with 240, 400 and 600 grit silicon carbide papers followed by polishing with 6 and 1 micron diamond paste. The mirror bright and flat nature of the prepared surface was critical to enhance precision and accuracy levels. The selected sample piece must be big enough to hold in the fingertips for this standard preparation method. The actual x-ray process was non-destructive to the sample, allowing for recycle back into subsequent melt lots. The segregated and nonhomogenous nature of cast articles was not a factor in any of our analysis. Figure 4 Figure 5 © Platinum Guild International USA 2002 All Rights Reserved 3 Platinum Manufacturing Process Platinum Day Symposium Volume IX N5 Because of the wide 8mm xray spot size, finished jewelry articles with concave or convex surfaces, that were potentially rhodium plated were not used for any analysis. ANALYTICAL PRINCIPLES: From an x-ray source perspective, a 50Watt system does not have enough excitation energy to use a secondary target. This was found to be a critical factor contributing to high precision and accuracy. A 400Watt system produced enough x-ray power to excite a secondary target and produce high energy filtered monochromatic x-rays. The clean Tin x-rays produced a high signal to noise ratio and were able to excite the heavy PGM elements. Contrast Figure 5 with Figure 2. The spectrums represent the same specimen. Figure 2 clearly shows distinct peaks with minimal noise and overlap versus Figure 5. From a statistical perspective, we always tried to achieve 1 million x-ray counts for analysis. This was achieved through 200800 seconds of x-ray acquisition. Analysis was done in triplicate to confirm precision on each x-ray session. A standard which was validated by alternative analytical methods was used to confirm minimal detector drift. Statistical manipulation of the x-ray counts is based on linear regression analysis techniques to maximize accuracy. Other methods such as standard less fundamental parameters (SLFP) have 10-20% relative accuracy. Even using the NBS-GSC method where a known standard is analyzed in tandem with the unknown specimen only yields 510% relative accuracy. To fulfill the requirements of the linear regression model, a specific number of standards that are validated by 4 Figure 6 a separate analytical procedure must be used. The model dictates that (n+1)2 standards are required, where n equals the number of distinct elements in the alloy, not just the number being analyzed. For a typical binary platinum alloy, 9 standards must be fabricated, validated by some accepted method such as I.C.P or AA spectroscopy and their spectrums acquired by XRF methods. For a three element alloy, 16 standards are required. The re- © Platinum Guild International USA 2002 All Rights Reserved sults of these acquisitions were numerically managed to produce a regression curve with a high coefficient of co-relation. This greatly enhanced accuracy. The goal was to establish a curve where unknowns are analyzed by interpolation as opposed to extrapolation. Figure 6 illustrates the concept. All elements in the alloy must be accounted for because the software will normalize the results to 100% and create inaccuracies. The same basic Platinum Manufacturing Process Platinum Day Symposium Volume IX N5 statistical principles are used with wet chemical methods such as A.A or ICP. FABRICATION OF STANDARDS: Our platinum binary alloys were made to ±0.50% of the nominal platinum content in 0.10% increments. We validated the standards by A.A. methods internally and had an external lab use I.C.P. to confirm our values. For the 90Pt-10Ir we had standards from 89.5% Pt to 90.5% Pt in 0.10% increments. Likewise for the 95% alloys, we had standards from 94.5%Pt to 95.5%Pt in 0.10% increments. This furnished high quality regression curves where linear interpolation was used to determine the platinum content of unknown specimens. All standards received surface preparation by grinding and diamond polishing from wrought microstructures produced through a series of cold mill rolling and furnace annealing operations to minimize potential segregation issues. Figure 7 EXAMPLE OF A SAMPLE ANALYSIS: Samples were loaded into an automated rotating carrousel. Analysis can be completed overnight without supervision. Figure 7 shows the equipment. The software is Windows based for a standard PC. Figure 8 illustrates a typical screen for the operator where batches are defined and routines selected by cursor and mouse movement methods. The unique sample ID can be entered and either standard or special acquisition parameters defined for each specimen. The results of a 600 second acquisition of 95Pt5Ru are summarized in Figure 9. Final calculations for platinum and ruthenium content are sum- Figure 8 © Platinum Guild International USA 2002 All Rights Reserved 5 Platinum Manufacturing Process Platinum Day Symposium Volume IX N5 marized on the screen shown in Figure 10. These results can be printed in report form or transferred with a computer network for storage. The entire process of standard acquisition and tripli- cate analysis of 6-8 unknown specimens can be completed in an unsupervised overnight routine. Each specimen may receive 30-40 total minutes of x-ray irradiation in three separate analysis that can Figure 9 be compared to confirm a high level of precision. Results are printed in the morning to quantify platinum content. Results of Analysis 1997 & 1998 via A.A spectroscopy versus 1999 & 2000 with XRF. Results summarized for each platinum alloy allow comparison between A.A spectroscopy in 1997-98 versus XRF in 19992000. Statistical indicators such as sample size (N), average, and standard deviation describe the attributes of the assay results (population). Tests such as the “F Test” and “T Test” were run to provide a measure of similarity between XRF and AA assay results. Results of assay populations can be considered in terms of the normal distribution shown in Figure 11. A range of ± 2σ, or 2 times the standard deviation around the average encompasses a range involving 95% of the results. If 100 XRF analysis sessions were conducted on the same sample, results form 95 of these would fall into the average ± 2σ range. The narrower this range the better. 95% PLATINUM 5% COBALT Results are summarized in Table 1. XRF platinum assays produced a superior average compared to the 95.20% Pt content target, a smaller standard deviation and a smaller ± 2σ range. Figure 12 shows a distinct trend. The results may be indicative of actual melting skill improvement, not specifically a difference between analysis methods. Sample sizes Figure 10 6 © Platinum Guild International USA 2002 All Rights Reserved Platinum Manufacturing Process Platinum Day Symposium Volume IX N5 95Pt- 5Co Year/method 2000/XRF 1999/XRF 1998/AA 1997/AA T bl 1 N 30 34 37 27 # kg 66 68 115 79 were very similar. There was minimal interference between platinum AVERAGE andSTD DEV +/-2SIGMA cobalt peaks. The narrow 95% 0.9515 range 0.00072 and quality0.9498-0.9529 average indicate 0.9522 that0.00086 XRF handles0.9504-0.9539 Pt determination accuracy and0.9506-0.9546 precision for this 0.9526 with0.0010 alloy. 0.9540 0.0017 0.9574-0.9506 This was symptomatic of peak overlap and interference issues that required attention. In 2000 a smaller population of results had a much lower average assay and standard deviation comparable to 1997 values with AA methods. After a period of adjustment the XRF method produced results of comparable accuracy and precision to standard AA practise. 95% PLATINUM 5% RUTHENIUM A similar trend occurred with this alloy. Results are summarized in Table 3 and Figure 14. The first year of XRF techniques (1999) showed a distinct jump in average assay and standard deviation of the population. By the end of the second year (2000) both the average and standard deviation values closely match those from AA in 1997-98 for a much large sample population. This level Figure 11 90Pt- 10Ir Year/method 2000/XRF 1999/XRF 1998/AA 1997/AA N 29 50 165 62 95Pt- 5Ru Year/method N 2000/XRF 111 1999/XRF 162 1998/AA 36 1997/AA 65 90% PLATINUM # Kg 160 309 1380 390 AVERAGE 10% STD DEV +/-2SIGMA IRIDIUM 0.9031 0.0014 are summarized 0.9059-0.9003 Results in Ta0.9048 ble 20.0018 0.9084-0.9012 and Figure 13. AA methods good results in 1997&98. 0.9020 produced 0.0085 0.9037-0.9003 large volumes0.9054-0.8990 handled in 1998 0.9022 The 0.0016 # kg 1082 1345 198 463 exhibited a good average and comparatively narrow standard deviation. Initial application of XRF AVERAGE techniques STD DEV in 1999 +/-2SIGMA experienced a period where sample 0.9528 learning 0.00148 0.9557-0.9498 preparation and standards were ad0.9536 0.0020 0.9576-0.9496 justed to compensate for a high av0.9527 0.0012 0.9551-0.9503 erage assay and standard deviation. 0.9527 0.0013 0.9553-0.9501 © Platinum Guild International USA 2002 All Rights Reserved 7 Platinum Manufacturing Process Platinum Day Symposium Volume IX N5 Figure 12 Fugure 13 Method N AVERAGE STD DEV XRF 22 0.9518 0.0010 A.A 22 0.9516 0.0012 t-test F-test Pass@ 95% CONF. 0.916<2.08 Pass @ 95% CONF. 0.694<2.46 t-test F-test Fail @ 95% CONF 2.93>2.09 Pass @ 95% CONF 0.429<2.30 T bl 4 Method N AVERAGE STD DEV XRF-AA 22 0.0011 0.0008 A.A 22 ±0.0003 0.0012 T bl 5 8 © Platinum Guild International USA 2002 All Rights Reserved indicated that XRF results have a high degree of precision and of accuracy comparable to conventional techniques. Another method of comparing the two analytical methods involved applying the “F-test” and “t-test” to a population of 22 assays of 95% Pt-5% Ru that were executed with both methods. Results are summarized in Table 4. The F-test evaluates the equivalence of two populations variances (1 set of results from the XRF assay of Pt, the other from the AA assay of the same Pt samples). Results pass the test, which implies that the two populations have a similar variance. In other words the standard deviation of Pt assays from XRF methods closely match that of AA methods. Likewise, the student t-test evaluates the equivalence of two population averages. Again, the XRF results closely match the AA results for this population of 22 assays completed by both methods. One last comparison was made between the difference in the XRF results versus the AA results. Results are summarized in Table 5. The difference in the averages for the populations was 0.0011 which closely approximated the target of 0.10% accuracy. The value of ±0.0003% was considered the absolute level of accuracy possible with correct application of the standard method. When comparing this 0.0006 range attainable to the 0.0011 range obtained those results failed the t-test for population similarity by a small margin. The deviations in the two populations passed the F-test, indicating a close match in values. Overall, results from the application of the XRF method in direct comparison with the standard AA method for this population of 22 platinum Platinum Manufacturing Process Platinum Day Symposium Volume IX N5 95Pt5Ru ASSAY RESULTS 95.4 0.0025 0.002 95.35 0.0015 95.3 0.001 95.25 0.0005 0 95.2 1997 1998 1999 2000 Figure 14 assays, compare favorably for accuracy. CONCLUSIONS AND FURTHER RESEARCH: From a large number of assay results over a 2 year period, we concluded that with proper care and attention to specific details, an XRF method could quantitatively assay platinum content to 0.10% accuracy with repeatable precision. A 4 year period of results allowed for comparison between population averages and standard deviations between AA spectroscopy and XRF. Again, results com © Platinum Guild International USA 2002 All Rights Reserved pared favorably. We believe that a high energy x-ray source capable of generating secondary mono-chromatic x-rays presented to a carefully prepared, wrought flat sample is required to attain the highest level of accuracy and precision. The method can be applied to binary platinum alloys with the use of 9 standards carefully matched to the specific alloy matrix validated by some standard wet chemical method such as AA or ICP. The method has been proven to work for the alloys 95% Pt-5%Co, 95%Pt-5%Ru & 90%Pt-10%Ir. The method may also apply to such alloys as 95Pt5%Pd, 95%Pt-5%Ir and 95%Pt5%HTA. We plan to continue testing the application of XRF methods to additional alloys and expand the population of direct comparison results that confirm accuracy and precision between novel and conventional analytical methods. Our ultimate goal is to have XRF methods accepted as standard practice in the trade. 9