January 12, Dr. Kevin Lackey TeckCominco Alaska Red Dog Mine Anchorage, AK Dear Dr. Lackey:

Transcription

1 January 12, 2004 Dr. Kevin Lackey TeckCominco Alaska Red Dog Mine Anchorage, AK Dear Dr. Lackey: Thank you for submitting another set of samples for the final evaluation of our QuanX EC EDXRF system. We have concluded our analysis and present the results in the attached report. In addition, we address all the questions and concerns you raised regarding XRF and the QuanX in particular. As promised in our first analysis report, we have optimized the latest method to achieve a two-fold reduction in analysis time without compromising performance. We are also pleased and relieved to report that multiple independent samplings of the same material (Unknown 2) yielded the same analytical results, which implies that the grinding performed on the materials is, in fact, the only sample preparation required for reliable analytical work. Of the five elements in this evaluation, only Silicon yielded less than outstanding results. Frankly, that was not surprising, since Si is a low-energy X-ray fluorescer, which makes it significantly more susceptible to matrix and sampling effects. Given that around 30% of the sample composition was not explicitly considered during the analysis, it is certainly possible for variations in that dark matter to affect Si specifically. Such complications are usually resolvable by adding standards to the calibration to better describe (or correct for) the matrix effects. According to the 2n+3 rule, the 14 standards provided allowed for application of no more than 5 matrix corrections. While five or fewer corrections were sufficient for the higher-energy analytes, Silicon will require about seven corrections or about 20 standards, to be conservative. Please, review the details below and feel free to contact Dave Leland or me with any questions. Sincerely, Anton Kleyn, Applications Engineer

2 EDXRF Analysis Report Thermo Electron Scientific Instruments Division INTRODUCTION Twenty-seven samples, all similar in appearance to the sample at right pictured in its analysis cup, were submitted for analysis of the five elements Ba, Pb, Zn, Fe and Si, ranging in concentrations from below 1% up to 50%. The method was calibrated using an Intensity Correction algorithm on the basis of fourteen standards provided by the customer. Three samples were analyzed as unknowns and another four samples were analyzed to evaluate the feasibility of the method for similar geological materials. PROCEDURE Instrumentation The samples were analyzed using the QuanX EC PCD system, which consists of: 50 kv Rhodium target X-ray tube <155 ev resolution Si(Li) Peltier-Cooled Detector Eight-position X-ray filter wheel Variety of sample-positioning options, including the 20-position automatic tray used for this analysis. Vacuum and/or helium chamber purge options The figure on the right illustrates the chamber environment. A personal computer controls all setup, data acquisition and analysis through Windows-based software. Sample Preparation In general, sample preparation is critical to accurate and reproducible XRF analysis. Since x-rays effectively probe only the first ~10 microns of the sample surface, any surface non-uniformities, contamination and variations in particle size could skew the analytical results. An ideal XRF sample would be a fused pellet, typically made by dissolving the sample powder in a lithium-tetraborate flux at high temperature and then cooling it to a glassy state. While fusing a sample is a laborious process, especially for high sample volumes, shortcuts in sample preparation often compromise the quality of results. Nevertheless, one goal of this evaluation was to determine if accurate and reproducible quantitative results could be obtained without the effort and time required by extensive sample preparation. Thus, no sample preparation was performed. The samples had been submitted as finely-ground powders, 325 mesh. These powders were transferred into standard xrf 31-mm plastic cups to a depth of about 1 cm and supported by Prolene-brand X-ray film. To minimize instrument contamination during rapid atmosphere changes (pumpdown or venting), the cups were covered with a microporous film that allows the air to escape from inside the cup without carrying the sample powder along. Red Dog.doc 1/12/04 1

3 Cost of Operation The cost of instrument operation per sample has two main sources: sample preparation supplies and instrument consumable(s): Sample Preparation Since the analysis requires no sample preparation beyond grinding, the cost of analysis for a single sample would that of the sample cup, the X-ray film, and the semipermeable film. Excluding operator time and assuming that supplies would be discarded after one use, the preparation-packaging cost would be less than $4 per sample. In theory, the sample cups are reusable, since they serve merely as containers for the powders and may be emptied, wiped and or blown-out rather easily. The exact amount of effort and care required to reuse the cups may be determined by a simple experiment that measures the potential of sample cross-contamination and its effects on analysis results. Even so, the cost of additional labor required to reuse the cups may not be justifiable. Instrument Consumables X-ray tube: Unlike other systems that employ shutters and supply constant power to the X-ray tube the only true consumable in the system - the QuanX EC powers the X-ray tube only during the actual analysis and automatically turns it off at the end of the run. At the quoted rate of ~1200 individual samples per month and 3 minutes of analysis time per sample, the X-ray tube would be powered for a mere 2 hrs/day on average. At this rate, it would be expected to last over 5 years. The current replacement cost of an X-ray tube is about $7,000. Over an estimated lifespan of 5 years, the cost of the tube per sample would be a mere $117 or less than 10 cents per sample. Detector: While not truly a consumable (i.e. increased use of a system will not shorten the detector lifespan), a detector s average life is about 5 years. Since detector replacement is a significant expense and is more or less predictable, it should be included in the operation cost estimate. At a replacement cost of about $10,000 (including labor), the detector s cost over 5 years would be $167 per month or 14 cents per sample. Service Contracts Several types of service contracts are available to minimize downtime and the cost of unexpected repairs. Service contracts also make the maintenance costs very predictable and amenable to yearly operational budgets. We typically recommend the Essential or the Comprehensive options, which differ only in terms of response time and the number of included visits per year. These contracts cost $7,500 and $10,000 per year, respectively. With the Comprehensive service contract, the operation cost would be $830 per month or 70 cents per sample. The overall analysis cost per sample, including instrument maintenance as part of a service contract is on the order of $5 per sample. Red Dog.doc 1/12/04 2

4 Additional Notes on Operation The QuanX EC system features several advantages from the standpoint of operational convenience:! Excellent analytical sensitivity and performance of traditional LN-cooled detectors without the LN! As a result, the instrument is a self-contained unit powered by a single standard electrical outlet. The PCD detector requires no babysitting, and unlike its distant cousin, the PIN detector, it shows the exclusive performance and reliability worthy of adult scientists.! Single-cable ethernet connection between the instrument and its computer allows for easier system setup, as well as faster and more robust data transfer. In fact, a QuanX EC can be taken out of a van, set up, cooled down, calibrated for an analysis, and placed back in the van all in the span of three hours. The ethernet implementation also permits easy use of customer s own computers, including a laptop, without the need to obtain or install special interface cards.! The autosampler is a very practical, single-motor carousel that is wholly enclosed in the analysis chamber. This configuration is truly optimized for high-throughput vacuum analysis, because the chamber is only pumped down once for up to twenty samples at a time. Thus, the 90-second pump-down becomes a mere 5-second delay per sample. Additional 20-sample trays can be provided to allow the operator to fill the next tray while the first one is being analyzed. With the samples packaged in their cups, it would take no more than a minute to place the cups into the tray and position the tray in the autosampler. The software is designed to work with multiple trays, so that the operator need only update the sample names to proceed with the next analysis.! With only two moving parts inside the sample chamber the autosampler and the filter wheel the possibility of mechanical breakdown is minimized. All sample chamber components are easily replaceable by a non-expert with a hex wrench. Red Dog.doc 1/12/04 3

5 Excitation Conditions Each sample was analyzed using the three excitation conditions shown in the following program image: The QuanX EC employs direct-filtered radiation to excite a sample and cause its constituent elements to fluoresce. Eight filters in the computer-controlled filter wheel optimize the excitation and peak-to-background for specific ranges of elements. Thus, a Pd filter was used to analyze Fe, Zn and Pb; a Cellulose filter optimizes the analysis of Ba, while no filter (i.e. direct excitation) is best for analysis of Al and Si. Every system is equipped with a default filter set specifically designed to make any QuanX EC a versatile, high-performance instrument that can be configured for any application within a few hours on-site. Analysis Time As indicated in the conditions diagram above, a complete analysis takes 30 x 3 = 90 seconds livetime, which is equivalent to 180 seconds or 3 minutes of real clock time. For single samples, up to 90 seconds may also be required to evacuate the chamber for analysis of light elements such as Al and Si. When a full tray of 20 samples is analyzed, the evacuation time would remain the same and only add 5 seconds to the analysis of each sample. Peak DeconvoIution and Integration Peak intensities were extracted using the XML peak-fitting and deconvolution method. The XML routine corrects for peak overlaps using a least-squares algorithm that fits stored reference spectra of pure analyte elements to the sample spectrum. All spectra are digitally filtered to remove any background before peak fitting is performed. Quantitative Analysis The method was calibrated using the fourteen standards provided by the customer. An Intensity-Correction algorithm was used to correct for the effect of the major sample components on each other. This calibration technique determines mathematical relationships between the intensity of one element and the intensity of another element. For example, an increase in lead intensity might correspond to a decrease in iron intensity because of increased absorption of iron x-rays by the lead atoms in the sample. Such a correlation would then be used to correct the iron intensity based on the observed lead intensity. Typically, all major sample constituents have a matrix effect on each other, even if not all of them are important from the analytical standpoint. Thus, a comprehensive analysis requires the measurement of Al, Sr and others only to correct for their effect on the Si, Fe, Zn, Pb, and Ba. Red Dog.doc 1/12/04 4

6 RESULTS The statistical results of ten consecutive measurements of each unknown sample are summarized in the following Excel spreadsheet. As reported in the first evaluation phase, the precision (1-sigma) for all elements is still within 1% relative for high concentrations and within 0.1% absolute for concentrations below 10% - but, of course, the analysis time has been cut by half, down to 3 minutes per sample. [Double-click the table below to access detailed results for all unknown samples] SiO2 Fe Zn Ba Pb Unknown 1 Average Sigma % RSD Unknown 2-1 Average Sigma % RSD Shaken Unknown 2-2 Average Sigma % RSD Shaken Unknown 2-3 Average Sigma % RSD Shaken Unknown 2-4 Average Sigma % RSD Shaken Unknown 3 Average Sigma % RSD Object 1 Summary of Unknown sample results As requested, Unknown 2 was analyzed as four roughly equally-sized samples of the material; to better simulate the conditions of daily operation, each sampling was analyzed in a different position in the autosampler tray. The results of ten consecutive measurements are reported in Object 1 as Unknown 2-1, 2-2, etc. After the ten replicate measurements, the same samples were shaken mercilessly, placed into another position in the autosampler tray and analyzed three consecutive times. The results of this subsequent analysis are reported in Object 1 as Shaken. Red Dog.doc 1/12/04 1

7 The results in Object 1 show that the four samplings of Unknown 2 were all within 3- sigma of each other, and most within 2-sigma. With the exception of Silicon, even after shaking the samples and setting them into different tray positions, the answers were still within 3-sigma or better. As expected, Silicon would be the analyte most sensitive to entrained air or material settling effects. The four flotation samples were analyzed quantitatively with the same method used for the other materials. These results are shown below in Object 2: SiO2 Fe Zn Ba Pb ZRT Assay Calculated Sigma ZF Assay Calculated Sigma PF Assay Calculated Sigma PRC Assay Calculated Sigma Object 2 Summary of Feasibility Study Given that three of the samples were clearly very different materials from those used to calibrate the method on the basis of the low Silica content alone the comparison between the Assay and Calculated results is quite good. Certainly, there is no indication that these materials would require a different approach or otherwise present an analytical challenge, provided that suitable standards are made available. Finally, the following results are provided in the Appendices: # Appendix A Detailed replicate measurements of the standards. # Appendix B Calibration curves for the five analytes of interest. Si, Fe and Zn were each calibrated with five corrections, while Ba and Pb required only four. # Appendix C Spectra of Standards 0 and 14 under each of the three analysis conditions. CONCLUSION The analysis results demonstrate the ability of the QuanX EC EDXRF system to analyze Fe, Zn, Pb, Ba and Si quickly, accurately and reproducibly. Red Dog.doc 1/12/04 2

8 APPENDIX A: STANDARD REPLICATES [Double-click the table below to access all individual measurements] SiO2 Fe Zn Ba Pb Standard 0 Given Calculated Sigma Standard 1 Given Calculated Sigma Standard 4 Given Calculated Sigma Standard 7 Given Calculated Sigma Standard 10 Given Calculated Sigma Standard 14 Given Calculated Sigma Standard 18 Given Calculated Sigma Standard 22 Given Calculated Sigma Standard 26 Given Calculated Sigma Standard 30 Given Calculated Sigma Standard 33 Given Calculated Sigma Red Dog.doc 1/12/04 1

9 APPENDIX B: CALIBRATION CURVES Calibration Curves for Fe (above) and Zn (below) Red Dog.doc 1/12/04 1

10 Calibration Curves for Ba (above) and Pb (below) Red Dog.doc 1/12/04 2

11 Calibration Curve for SiO2 Red Dog.doc 1/12/04 1

12 APPENDIX C: SAMPLE SPECTRA Figure C1 Spectrum of Standards 14 (in blue) and 0 (in red) acquired under conditions optimized for Al and Si, although S and secondary lines from Pb and even Zn also appear here. The small Rh peak is due to X- ray tube (Rh-target) radiation scattered from the sample. Red Dog.doc 1/12/04 1

13 Figure C2 Spectrum of Standards 14 (in blue) and 0 (in red) acquired under conditions optimized for Ba. Fe also appears under these conditions, but the higher background under the Fe peaks would make the analysis less than optimal. Red Dog.doc 1/12/04 2

14 Figure C3 Spectrum of Standards 14 (in blue) and 0 (in red) acquired under conditions optimized for Zn and Pb, as well as Fe. Peaks for Rb and Sr appear next to the Pb peaks in Standard 0. For reference, Sample 14 assayed 6.24% Pb and 12.93% Zn, while Standard 0 assayed 0.07% Pb and 0.29% Zn. The detection limit for Pb under these conditions with an analysis time of only 30 seconds is 12 ppm! Red Dog.doc 1/12/04 1