ASTM D7343 Standard Practice for Optimization, Sample Handling, Calibration, and Validation of X-ray Fluorescence Spectrometry Methods for Elemental Analysis of Petroleum Products and Lubricants
9. Measurement of Standard/Sample
9.1 Ensure all support gases are turned on, for example, helium or nitrogen if this is required, or the supply gas for the flow counter. Check that gas pressure is set to the appropriate level for the instrument. Also, check the cooling medium for the detectors (if required), and verify the temperature stability of the instrument.
9.2 Handling of samples once prepared for measurement requires consideration prior to the establishment of an analytical method. Some petroleum products are unstable at room temperature and pressure. Samples can lose analytes preferentially giving rise to erroneous results.
9.2.1 When measuring volatile samples, a measurement protocol should be developed to ensure that sample loss will not affect a result. Suggested procedures to consider are as follows:
9.2.1.1 Only prepare a sample just prior to analysis. All standards and samples should follow a strict time-based preparation procedure. If a sample is prepared outside this time base or a measurement is delayed for any reason, then a new sample should be prepared. This should include the preparation of a new sample cup and film.
9.2.1.2 Design a measurement strategy to monitor the effect of sample loss during analysis. The most common method for this is the measurement of a matrix sensitive portion of the spectrum such as the Compton peak. The ratio of this peak to the analyte peak can correct for changes in samples, however constant mass or volume should be used if this method is to provide accurate correction for physical changes in a sample during measurement.
9.2.1.3 Use of an automated sample handling system should be avoided if a sample undergoes any changes during the time from preparation to completion of analysis, or if sample loss is expected. Common matrix types that this would apply to are gasoline, diesel, and sample types that cause any change to the sample support film, like relaxation or permeation. This is because samples will be stored in the sample handler for varying times giving rise to drift or bias in results.
10. Calibration of Acquired Data (Regression)
10.1 Once intensity data has been generated using either a WDXRF or EDXRF spectrometer, the treatment of the data is essentially the same. Within regression, a number of corrections can be made for background, line overlap, and matrix. For each correction made, there is an associated error. It is important to minimize the number of corrections to minimize the overall error on any element analysis.
10.1.1 Background corrections can be carried out in many ways. WDXRF makes discrete measurements of background around and near the element peak of interest, determines the amount of background existing beneath a peak, and removes it. EDXRF deconvolutes a spectrum and uses various mathematical background fitting models to remove background from a result. Background can also be removed by measuring a region of background close to the peak of interest, which does not contain any element, and establishing a relationship between this background region and the background that exists beneath the peak of interest. This latter technique is referred to as measured background subtraction. For trace element analysis where very small peak intensities often exist on relatively high backgrounds, the method used to measure and remove background is critical. A careful assessment of the most appropriate method should be undertaken by looking at the associated errors involved and whether background removal is needed at all. In petroleum products, backgrounds are often considerable.
10.1.2 Line overlaps of any type require the measurement of a component of an overlapping line that is related directly to the overlap of the element of interest. The overlapping line is then subtracted from the element of interest by the measurement of the related component and applying an overlap factor to remove the overlap. This is often achieved via the measurement of a K series line where the L series overlap. It is important that the statistical counting error of the overlap function does not add significantly to the error of the overall analysis. The coefficient of variation for each overlapping line should be at least equal to that of the analyte line of interest.
10.1.3 Various matrix correction models are available to spectroscopists for XRF. The two most common are the application of theoretical concentration alphas and ratio to Compton or Compton Correction. Theoretical alphas have the advantage of not adding additional error via the correction itself but are restricted in that all major constituents need to have a concentration associated with them and it is not advisable to apply them over very wide concentration ranges. Compton correction requires the measurement of the X-ray tube most sensitive α Compton line. This is often the Kα line. For low atomic number tube types, this line is often not available for measurement. Statistical counting errors should be carefully assessed if this method is used. For elements with lines lower in energy and greater in wavelength than the first major element in a sample, this method should not be used.
10.1.4 Infinite thickness is an important factor for the determination of high atomic number elements in petroleum products. To use matrix correction methods and to correctly assess and remove backgrounds and overlaps, samples should be infinitely thick with respect to all elements measured. Where this is not the case, using a fixed mass standard/sample should be employed.
11. Validating an Analysis
11.1 Once a standard or sample has been measured, a procedure should be carried out to validate that measurement. This requires the operator to check for obvious signs of damage to the sample such as leaking sample cells, crinkled sample cell window (especially important for low atomic number elements atomic number <13) and inspection of any secondary film.
11.2 If a measured result is considered to be outside the range of expected results, the sample should be prepared again and reanalyzed.
11.3 Regular checks should be carried out to ensure that purging gas performance is within the instrument manufacturer's specification.
11.4 Quality control standards/monitors should be run on a regular basis. The tolerance levels of the checks made using these monitors should be such that a protocol of either drift correction or total recalibration is carried out if the results fall outside these levels. All measurements between the last accepted monitor result and the point of non-compliance should be repeated, if a current monitor measurement proves to be outside acceptable levels.
11.4.1 Selection of calibration samples and elemental interferences are addressed in the test methods listed in the Referenced Documents (2).
11.4.2 The use of correct statistical quality control (SQC) techniques with all XRF methods is highly encouraged. It is recommended that at least one type of quality control sample be analyzed that is representative of typical laboratory samples as defined in Practice D6299.
12. Keywords
12.1 calibration; EDXRF; regression; validation; WDXRF
