The physical principles of XRF calibrations

Full text: XRF Control and calibration software has come a long way in recent years. Advances in the multiple regression software sophistication and speed of computers have provided an essential resource to the XRF analyst. Over recent years there has been a trend amongst some analysts to develop XRF calibrations based exclusively on the statistical information given by calibration software. Multiple regression statistics are designed for non-correlated data sets but give unpredictable results if there are significant correlations in the standards used. This is typical for calibrations weighted towards certified reference materials (CRM's) which, being natural materials, contain correlated concentrations. Purely statistical methods in calibration development have applicability only over very short concentration ranges and for materials whose composition varies little. Beyond these ranges, the calibration has the potential to be unstable and has been known to produce significant deviations in analysis of unknown samples. The statistical information generated during XRF calibrations can be a very useful tool when used in conjunction with knowledge of the physics behind the correction factors applied. The matrix coefficients represent physical absorption/enhancement effects within the sample and are not arbitrary numbers used to get a good fit to the calibration line. Inappropriate use of matrix factors and overlap factors can produce low RMS values but erroneous results in unknown samples. This talk will contain examples to demonstrate hazards with different calibration strategies and will include coverage of the following topics: physical effects occurring within the sample as a result of X-ray irradiation; use of multiple regression statistics and what role it plays in the calibration; calibration strategies using synthetic and CRM standards; determining appropriate theoretical and semi-empirical matrix corrections and line overlap factors. Copyright (1999) Australian X-ray Analytical Association Inc