Evaluating obsidian calibration sets with portable X-Ray fluorescence (ED-XRF) instruments

Abstract

Available calibrations for obsidian transform X-ray fluorescence (XRF) spectral data into analytical units such as parts per million or weight percent through complex algorithms built around known geologic rock references (USGS and NIST) or independent matrix matched sets (Bruker from MURR and PYRO from Yale) characterized by other analytical methods, complex physics-based models of atomic energy dispersion, or both. The current portable XRF (pXRF) industry uses two main part per million predictive methods: Lucas-Tooth empirical calibration (regression-based) and Fundamental Parameters with standards. These procedures are often exclusive to a given pXRF manufacturer and claims of superiority are often cited. In addition to these associations, obsidian calibration sets are also available to either create your own regression calibration or fine-tune a Fundamental Parameters algorithm using a post-processing slope correction, thus potentially complicating pXRF accuracy through unique user customizations beyond claims by the manufacturer. This paper reviews calibration methods and examines the accuracy of three in-use pXRF instruments with respect to calibration methods and obsidian calibration sets thereby independently evaluating claims of accuracy by manufacturers. Research of this kind serves to inform new users of how to evaluate and disclose instrument and calibration accuracy before the analysis of archaeological materials.