Portable x-ray fluorescence calibrations: Workflow and guidelines for optimizing the analysis of geological samples

Abstract

The use of portable Energy Dispersive X-Ray Fluorescence (pXRF) spectrometers has exploded within the last decade for e.g., environmental, geological, and archaeological field research. Producing reliable geochemical data using a pXRF is a challenging process that requires the development of dedicated and matrix-matched empirical calibrations. Although a large field of applications, several research papers are still published using unquantified pXRF data, using manufacturers' built-in calibrations unsuitable for targeted rocks, or interpreting chemical concentrations without checking the limit of quantification of the pXRF spectrometer and its calibration model. In this paper, we present a general workflow for building an empirical calibration model for a pXRF spectrometer, from choosing reference materials to validating calibration's efficiency and quantifying its limit of quantification. We also compare the geochemical results on a sedimentary and an igneous rocks suite obtained from thirty-two calibration models that were built using different software available, including a simple homemade spreadsheet, the EasyCal calibration software, a free and open-source calibration app named CloudCal, and the built-in GeoMining and GeoExplorer calibrations provided by Bruker manufacturer on the Tracer 5 pXRF. This decoupled approach allows (i) to explore the pros and cons of each calibration model and determine which empirical calibration model brings the most accurate chemical concentration for a sedimentary and an igneous rock suite, (ii) to weigh the impact of each calibration parameter (type of regression line, spectral interferences, normalization, influence coefficients, multiple conditions of excitation and acquisition) on the accuracy of geochemical concentrations on rock samples and (iii) to bring innovative tools to assess the limits of quantification of a newly built calibration model.