The chemical analysis of rock powders by automatic X-ray fluorescence

Abstract

This account presents detailed calibrations which can be used in other laboratories for the ultra-rapid X-ray chemical analysis of rock powders for 38 major and trace elements. The major element calibrations are based on 16,000 measurements on about 400 previously analysed rocks and working equations are given for rhyolites, granodiorites, granites etc., amphibolites, basalts, dolerites, pelites, semipelites, carbonate-rocks, and psammites. Trace elements have been calibrated by spiking nine carefully analysed rocks with 1,500 different amounts of added spectrographically pure chemicals. All the calibrations have been computer-calculated and are given in terms of ratios of counts relative to one rock, an amphibolite, which is available for general distribution and thus enables other laboratories to use the calibrations. By this ratio technique very high precision has been obtained. The results of extensive investigations into the optimum conditions for the determination of the 38 elements as regards X-ray tubes (Cr, W or Mo); kV and mA; analysing crystals; peak and background positions; counters etc. are summarised in tables. A new specimen preparation technique enables all elements above F to be determined on one permanent rock pellet which is 86% rock, 14% binder. Detection limits are generally low for an X-ray technique, being 0.02% for MgO, 0.05% for Na 2O and below 10 p.p.m. for all elements above Cl in atomic number. Mass absorption corrections are applied in a novel way that is independent of international rock standards. It is shown, however, that for major oxides, mass absorption corrections are negligible within fairly wide ranges of a few rock types and are trivial compared with other effects, including mineralogy and pellet variation. It is demonstrated that octahedral and tetrahedrally-coordinated alumina give rise to different alumina calibrations and this effect is, for alumina, more significant than mass absorption effects, within the common range of rock compositions. Precision and accuracy are evaluated by a most extensive series of checks and statistical treatment of the data. The accuracy varies from extremely good (e.g., TiO 2, K 2O, Rb and Sr) to poor (S and P 2O 5) but overall the moderate accuracy and high speed justify the use of the method for serial analyses and element variation maps. Some 1,550 rocks have each been analysed for 23 elements in nine months.