Matrix-matched calibration in LA-ICP-MS of silicate, phosphate and carbonate minerals: application of G-Probe samples

Laser ablation (LA) sampling provides fast microelement ICP-MS analysis of a wide range of solid materials without their dissolution, thus decreasing contamination from water and reagents as well as reducing polyatomic isobaric interferences from acid solutions. However, the issue of matrix-matched calibration becomes crucial for LA-ICP-MS due to differences in behaviour during laser interaction and evaporation of solid samples. There are several approaches to LA calibration: simultaneous supply of standard solutions into a spray chamber; calibration using a set of NIST 61х synthetic glasses and glasses prepared from natural rocks and minerals (basalt, nephelinite, etc.) or pressed synthetic samples (calcium carbonates, phosphates and sulphides produced by USGS). A set of natural glasses for microanalysis is available from the International Association of Geoanalysts (IAG) in co-operation with the USGS. The G-Probe proficiency testing programme has been operating since 2008 and deals with solid samples for microanalysis (LA-ICP-MS, EPMA, EDS-SEM). A number of samples of different compositions were distributed: BBM-1G and BSWIR-1G natural basaltic glasses, GSM-1 gabbro; NIST SRM-based basaltic and diabase glasses; GP-MACS synthetic pressed calcium carbonate, GP-MAPS phosphate and some others. The aim of the present work was to estimate the LA-ICP-MS analysis quality using matrix-matched calibration with G-probe samples of various composition. All G-Probe samples were analysed using an ELAN 9000 Q-ICP-MS combined with a LSX-500 (Nd:YAG, 266 nm) laser ablation system. For silicate rocks, TB-1 basaltic glass was used for calibration; the remaining samples were analysed as unknowns. MAPS-4 calibration material were used for phosphate rock analysis. A combination of external matrix-matched calibration and internal normalisation was used for calculating element concentrations. LA-ICP-MS analysis quality was estimated using z-scores. Most of the results obtained were in a good agreement with assigned values.