Quasi 'non-destructive' laser ablation-inductively coupled plasma-mass spectrometry fingerprinting of sapphires

A homogenized 193 nm excimer laser with a flat-top beam profile was used to study the capabilities of LA-ICP-MS for 'quasi' non-destructive fingerprinting and sourcing of sapphires from different locations. Sapphires contain 97-99% of Al2O3 (corundum), with the remainder composed of several trace elements, which can be used to distinguish the origin of these gemstones. The ablation behavior of sapphires, as well as the minimum quantity of sample removal that is required to determine these trace elements, was investigated. The optimum ablation conditions were a fluency of 6 J cm(-2), a crater diameter of 120 mum, and a laser repetition rate of 10 Hz. The optimum time for the ablation was determined to be 2 s, equivalent to 20 laser pulses. The mean sample removal was 60 mn per pulse (approx. 3 ng per pulse). This allowed satisfactory trace element determination, and was found to cause the minimum amount of damage, while allowing for the fingerprinting of sapphires. More than 40 isotopes were measured using different spatial resolutions (20-120 mum) and eight elements were reproducibly detected in 25 sapphire samples from five different locations. The reproducibility of the trace element distribution is limited by the heterogeneity of the sample. The mean of five or more replicate analyses per sample was used. Calibration was carried out using NIST 612 glass reference material as external standard. The linear dynamic range of the ICP-MS (nine orders of magnitude) allowed the use of Al, the major element in sapphire, as an internal standard. The limits of detection for most of the light elements were in the mug g(-1) range and were better for heavier elements (mass > 85), being in the 0.1 mug g(-1) range. The accuracy of the determinations was demonstrated by comparison with XRF analyses of the same set of samples. Using the quantitative analyses obtained using LA-ICP-MS, natural sapphires from five different origins were statistically classified using ternary plots and principal multi-component analysis. (C) 2001 Elsevier Science B.V. All rights reserved.