On the performance of laser-induced breakdown spectroscopy for quantitative analysis of minor and trace elements in glass

Abstract

The analytical figures of merit of laser-induced breakdown spectroscopy (LIBS) for elemental analysis of glass have been evaluated using a laboratory prototype of the LIBS instrument for the quantification of 4 elements, Ti, Cr, Ca and Ba. Two sets of samples were prepared or collected for the assessment. The first one consisted of 10 laboratory-prepared fused beads with the elemental content determined by X-ray fluorescence (XRF), an established analytical technique which was considered in our study as the reference technique for the assessment of the LIBS technique. Among them, 8 were used as reference samples and 2 as “unknown” samples for test. The calibration curves were thus established with the references. The counter calibration led to the determination of the elemental content in the unknown samples. Such a calibration procedure allowed assessing the figures of merit of LIBS together with the used setup and measurement protocol about a certain number of key parameters, such as the correlation with a linear regression of the calibration data, limit of detection (LoD), repeatability, reproducibility and relative accuracy. The second set of samples was collected from different origins and consisted of 8 bottle glass fragments, which were different in appearance (color and surface) and in content for the 4 analyzed elements. Their elemental concentrations were first determined using XRF. The LIBS calibration curves established with the fused beads were thus used to perform the analysis of 2 glass fragments with elemental contents lying around the range of the calibration concentration. Further analysis of the ensemble of glass fragments allowed assessing the matrix effect introduced by the different types of glasses and extending the calibration curves over a very large concentration range from several ppm to several percent. We show that the self-absorption effect observed over such a large concentration range can be taken into account by using quadratic regression.