Forensic analysis of glass by μ-XRF, SN-ICP-MS, LA-ICP-MS and LA-ICP-OES: evaluation of the performance of different criteria for comparing elemental composition

Four interlaboratory tests were designed to evaluate the performance of match criteria for forensic comparisons of elemental composition of glass by μ-XRF, solution nebulization SN-ICP-MS, LA-ICP-OES and LA-ICP-MS. A total of 24 analysts in 18 laboratories participated in the tests. Glass specimens were selected to study the capabilities of the techniques to discriminate glass produced in the same manufacturing plant at different time intervals and to associate samples that originated from a single source. The assessment of the effectiveness of several match criteria included: confidence interval (±6s, ±5s, ±4s, ±3s, ±2s), modified confidence interval, t-test, range overlap, and Hotelling's T2. Error rates are reported for each of these criteria. Recommended match criteria were those found to produce the lowest combinations of type 1 and type 2 error rates. Performance of the studied match criteria was dependent on the homogeneity of the glass sources, the repeatability between analytical measurements, and the number of elements that were measured. The best results for μ-XRF data were obtained using spectral overlay followed by a ±3s confidence interval or range overlap. For ICP-based measurements, a wider match criterion, such as a modified confidence interval based on a fixed minimum relative standard deviation (±4s, >3–5% RSD), is recommended due to the inherent precision of those methods (typically <1–5% RSD) and the greater number of elements measured. Glass samples that were manufactured in different plants, or at the same plant weeks or months apart, were readily differentiated by elemental composition when analyzed by these sensitive methods.