Abstract:
Glass is a common piece of trace evidence encountered in forensic casework, found in crimes such as hit-and-run accidents and burglaries. Glass fragments can be analysed in many ways such as; by observing physical characteristics (e.g. thickness, colour, or edges); analysing optical properties (e.g. refractive index or density) and by elemental analysis by techniques such as scanning electron microscopy energy dispersive spectroscopy (SEM-EDS), micro X-ray fluorescence (μXRF) and laser ablation inductively coupled plasma (LA-ICP-MS). Another technique gaining popularity in scientific fields such as archaeology, biology, geology and forensic science for the analysis of samples is laser induced breakdown spectroscopy (LIBS). LIBS is an atomic emission technique gaining popularity in forensic science due to advantages such as minimal sample preparation, high throughput and low cost. However, the disadvantages of using LIBS includes self-absorption and poor precision and repeatability resulting in variable spectra for the same sample for successive runs. This thesis continues a study by Gupta et al (2017), whom investigated the claim that there is higher variation in analyses conducted on different days compared to analyses conducted on the same day. The study examined variability in spectra produced from runs on the same day (intra-day) compared to runs on different days (inter-day) using standard reference glass and found inter-day variation was not significantly greater than intra-day variation at 5% significance level. Instead of analysing one glass type, this study analysed four glass types: standard reference glass, bottle glass, vehicle glass and building/window glass. The data was statistically analysed using additive log ratio (ALR) transformations along with frequentist and Bayesian statistics. A second study investigated whether fragment size influences LIBS spectra using standard window glass only. This thesis found that inter-day variation is not greater than intra-day variation for all glass types at 5% significance level. From these results, we can infer that LIBS could be a reliable tool for forensic casework and spectra from different samples could be used to build a database which is an essential tool for the interpretation of forensic glass evidence. Further research in this area must be carried out so that the overall objectives are attainable.