Forensic gemstone analysis faces many challenges. One of the most critical aspects of gemological research is the classification of a gemstone. It is necessary to understand the mineral species, purity, origin, and identification of treatments to identify and classify a gem correctly, as well as assign a monetary value to it. However, the identification methodology must be neither destructive nor invasive to the material. Raman spectroscopy offers several advantages, such as fast, non-destructive analysis, requiring little material, being fully reproducible, and requiring no sample preparation. Furthermore, the development of portable Raman spectrometers allows the analysis to be carried out anywhere. Fourteen Brazilian gemstones (agate, amazonite, amethyst, yellow calcite, orange calcite, citrine, hematite, yellow jasper, tiger's eye, onyx, blue quartz, rose quartz, green quartz, and sodalite) were analyzed using the TacticID N 785 nm portable Raman system. For each of the fourteen types of gemstones, three fragments of rough and three polished gemstones were analyzed. Three spectra of each piece in different faces were taken, totaling nine spectra for each species and 252 spectra in all. We used one univariate technique (Pearson correlation coefficient) and one classificatory technique-Soft Independent Modelling of Class Analogy (SIMCA) to evaluate the spectra. Results showed that the univariate method was not able to discriminate the classes. However, SIMCA analysis showed promising results for classification. These results were achieved after removing outliers and making a variable selection. More specifically, some samples presented anomalous behavior due to the nature of the technique applied. Some regions of the spectrum did not offer information, or that information was redundant; because of these, it was removed. This suggests that Raman spectroscopy and multivariate methods can be essential tools for classifying gemstones, not only Brazilian ones but of any origin.
Read full abstract