Raman spectroscopy is a valuable tool for elucidating the chemical structure and more of an unknown sample. A comprehensive collection of data, or library, is required for the proper identification of any material. Searchable spectral libraries have demonstrated value for the identification of a plethora of different forensic samples, such as drugs, organic pigments, and polymers. Mineral analysis presents another opportunity where a comprehensive searchable Raman spectral library could aid in the identification of samples for both geological and forensic purposes. However, while there have been several collections of mineral spectra created, there remains to be a comprehensive searchable Raman spectral library. Software programs like KnowItAll currently only have a few hundred mineral spectra, while other online databases like RRUFF2 allow one to view thousands of varying mineral spectra, but do not have a capability to compare an unknown sample to the library. This demonstrates the need for a comprehensive searchable mineral database to provide for their automated identification.
The purpose of this research is to develop and evaluate a searchable mineral library of Raman spectra to assist in identification of the minerals found in unknown soil samples through our ongoing research with Particle Correlated Raman Spectroscopy (PCRS). PCRS is an analytical technique which combines automated image analysis with Raman spectroscopy to provide morphological measurements and chemical identification of discrete particles within a complex mixture. Reliable and automated Raman spectral identification is critical for successful PCRS, however it must also be rapid (no more than a few seconds for Raman analysis) because of the large number of particles to be targeted for analysis. Another challenge for mineral identification by Raman spectroscopy is the natural variations within a mineral variety which can cause spectral differences. Factors such as polymorphism and twinning can cause minerals of the same composition to have different spectral peaks from Raman spectroscopy. Additionally, the fluorescent properties of some minerals can cause effects on the spectra that overtake the weaker Raman scattering effect, making it more challenging to identify some minerals. Thus, it is important to evaluate the capabilities of a purpose-built searchable and comprehensive Raman spectral library of minerals.
The capabilities for the automated spectral library identification of mineral grains in soil were evaluated using a known set of 60 comminuted soil minerals. The Rruff database was imported into KnowItAll to create the comprehensive mineral library, and the ability to correctly identify the 60 comminuted minerals was evaluated using two laser excitation wavelengths (785nm and 532nm). When comparing the automated mineral library correct identification rate before and after the addition of the Rruff database, as expected there was a noted increase in correct identifications of different minerals and a greater hit quality index regarding the strength of identification. With the addition of the Rruff mineral database to KnowItAll’s searchable library, correct identifications went from 15 of 60 (25%) to 47 of 60 (78.3%) minerals. The ability of this Raman searchable library to identify more mineral spectra demonstrates the importance of having such a comprehensive library available for analysis of unknown samples. The results of this research have the potential to be applied to an overall mineral identification procedure that would be useful for both the forensic science and geology fields.