Chemometric enhancement for blind signal resolution from non-invasive spatially offset Raman spectra

Spatially offset Raman spectroscopy (SORS) is a promising spectroscopic technique that enables the collection of Raman signals from deeper layers of materials compared to conventional Raman spectroscopy. SORS equipment allows acquiring Raman spectra of substances through packaging, making it a non-invasive analytical technique. Note that the acquired spectrum is always the result of mixing two contributions: (1) the spectrum of the container material, and (2) the spectrum of the substance inside the container. Nowadays, SORS equipment are supported by software capable of removing the surface contribution from the recorded spectral data. However, the optimal extraction of this contribution is not achieved in all cases. This study explored the potential of two chemometric methods, Multivariate Curve Resolution (MCR) and Independent Components Analysis (ICA), for resolving the mixed spectra acquired by Vaya Raman equipment (Agilent) of four standard substances (sucrose, anhydrous sodium sulphate, ethanol 96% and glycerol) analyzed through containers of two different plastic materials (polypropylene and polyethylene terephthalate). The two resulting resolved spectra (by MCR and by ICA) and the one resolved by the equipment software were compared by similarity analysis with the Raman spectra of the standard substances available in recognized databases intended as target spectra. Similarity analysis was performed by calculating four similarity indexes, namely: cosine, arccosine, coefficient of determination and nearness index. Both MCR and ICA methods successfully extracted pure Raman spectra of the test substances more similar to the standard substances than the spectra resolved by the equipment software. These results highlighted the potential of both methods for resolving complex mixed signals, such as those acquired through the SORS technique, thereby enhancing its utility.