Abstract
This work explores the potential use of spatially offset low-frequency anti-Stokes Raman spectroscopy (SOLFARS) to detect subsurface composition below an emissive surface. A range of bilayer tablets were used to evaluate this approach. Bilayer tablets differed in both the underlying layer composition (active pharmaceutical ingredient to excipient ratio, celecoxib: alpha-lactose monohydrate) and the upper layer thickness of the fluorescent coating (polyvinylpyrrolidone mixture with sunset yellow FCF dye). Two low-(<300 cm-1) plus mid-(300 to 1800 cm-1) frequency Raman instrumental setups, with lateral displacements for spatial analysis of solid dosage forms, using different excitation wavelengths were explored. The 532 nm system was used to illustrate how the low-frequency anti-Stokes Raman approach works with samples exhibiting extreme fluorescence/background emission interference, and the 785 nm system was used to demonstrate the performance when less extreme fluorescence/emission is present. Qualitative and quantitative chemometric analyses were performed to evaluate the performance of individual spectral domains and their combinations for the determination of the composition of the subsurface layer as well as the coating layer thickness. Overall, the commonly used midfrequency region (300-1800 cm-1) proved superior when using 785 nm incident laser for quantifying the coating thickness (amorphous materials), whereas a combined Stokes and anti-Stokes low-frequency region was found to be superior for quantifying underlying crystalline materials. When exploring individual spectral regions for subsurface composition using spatially offset measurements, the anti-Stokes LFR spectral window performed best. The anti-Stokes low-frequency range also demonstrated an advantage for models composed of data exhibiting high levels of fluorescence (e.g., data collected using 532 nm incident laser), as the Stokes scattering was masked by fluorescence. Transmission measurements were also explored for comparison and showed the best applicability for both upper and lower layer analysis, attributed to the inherently larger bulk sampling volume of this setup. From a practical perspective, these results highlight the potential adjustments that can be made to already existing (in-line) Raman setups to facilitate similar analysis in pharmaceutical industry-based settings.