Applying Mechanistic Understanding of Optical Absorption and Scattering Phenomena to Enhance the Spectroscopic Analyses of Pharmaceutical Materials

The dissertation uses spatially-resolved spectroscopy to separate absorption and scattering behaviors when NIR light interacts with pharmaceutical materials. The separated absorption and scattering were utilized to enhance mechanistic understanding of NIR diffuse reflectance spectroscopy and improve practical spectroscopic analysis in pharmaceutical applications. Near-Infrared (NIR) chemical imaging based spatially-resolved spectroscopy was used to measure radially-diffused reflectance on pharmaceutical materials. A Monte Carlo simulation based partial least square (PLS) model was constructed to determine the absorption and reduced scattering coefficients in pharmaceutical samples from the measured radially-diffused reflectance. The separated absorption and reduced scattering coefficients were combined with Monte Carlo simulation to provide understanding of the effects of physical properties (e.g., particle size and tablet density) on NIR spectral responses, including absorption and depth of penetration profiles. It was discovered that absorption and reduced scattering coefficients are the dominant factors in determining NIR absorbance and depth of penetration profiles, respectively. Both empirical measurements and Monte Carlo simulation were used to explore the photon radial movements in a chemical imaging setting. It is well understood that radial photon movements among adjacent pixels leaded to blurred 2-D chemical images. A Monte Carlo simulation based deconvolution filter was developed to sharpen a blurred feature in a 2-D image while maintaining the original chemical content of that feature. A new scattering correction method via the reduced scattering coefficient was proposed to specifically reduce physical interference with predictions of chemical properties. The wavelength- and absorption- dependent properties of the reduced scattering coefficient were found to allow specific suppression of physical interference and maintain the original chemical information. Combing the separated optical coefficients with contemporary efficient calibration approaches was found to simplify multivariate model calibration using a reduced calibration dataset, allowing parsimonious multivariate models, and reaching the same or even lower prediction error. To the author\u27s best knowledge, this work is the first example of the application of spatially-resolved spectroscopy to the pharmaceutical field. The enhanced understanding and improved spectroscopic analysis demonstrated in this dissertation is expected to provide groundwork for a wide variety of applications of spatially-resolved spectroscopy in pharmaceutical analyses