Application of multivariate statistics to hyperspectral imaging and Raman hydration shell spectroscopy

Advances in data acquisition and storage technologies have motivated the development of multivariate statistical techniques to handle and interpret large data sets. Such statistical methods may be readily applied to the spectral data routinely generated in analytical and physical chemistry to reveal underlying chemical information and to facilitate data compression and storage. This work focuses on the application of chemometrics to Raman hydration shell spectroscopy, multivariate hyperspectral imaging, and two-dimensional microwave spectroscopy. The former hydration shell spectroscopy is analyzed with multivariate curve resolution to decompose mixture spectra into pure component contributions. Subtle features not evident in the unprocessed spectra are revealed, including spectral features indicative of pi-hydrogen bonding in dilute aqueous benzene solutions and solvent-separated hydrophobic aggregation in tert-butyl alcohol mixtures. Chemometrics are also applied to hyperspectral imaging to develop a strategy which utilizes hardware-integrated data compression (using a digital micromirror device) to reduce the time necessary to classify and quantify components by orders of magnitude. Procedures including spectral aliasing and compressive sensing are applied to two-dimensional microwave spectroscopy to reduce and interpret the large data sets generated with the broadband method realized herein. The results of this work demonstrate the power and applicability of multivariate statistics to various disciplines of chemistry.