Resolution and non-linear spectral studies in multivariate optical computing

Multivariate optical computing (MOC) is a method of multivariate calibration that utilizes a specially designed interference filter called a multivariate optical element (MOE) to perform an optical regression for the prediction of an analyte characteristic of interest. The first portion of this dissertation inspects two areas of the MOC methodology. First, there is an examination of the effects of spectral resolution with regard to the design of MOEs. In this study, the question of whether or not low resolution calibration data can produce MOEs with predictive ability for high resolution applications is tested on a binary system of naphthalene and pyrene in the near infrared (NIR). Validation results are analyzed using traditional chemometric methods (e.g., principal component regression) in addition to MOC. It was found that it is possible to use lower resolution calibration data to build both digital and optical calibration models, however, there is sensitivity with respect to sample pathlength.;Second, the effects of spectral non-linearity on the design of MOEs are investigated. Causes of spectral non-linearity are discussed as well as how spectral non-linearity is a concern for MOC because of its nature to use raw spectral data. Several new methods are proposed to avoid the linearity requirement - using both digital chemometric methods, as well as MOC. As a result, these new methods stand to be new, and yet computationally less intensive methods to construct calibration models for systems with a degree of spectral non-linearity.;The next section of this dissertation describes the building of a single-phytoplankton fluorescence excitation spectroscopy instrument, and preliminary data collected from the instrument. Two common phytoplankton species, T. pseudonana and E. huxelyi are used in this study to determine if the use of an optical trap in conjunction with continuous wavelength excitation can provide a means of discrimination between phytoplankton species. Preliminary chemometric work shows that in fact, different species can be separated using spectroscopic means alone.;In addition to the MOC work, this dissertation also discusses the development of an undergraduate physical chemistry experiment that introduces undergraduates to chemical polymorphism. Differential scanning calorimetry (DSC) is employed to probe the heat of fusion and melting point temperature of two chemical polymorph pairs: acetaminophen (i.e., Tylenol) and neotame. The polymorphic relationship of these two polymorphic forms are determined by students using the Burger and Ramberger rules and the data collected, and the students report their findings and argue for which relationship best applies to acetaminophen or neotame.