STUDIES IN PEAK RECOGNITION FOR GAS AND LIQUID CHROMATOGRAPHY: LIBRARY SEARCHING OF VAPOR PHASE INFRARED SPECTRA AND SELECTION OF REPRESENTATIVE WAVELENGTHS FOR ULTRAVIOLET-VISIBLE SPECTRA (CHEMOMETRICS,OPTIMIZATION).

摘要

A series of studies addresses the issue of peak recognition for solutes eluting from gas or liquid chromatographs. In the first part of the dissertation, vapor phase infrared (VPIR) spectra are used for the recognition of solutes eluting from a gas chromatograph. In the second part, UV-visible spectra are used in conjunction with liquid chromatography. The goal of qualitative solute identification is common to both sections, although the specific approaches used for identification differ.; A semi-automated approach is used to generate a correlation table for the hydroxyl stretching region of VPIR spectra for 416 hydroxyl containing compounds. For VPIR spectra, the hydroxyl stretching region is found to be of significant value for structure elucidation. Peak shifts due to intramolecular hydrogen bonding and steric crowding are observed and discussed.; A reduced representation for VPIR spectra is presented, based on the suppression of intensity information to peak present/absent and the encoding of a fraction of the width of each spectral band. The optimal degree of width encoding is predicted, and then tested by use of the width enhanced binary spectra in a library search (LS) system.; A procedure is presented for the quantitative evaluation of the performance of LS systems employed for molecular structure elucidation. This procedure allows the evaluation of any form of spectral representation or any spectral comparison metric in a quantitative and objective manner.; Four alternatives for the selection of representative wavelength sets from a collection of 101 UV-visible spectra are surveyed. The cumulative information content, corrected for correlations between selected wavelengths, is used to evaluate the set of representative wavelengths selected by each procedure.; Building upon the results of the survey study, the selection procedures are applied to a set of UV-visible spectra for nine cephalosporin solutes, for the purpose of selecting wavelengths to be used in the monitoring of absorbance ratiograms during liquid chromatographic separations. The key set factor analysis approach is considered in detail. The limitations of this approach are explored, and an alternative procedure is presented which provides more effective wavelength selections.