Computer-aided carbon-13 nuclear magnetic resonance spectrum simulation

This thesis describes research involving computer-assisted simulation of carbon-13 nuclear magnetic resonance ({dollar}sp{lcub}13{rcub}{dollar}C NMR) spectra. This work combines aspects of multivariate statistics, applied mathematics, and computational techniques for generating model equations for {dollar}sp{lcub}13{rcub}{dollar}C NMR chemical shift prediction.; Spectrum simulation methods allow approximate spectra to be constructed for comparison with observed spectra, thereby facilitating their interpretation. The spectrum simulation methodology described here is based upon the development of linear models relating observed {dollar}sp{lcub}13{rcub}{dollar}C NMR chemical shifts to groups of numerically encoded structural parameters (descriptors). This approach has been implemented into a computer-based {dollar}sp{lcub}13{rcub}{dollar}C NMR spectral simulation system.; The work presented in this thesis involves testing, expanding, and improving this spectrum simulation methodology. The first portion of this research examines the development of predictive models for two chemically important classes of compounds, while the second segment of this work focuses upon the assembly and use of a unified spectral simulation data base, incorporating information from previous studies.; In one study, highly accurate predictive models and simulated {dollar}sp{lcub}13{rcub}{dollar}C NMR spectra are obtained for a series of cyclohexanones and decalones, expanding the capabilities of this simulation system to include carbonyl-containing compounds. Another project focuses upon simulating spectra for alkyl-substituted aromatic molecules and involves the development of several new structural descriptors. Both of these studies examine the use of alternate atom subsetting schemes and simulated subspectra.; The second portion of this dissertation deals with the construction and use of a centralized {dollar}sp{lcub}13{rcub}{dollar}C NMR spectral simulation data base for use in evaluating and broadening the scope of this methodology. All of the structures, spectra, and predictive equations used in previously separate simulation studies are merged into one large data base. Methods for the automated selection of stored models for simulating spectra for unknown molecules are presented and evaluated. The results suggest several areas of future research, critical to the continued development of this spectral simulation system.; The work presented in this thesis demonstrates the utility of this spectral simulation methodology as a valuable means of aiding in the interpretation of {dollar}sp{lcub}13{rcub}{dollar}C NMR spectra and the identification of unknown compounds.