Prediction and description of enantioselective separations on amylose based HPLC chiral stationary phases

Amylose based chromatographic chiral stationary phases (CSPs) are widely used throughout liquid chromatography to separate enantiomers, although little is known about the interaction processes operating on these phases.This thesis describes a systematic study into the retention and enantioselective separation capabilities of the tris (3,5-dimethylphenylcarbmate), tris (R-phenylethylcarbamate) and tris (S-phenylethylcarbamate) stationary phases.The underlying aim of this work was to obtain an understanding of the solute-stationary phase interactions which are responsible for the chromatographic retention and separation of enantiomers. By development of quantitative structure-enantioselective retention relationships (QSERR), it was possible to identify the primary components involved in the molecular interaction process and describe those interactions in terms of non-empirical solute descriptors.Results for several different series of racemic compounds indicate that multiple mechanisms are possible and that retention may be described by combinations of several key classes of interaction including hydrogen bonding, electrostatic, lipophilic, charge transfer and steric interactions. Progressive development of these properties from classical point interactions to molecular interactions, supports a proposed theory of chiral recognition which extends from the basic three-point model to a more complex dynamic molecular model. This new model, derived to explain obtained experimental results, addresses issues in chirality which are not adequately covered by previous theory. It is proposed that the original theory is a static model which does not authentically reflect real systems and as such, cannot account for all experimentally observed molecular interactions of an enantiospecific nature. A modified and improved theory of chiral recognition has been fashioned, which includes the vital aspects of conformational adjustment and molecular fit.Finally, the first reported application of artificial neural networks in chiral chromatography is documented. Multi-layer feed forward neural networks, with error back propagation, have been used in combination with QSERR as a basis to development of chiral chromatographic expert systems. Resulting networks contained cross validated predictive ability ranging from 94 to 97%