| dc.description.abstract | Statistical rate theories such as transition state theory (TST) are central to the understanding of the rates of chemical reactions. This dissertation describes an investigation of a variety of observations in simple organic reactions that cannot be understood with statistical rate theories. Such observations are referred to as dynamic effects.
The long history of experimental observations in carbocation reactions has treated ion pairs as ordinary intermediates reacting statistically. The combined experimental, computational, and dynamic trajectory study of one such reaction here, suggests that this view is incorrect due to the importance of nonequilibrium solvation and solvent dynamics in the mechanism. This idea complicates our understanding of the nature of charged intermediates in solution and suggests a reinterpretation of a series of historically notable observations in carbocation reactions.
The fundamental idea of symmetry in chemistry breaks down on a sufficiently short time scale, but this theoretical pedantry is assumed to be irrelevant to reactive chemistry. However, for the carboborative ring contraction of cyclohexenes studied here, the apparent symmetry of an intermediate structure is lost due to its short lifetime. This reaction affords an unequal mixture of two equivalent products, and molecular dynamics studies are able to account for the product mixture based on the formation of a nominally symmetrical carbocation structure with a nonsymmetrical distribution of vibrational energy. This dynamical asymmetry complicates the understanding and interpretation of stereochemical observations in chemistry.
Machine-learning transition state theory (MLTST) is an approach to the numerical prediction of dynamic effects and their qualitative understanding, while retaining the strengths of statistical rate theories. The central idea of MLTST is that machine learning algorithms are used to divide the transition state hypersurface into regions associated with each possible outcome of a trajectory crossing the transition state. This mapping of the surface allows the prediction of trajectory outcomes from the first point of the trajectory, sidestepping the numerous and computationally expensive trajectories needed to achieve statistical significance. This approach allows for an understanding of selectivity controlling elements outside of the TST paradigm. | en |
