Direct Dynamics with Applications to Photochemical Reactivity

Barry Robert Smith
January 1998
Department of Chemistry
King's College London

Abstract

Traditional techniques for computing classical trajectories are problematic when dealing with many degrees of freedom. One might usually expect to map-out the entire potential energy surface (PES) before starting the calculation. Direct dynamics is a method which overcomes this problem by removing the need to compute the PES explicitly.

When studying photochemical systems one encounters a fundamental problem: the detailed bond-breaking / bond-making processes which occur in these reactions must be modelled using some form of quantum-mechanical potential surface. An efficient solution is to employ a hybrid potential, which utilises a combination of quantum and classical theory to generate the electronic structure data required for a trajectory.

This thesis describes a method which combines direct dynamics with a hybrid potential -- MM-VB (molecular mechanics with valence bond) -- making it possible to study the dynamics of quite large organic photochemical systems without neglecting any degrees of freedom at all. Conical intersections (surface crossings) play a fundamental role in photochemical mechanisms. To simulate transfer through an intersection, a trajectory-surface-hopping (TSH) algorithm is used.

Our aim is to understand the factors which control photoproduct distribution, and to examine general dynamic effects caused by the reaction path and surface topology. The first four reactions we have studied involve the photochemistry of azulene, benzene, cyclohexadiene/hexatriene and all-trans octatetraene, respectively. Full dynamics involving all degrees of freedom have never before been computed for any of these systems. The fifth application chapter documents some interesting soliton-type dynamics which we predict can occur in linear polyenes, via a conical intersection mechanism. The final project, a host-guest study using a cyclodextrin cavity, demonstrates the effectiveness of MM-VB direct dynamics for studying large systems.