New Algorithms and their Use in the Study of Nonadiabatic Processes Influenced by Conical Intersections

点击翻译按钮获取中文摘要

David Yarkony of Johns Hopkins University is supported by an award from the Chemical Theory, Models and Computational Methods program in the Chemistry Division to develop theoretical and computational approaches to study chemical reactions that are initiated when a molecule is excited by absorbing energy from light. Scientists can describe a chemical reaction as a marble rolling on a surface: reactants point A; products point B; moving downhill is easy (extract chemical energy); moving uphill is hard (need to add energy). For many light-initiated processes such as photosynthesis, vision, and solar energy conversion, this notion must be generalized. In this case, as the result of absorbing light, the marble starts on a higher surface. Scientists want to know how the marble gets back to the lower surface and where it ends up on that surface, that is, what molecules are made and how much energy they have. If the light is provided by a laser, the task at hand may be laser control of a chemical reaction. If the light comes from the sun, energy conversion, photosynthesis or cell damage may be the issue. The passage from the upper surface to the lower surface not unlike spiraling down a drain - it involves funneling by a double cone-like structure (two cones joined at their vertices) called a "conical intersection". The mathematical description of the resulting multi-surface process requires detailed knowledge of the surfaces and the connecting cones. The more accurate the description of the surfaces and cones the more likely the computer simulation will correspond to reality. The principal goal of this project is the representation of the surfaces and connecting cones using the most accurate tools available and enabling computers to reliably simulate energetic processes. A principal issue in nonadiabatic dynamics is the quality of the electronic structure data (energies and interstate couplings) used. When direct or on-the-fly dynamics techniques are employed, the quality of electronic structure data is subordinated to the need to obtain that data quickly. The alternative is to use fitted coupled potential energy surfaces. This method can use electronic structure descriptions more accurately than those used in direct dynamics and can smooth the irregularities that may occur in the electronic energies due to the orbital changes inherent in nonadiabatic processes. The challenge is to determine the fit. By combining least squares fitting with exact data representation and incorporating derivative coupling data directly into the fitting process, an algorithm has been developed that provides a quantifiably quasi-diabatic fitted representation of electronic structure data including a reliable description of seams of conical intersections. That approach is suitable for 4-5 atom systems. The goal of this research is to extend this algorithm to treat significantly larger molecules, such as the photodissociation of phenol. The proposed fit surfaces algorithm will make high quality electronic structure data available for the study of nonadiabatic dynamics in nucleobases and a multitude of comparably sized systems. Determining how far this methodology can be extended, that is how large a system can be treated, is one of the principal goals of this research.