NERSC Initiative for Scientific Exploration (NISE) 2011 Awards

Semiclassical Approaches for Clean Energy Resources

Jian Liu, University of California Berkeley

Associated NERSC Project: Semiclassical approaches for condensed matter dynamics and for molecular reaction dynamics (mp14)
Principal Investigator: William Miller, University of California Berkeley

The research project described in this proposal is mainly aimed at developing new semiclassical methods and models to properly include quantum effects into classical molecular dynamics simulations of dense plasmas (e.g., fusion plasmas) and complex molecular systems. The research will also focus on the applications of these semiclassical approaches to specific problems of clean energy interests and water-related technologies. The accomplishment of this project will greatly advance our understandings of fundamental mechanisms of quantum effects in fusion plasmas and photochemical dynamics related to solar energy conversion; therefore will fulfill the mission of the Department of Energy.

We have made a significant advance in the development of more efficient semiclassical methodology that is capable of describing true quantum coherence and decoherence in the dynamics of large molecular systems. Semiclassical initial value representation (SC-IVR) methods take use of classical trajectories to describe quantum dynamics. A simple model is presented for treating local imaginary frequencies that are important in the study of quantum effects in chemical reactions and various dynamical processes in molecular liquids. It significantly extends the range of accuracy of conventional local harmonic approximations used in the linearized semiclassical initial value representation (LSC-IVR)/classical Wigner approximation for real time correlation functions. The key idea is realizing that a local Gaussian approximation (LGA) for the momentum distribution (from the Wigner function involving the Boltzmann operator) can be a good approximation even when a local harmonic approximation (LHA) for the potential energy surface fails. Along these lines, we have recently proposed three novel approaches for generating trajectory-based dynamics which conserves canonical distribution in the phase space formulation of quantum mechanics. They provide the framework to go beyond traditional quantum statistical potential approach to capture quantum effects in real time.

Applications of semiclassical methodologies to fusion plasmas will provide deeper insights on our understanding the dynamical properties of fusion plasmas. Currently we are collaborating with Lawrence Livermore National Laboratory to apply the semiclassical approaches to study energy fluxes in fusion plasmas to help us understand how energy is transferred between different modes and transported in the macro scope. We are also developing algorithms to improve the efficiency and scalability of the methodologies. For instance, when dealing with systems with a large number of particles on massively parallel computers, the scalability of the short- and long-range force terms is very different. Typically, 1,000 ~ 10,000 particles will be the standard size of the system that we are studying. We will divide the processes into two subsets.

The important role of liquid water in many areas of science from chemistry, physics, biology, geology to climate research, etc., has motivated numerous theoretical studies of its structure and dynamics. The significance of quantum effects on the properties of water, however, has not yet been fully resolved. In this paper we focus on quantum dynamical effects in liquid water based on the linearized semiclassical initial value representation (LSC-IVR) with a quantum version of the simple point charge/flexible (q-SPC/fw) model for the potential energy function. The infrared (IR) absorption spectrum and the translational diffusion constants have been obtained from the corresponding thermal correlation functions, and the effects of inter-molecular and intra-molecular correlations have been studied. Currently we are applying LSC-IVR to study liquid water using an ab-initio based, flexible, and polarizable force field. The system size consists of hundreds of molecules (or thousands of atoms). The computation also requires massively parallel computers.