Simulating nonadiabatic dynamics using the Meyer-Miller-Stock- Thoss Hamiltonian: a comparison of algorithms Lauren Cook 1 , Johan E. Runeson 2 , Jeremy O. Richardson 2 and Timothy J. H. Hele 1 1 Department of Chemistry, University College London, UK, 2 Laboratory of Physical Chemistry, ETH Zürich, Switzerland Mixed quantum-classical models are commonly used to simulate nonadiabatic dynamics. Often, these approaches use a mapping to describe the electronic dynamics by time-propagating a set of classical variables, where averaging over many trajectories allows the approximation of thermal equilibrium properties through correlation functions. Here, we compare three time-propagation algorithms for the Meyer-Miller-Stock-Thoss Hamiltonian: the MInt, Split-Liouvillian (SL), and Degenerate Eigenvalue (DE) algorithms. 1–3 We determine that the MInt is the most accurate algorithm based on the symplecticity, energy conservation, computational cost, and accuracy of correlation functions. Despite not being symplectic, the SL algorithm obtains similar results for a lower computational cost and in some cases, better energy conservation. Approximations within the DE algorithm results in inaccurate dynamics, poor energy conservation and a higher computational expense for systems with weak electronic coupling. References
1. M. S. Church, T. J. H. Hele, G. S. Ezra, and N. Ananth, J. Chem. Phys. 148 , 102326 (2018). 2. J. O. Richardson, P. Meyer, M.-O. Pleinert, and M. Thoss, Chem. Phys. 482 , 124 (2017). 3. A. Kelly, R. van Zon, J. Schofield, and R. Kapral, J. Chem. Phys. 136 , 084101 (2012).
P10
© The Author(s), 2023
Made with FlippingBook Learn more on our blog