Variable timestep algorithm for molecular dynamics simulation of non-equilibrium processes

Abstract

a b s t r a c tA simple, yet robust variable timestep algorithm is developed for use in molecular dynamics simulations of energetic processes. Single-particle Kepler orbits are studied to study the relationship between trajectory properties and the critical timestep for constant integration error. Over a wide variety of conditions the magnitude of the maximum force is found to correlate linearly with the inverse critical timestep. Other quantities used in the literature such as the time derivative of the force and the product of the velocity and force also show reasonable correlations, but not to the same extent. Application of the corresponding metric jjF max jjDt in molecular dynamics simulation of radiation damage in graphite shows that the scheme is both straightforward to implement and effective. In tests on a 1 keV cascade the timestep varies by over two orders of magnitude with minimal loss of energy conservation.

Dynamic Change of Virtual Time When Modeling Vehicle Operating Cycles in the Program ‘Stamm’

“…• Unlike well-known modelling systems that use adaptive [3,4,8] change of the model time step and calculate its value before performing the next step, the Stamm program changes the value of the 'Step' variable depending on a logical condition or with reference to a specific value of the current model time.…”

Modification of nanodiamonds by xenon implantation: A molecular dynamics study

“…The algorithm for the timestep automatically adjusts the timestep using the metric ||F max ||∆t as described in detail in Ref. [17]. During close approaches the timestep was reduced to values as small as 10 −5 fs, and was subsequently increased as kinetic energy was dispersed and the system equilibrated.…”

“…The simulations were performed using an in-house Molecular Dynamics package developed by one of the authors (NAM). All calculations were performed in an NVE ensemble (constant number of particles, volume and energy), using Verlet integration and a variable timestep [17]. The algorithm for the timestep automatically adjusts the timestep using the metric ||F max ||∆t as described in detail in Ref.…”