Browsing by Subject "Hyperdynamics"

The purpose of this study is to analyze performance benefits of Collective Variable-driven Hyperdynamics (CVHD) method over standard molecular dynamics (MD) simulation. Theory of CVHD is an implementation of the hyperdynamics method with some beneficial features of metadynamics. The original implementation of CVHD was modified to work as an addon for COLVARS package of the LAMMPS simulation software with current software versions. About 70 simulations were run and analyzed to verify functionality of CVHD and compare results with CVHD to those without CVHD. The simulated system was the adatom self-diffusion on Copper 001 surface. It was found out that CVHD provides a significant performance boost (several order of magnitudes) over standard MD while preserving physical accuracy for simulation of the diffusion, but only in limited temperature range. With high temperatures CVHD doesn’t speed up the simulation at all compared to standard MD. With low temperatures, it is possible to achieve statistically meaningful number of diffusion events in temperatures where the same with standard MD would require unreasonable long simulations. But also, CVHD slows down at low enough temperatures so that it is impractical. It was also found out that the collective variable used in this context is suitable for counting number of adatom diffusion events, which helps analysis of adatom trajectories. It would be interesting to investigate CVHD more by trying different parametrization and by applying it to also other phenomenon than surface diffusion. The code of CVHD provides possibilities for performance optimizations, for example by utilizing parallel computation.