Parallel algorithms for hyperdynamics and local hyperdynamics.

Area of Science:

• Computational Physics
• Materials Science
• Chemistry

Background:

• Standard molecular dynamics (MD) simulations are limited by slow timescales, especially for systems with infrequent transitions between energy states.
• Hyperdynamics (HD) offers a method to accelerate MD by enabling faster escapes from energy basins while ensuring statistical accuracy.
• Local HD was developed to improve the scalability of HD for larger systems by exploiting intrinsic locality.

Purpose of the Study:

• To formulate parallel algorithms for both standard Hyperdynamics (HD) and local HD.
• To implement these parallel HD methods within the LAMMPS molecular dynamics code.
• To enable efficient, large-scale simulations utilizing HD's time acceleration capabilities.

Main Methods:

• Development and implementation of parallel algorithms for Hyperdynamics (HD) and local HD.
• Integration of these parallel methods into the LAMMPS molecular dynamics simulation package.
• Execution of large-scale (million-atom) simulations to demonstrate the methods' efficacy.

Main Results:

• Successfully implemented parallel HD and local HD in LAMMPS, compatible with all LAMMPS interatomic potentials.
• Achieved significant time acceleration (orders of magnitude) for simulations at a modest computational cost (2-4x standard MD).
• Demonstrated application through 80 μs and 160 μs simulations of Pt adatom diffusion and clustering on Pt(100).

Conclusions:

• Parallelized HD and local HD provide an efficient means to accelerate large-scale molecular dynamics simulations.
• The LAMMPS implementation makes these powerful simulation techniques accessible for a wide range of materials science and chemistry problems.
• These methods enable the study of rare events and long-timescale phenomena previously inaccessible with standard MD.