Accurate and efficient integration for molecular dynamics simulations at constant temperature and pressure.

Area of Science:

• Computational Physics
• Materials Science
• Chemical Physics

Background:

• Molecular dynamics simulations commonly use thermostats and barostats to control temperature and pressure.
• Traditional methods update these variables at every simulation step, often on longer timescales than particle motion.

Purpose of the Study:

• To present a novel framework for molecular dynamics simulations that decouples thermostat and barostat updates from particle motion.
• To enhance simulation performance, accuracy, and software design.

Main Methods:

• Developed a framework with separate updates for barostat, thermostat, and Newtonian particle motion.
• Implemented infrequent updates for thermostat and barostat variables.
• Tested on general-purpose and special-purpose hardware.

Main Results:

• Infrequent updates significantly improve performance, especially in parallelized simulations requiring inter-processor communication.
• Reduced errors associated with limited-precision arithmetic and truncation errors.
• Achieved closer time-average pressure to the target value.

Conclusions:

• The proposed framework offers substantial performance gains and improved accuracy for molecular dynamics simulations.
• Decoupled updates simplify software complexity and enhance modularity.
• This approach is beneficial for large-scale, parallelized simulations.