Random-batch list algorithm for short-range molecular dynamics simulations.

Area of Science:

• Computational Physics
• Materials Science
• Chemical Physics

Background:

• Molecular dynamics (MD) simulations are crucial for understanding material properties at the atomic level.
• Calculating short-range interactions efficiently is a key challenge in large-scale MD simulations.
• Classical neighbor-list methods, like Verlet lists, can be computationally expensive to construct.

Purpose of the Study:

• To develop a novel, fast method for calculating short-range interactions in MD simulations.
• To reduce the computational cost associated with neighbor searching and interaction calculations.
• To maintain or improve the accuracy of MD simulations while increasing speed.

Main Methods:

• Introduced the random-batch list method, a stochastic alternative to traditional neighbor lists.
• Implemented a two-level neighbor list system with core and shell regions for particles.
• Utilized random batches of interacting particles in the shell region to decrease pair computations.
• Provided an error estimation for the developed algorithm.

Main Results:

• Demonstrated significant acceleration (several-fold) of MD simulations for Lennard-Jones fluids.
• Showcased that the random-batch list method achieves this speedup without loss of accuracy.
• Confirmed the method's simplicity of implementation and compatibility with linked-cell techniques.

Conclusions:

• The random-batch list method offers a substantial speedup for large-scale molecular dynamics simulations.
• This approach is versatile, easily integrated with other methods, and extensible to various interaction types.
• The method holds promise for advancing the efficiency and scalability of computational materials science and physics research.