"Swarm relaxation": Equilibrating a large ensemble of computer simulations^⋆

Area of Science:

• Computational physics
• Statistical mechanics
• Molecular modeling

Background:

• Traditional simulations use multiple, independently initialized runs to enhance microstate sampling.
• Optimizing sampling efficiency is crucial for accurate equilibrium property evaluation.

Purpose of the Study:

• To investigate the effectiveness of using a large ensemble of short, independently initialized simulations (a "swarm") for improved sampling.
• To determine the conditions under which this swarm approach can rapidly achieve and confirm equilibrium.

Main Methods:

• Employing a large ensemble (swarm) of M independent simulations, each relaxed to equilibrium.
• Monitoring the swarm's relaxation process to equilibrium and verifying its attainment.
• Analyzing the ensemble of final microstates for equilibrium property evaluation.

Main Results:

• A swarm size of order N allows monitoring and confirmation of equilibrium within O(tau) time, where tau is the equilibrium relaxation time.
• The final microstates from an equilibrated swarm are sufficient for most equilibrium property calculations.
• This method reduces wall-clock time by several hundred-fold compared to single long simulations, with only a modest increase in total computational effort.

Conclusions:

• The swarm approach offers a significant acceleration of simulation studies, reducing wall-clock time dramatically.
• This strategy is highly parallelizable and suitable for modern high-throughput computing systems.
• It provides a viable method for obtaining high-precision simulation results in minimal time.