pyMSER─An Open-Source Library for Automatic Equilibration Detection in Molecular Simulations.

Area of Science:

• Computational Materials Science
• Chemical Engineering
• Statistical Mechanics

Background:

• Automated molecular simulations are crucial for predicting material properties.
• Simulations typically involve dynamic equilibration followed by a steady state.
• Accurate property extraction requires reaching a steady state, but predicting this point a priori is challenging due to condition-dependent equilibration.

Purpose of the Study:

• To demonstrate the application of the Marginal Standard Error Rule (MSER) for automatically identifying the optimal truncation point in simulations.
• To ensure objective, accurate, and reproducible extraction of figures of merit from simulation data.
• To reduce computational cost in automated molecular simulations.

Main Methods:

• Application of the Marginal Standard Error Rule (MSER) to identify steady-state in simulations.
• Utilized Grand Canonical Monte Carlo (GCMC) simulations for gas adsorption in metal-organic frameworks as a case study.
• Developed an open-source Python implementation, pyMSER, for broader applicability.

Main Results:

• MSER successfully and automatically identifies the optimal point to truncate simulations, ensuring steady-state has been reached.
• In GCMC simulations of gas adsorption, MSER reduced computational cost by up to 90%.
• MSER statistics are independent of the simulation method, indicating wide applicability to time-series data requiring equilibration truncation.

Conclusions:

• MSER provides a novel, automatic procedure for determining simulation steady-state, enhancing objectivity and reproducibility.
• This methodology offers significant computational savings, particularly in complex simulations like gas adsorption.
• The pyMSER library is available for use in any field requiring time-series analysis with equilibration truncation.