pyMSERAn Open-Source Library for Automatic\nEquilibration Detection in Molecular Simulations

Automated\nmolecular simulations are used extensively for predicting\nmaterial properties. Typically, these simulations exhibit two regimes:\na dynamic equilibration part, followed by a steady state. For extracting\nobservable properties, the simulations must first reach a steady state\nso that thermodynamic averages can be taken. However, as equilibration\ndepends on simulation conditions, predicting the optimal number of\nsimulation steps <i>a priori</i> is impossible. Here, we\ndemonstrate the application of the Marginal Standard Error Rule (MSER)\nfor automatically identifying the optimal truncation point in Grand\nCanonical Monte Carlo (GCMC) simulations. This novel automatic procedure\ndetermines the point at which a steady state is reached, ensuring\nthat figures of merit are extracted in an objective, accurate, and\nreproducible fashion. In the case of GCMC simulations of gas adsorption\nin metal–organic frameworks, we find that this methodology\nreduces the computational cost by up to 90%. As MSER statistics are\nindependent of the simulation method that creates the data, this library\nis, in principle, applicable to any time series analysis in which\nequilibration truncation is required. The open-source Python implementation\nof our method, pyMSER, is publicly available\nfor reuse and validation at https://github.com/IBM/pymser.