ReaxFF Parameter Optimization with Monte-Carlo and Evolutionary Algorithms: Guidelines and Insights

Abstract

ReaxFF is a computationally efficient force field to simulate complex reactive dynamics in extended molecular models with diverse chemistries, if reliable force-field parameters are available for the chemistry of interest. If not, they must be optimized by minimizing the error ReaxFF makes on a relevant training set. Because this optimization is far from trivial, many methods, in particular genetic algorithms (GAs), have been developed to search for the global optimum in parameter space. Recently, two alternative parameter calibration techniques were proposed, i.e. Monte-Carlo Force Field optimizer (MCFF) and Covariance Matrix Adaptation Evolutionary Strategy (CMA-ES). In this work, CMA-ES, MCFF and a GA method (OGOLEM) are systematically 1 compared using three training sets from the literature. GA shows the smallest risk of getting trapped into a local minimum, whereas CMA-ES is capable of reaching the lowest errors for two third of the cases. For each method, we provide reasonable default settings and our analysis offers useful guidelines for their usage in future work.An important side effect impairing the parameter optimization is numerical noise. A detailed analysis reveals that it can be reduced, e.g. by using exclusively unambiguous geometry optimizations in the training set. Even without this noise, many distinct near-optimal parameter vectors can be found, which opens new avenues for improving the training set and detecting overfitting artifacts.

New Reactive Force Field for Simulations of MoS2 Crystallization

“…For the ReaxFF parameter development, we utilized a Monte-Carlo-like optimization scheme, inspired by ref ( 41 ), but not strictly following the same procedure. The training set originally included a set of experimental and hypothetical crystalline Mo x S y structures.…”

First principles study of layered silicon carbide as anode in lithium ion battery

“…It computes the fitness function with its first and second derivatives that depends on the values force field (FF) parameters [26]. The Monte Carlo force field (MCFF) global optimization method was utilized to find the best fit FF for the particular training set [27]. The ReaxFF methodology enables energy calculations of the system as given below:…”

“…It computes the fitness function with its first and second derivatives that depends on the values force field (FF) parameters [26]. The Monte Carlo force field (MCFF) global optimization method was utilized to find the best fit FF for the particular training set [27]. The ReaxFF methodology enables energy calculations of the system as given below: where, these energy distributions define bond energy, lone‐pair energy, over‐coordination penalty energy, under‐coordination stability energy, valence energy, torsion energy, Van der Waals energy and coulomb energy, respectively.…”

Mixing ReaxFF parameters for transition metal oxides using force-matching method

“…The number of parameters needed for the ReaxFF description of the potential of the system is often in the order of hundreds, what makes the initial parametrization extremely difficult. Several methods have been investigated [ 9 ], including Monte Carlo–based simulated annealing algorithm [ 10 , 11 ], genetic algorithms [ 12 – 15 ], covariance matrix adaptation [ 16 – 18 ] and machine learning [ 19 ]. However, due to the complexity of many-variable function optimization, the usual approach is to refit only some of the parameters to suit a particular system of interest.…”