Gradient-Based Optimization of ReaxFF Parameters Using Pytorch for the Study of Silica Precipitation

Silica precipitation is a subject of big interest since it occurs in a wide variety of environmental and industrial processes: ceramic and catalytic applications, water heater scaling, biomineralization, coating applications to improve adhesion and wetting properties, and so forth. Even though there are many advances in atomistic simulation research of different forms of silica, the mechanism of silica precipitation has not been fully understood. We propose to study the following process using reactive force-field method (ReaxFF). Despite being a classical force field, ReaxFF can achieve quantum chemical accuracy once the optimal potential coefficients are found. However, the fitting of ReaxFF parameters is a challenge due to the complex functional form of the potential. Several techniques have been proposed to solve this problem, such as genetic algorithms and Monte-Carlo methods. The stochastic nature of these methods requires millions of error evaluations to fit the parameters, which results in excessive optimization times. Recent advances in machine learning made it possible to drastically speed up the process by utilizing the gradient of the potential. In this work, the gradient-based optimization of reactive force-field parameters is performed using Pytorch to apply the method to study silicate polymerization reactions in a water solvent. We have implemented the current ReaxFF potential as a Pytorch model. The model’s parameters were fitted to the dataset, which comprises 210K molecular clusters with forces and energies calculated using a long-range corrected hybrid functional ωB97XD3. We have validated the obtained parameters against the kinetic and thermodynamic properties of water and solid silica systems. The optimized potential will be further used to gain insights into the mechanism of silica precipitation.