DFT Accurate Interatomic Potential for Molten NaCl from Machine Learning

Abstract

Molten alkali chloride salts are a critical component in concentrated solar power and nuclear applications. Despite their ubiquity, the extreme chemical reactivity of molten alkali chlorides at high temperatures has presented a significant challenge in characterizing atomic structures and dynamic properties experimentally. Here, we investigate molten NaCl by performing high-temperature molecular dynamics simulations using a Gaussian approximation potential (GAP) trained on density functional theory (DFT) data sets. Our GAP model, trained with 1000 atomic configurations, arrives near DFT accuracy with a mean absolute error of 1.5 meV/atom, thus enabling fast analysis of high-temperature salt properties at large length (5000 ion pairs) and time (>1 ns) scales, currently inaccessible to ab initio simulations. Calculated structure factors and diffusion constants from our GAP model simulations show excellent agreement with experiments. Our results indicate that GAP models are able to capture the many-body interactions required to accurately model ionic systems.

Modelling Bulk Electrolytes and Electrolyte Interfaces with Atomistic Machine Learning

“…Along the same line, Tovey et al . had used SOAP descriptors with GPR to generate a Gaussian approximation potential for molten NaCl with a RMSE of energy as 1.5 meV/atom [133] . They showed that the computed self‐diffusion coefficients of Na + and Cl – ions from MD simulations using this potential agree well with experimental results.…”

Automated Development of Molten Salt Machine Learning Potentials: Application to LiCl

“…where V is the volume, j(t) = Σ N i=1 q i • v i (t) is the current with the q i being the ionic charge and v i (t) velocity at time t. The benchmark of the code implementation is discussed elsewhere. 21…”

“…18,20 Recently, we showed that many-body polarization in NaCl melts can be captured accurately by using MLP simulations scaled to ∼10,000 atoms with near Density Functional Theory (DFT) accuracy. 21 At these spatiotemporal scales, structural heterogeneity and ensemble averaging can be incorporated meaningfully into calculations, combining the accuracy of DFT with the statistical sampling of classical interatomic potentials. Despite these successes, the application of MLPs to the broader chemical space of molten salts is challenged by the difficulty of MLP parametrization.…”

On the role of long-range electrostatics in machine-learned interatomic potentials for complex battery materials

“…38 Consequently, short-ranged ML potentials are now commonly applied to study polar and even ionic systems. 34,[39][40][41] However, the importance of long-range effects will clearly depend on the material and property of interest and thus demands more systematic scrutiny. 42 For example, ionic diffusion in electrolytes may lead to the transient local accumulation and depletion of charges, which can break the isotropy of the electrostatic environment.…”