A Critical Assessment of Neural Network Potentials for Water and the Role of Nuclear Quantum Effects Through the Van Hove Correlation Function

Recent advances in aqueous electrolytes for energy storage devices like supercapacitors and Li-ion batteries have further motivated improving our understanding of the transport properties of water. High-accuracy studies of the properties of liquid water should combine the accuracies achievable by <i>ab initio</i> molecular dynamics (AIMD) with the long time and large length scales achievable by classical MD. To elucidate the correlation between molecular interactions and the macroscopic transport properties of water, we need to examine the spatiotemporal molecular correlations like the Van Hove correlation function (VHF).<br/>Here, we assess two different neural network potential (NNP) based modeling strategies within classical MD in predicting the VHF. In principle, the NNP’s can deliver <i>ab initio</i> quality results at significantly larger time and length scales than would be directly accessible using <i>ab initio</i> methods. We apply the DeepMD [1] and NequIP [2] approaches and critically assess their efficacy, particularly regarding the size of training set and accuracy of the final predictions. Whereas DeepMD uses a fully-connected feedforward NN’s, NequIP employs SE(3) equivariant graph convolutional NN’s to account for the interactions of geometric tensors, resulting in a more reliable representation of the chemical environment. By varying the training sets and using path integral approaches, we analyze the role of nuclear quantum effect on the VHF. These results are directly contrasted with recent inelastic X-ray scattering data [3].<br/>This research was sponsored by the Fluid Interface Reactions, Structures, and Transport (FIRST) Center, an Energy Frontier Research Center funded by the U.S. Department of Energy (DOE), Office of Science, Office of Basic Energy Sciences.<br/>[1] Wang, Han, et al. "DeePMD-kit: A deep learning package for many-body potential energy representation and molecular dynamics." <i>Computer Physics Communications</i> 228 (2018): 178-184.<br/>[2] Batzner, Simon, et al. "SE (3)-equivariant graph neural networks for data-efficient and accurate interatomic potentials." <i>arXiv preprint arXiv:2101.03164</i> (2021).<br/>[3] Iwashita, Takuya, et al. "Seeing real-space dynamics of liquid water through inelastic x-ray scattering." <i>Science advances</i> 3.12 (2017): e1603079.