Dataset of Random Relaxations for Crystal Structure Search of Li-Si System
This dataset and method address the challenge of crystal structure search for materials scientists, offering a more efficient way to identify stable structures, which is an incremental improvement in materials design.
This paper introduces a dataset of over 100,000 density functional theory (DFT) structural relaxations for Li-Si battery anode materials, starting from randomized structures. Using this dataset, the authors trained graph neural networks that can predict structural relaxations, including both ionic positions and lattice vectors, achieving up to two orders of magnitude lower error compared to models trained on molecular dynamics simulations for this specific task.
Crystal structure search is a long-standing challenge in materials design. We present a dataset of more than 100,000 structural relaxations of potential battery anode materials from randomized structures using density functional theory calculations. We illustrate the usage of the dataset by training graph neural networks to predict structural relaxations from randomly generated structures. Our models directly predict stresses in addition to forces, which allows them to accurately simulate relaxations of both ionic positions and lattice vectors. We show that models trained on the molecular dynamics simulations fail to simulate relaxations from random structures, while training on our data leads to up to two orders of magnitude decrease in error for the same task. Our model is able to find an experimentally verified structure of a stoichiometry held out from training. We find that randomly perturbing atomic positions during training improves both the accuracy and out of domain generalization of the models.