DFT training data and fine-tuned MACE model for lithium storage and transport in Cu8B14 monolayers

Abstract

This dataset accompanies the manuscript “Unraveling Lithium Storage and Defect-Modulated Transport in 2D Copper Boride: From First Principles to Machine-Learned Molecular Dynamics”. The archive contains the DFT-labelled configurations used to fine-tune a MACE-MP-0 interatomic potential for Cu-B-Li systems, including the complete 1000-configuration dataset and the corresponding training, validation, and held-out test subsets. It also contains the final fine-tuned MACE model used for the reported machine-learning-interatomic-potential molecular-dynamics simulations, together with the principal training parameters. An independent 6000-frame DFT-AIMD dataset for the line-defected Cu-B structure containing 27 Li atoms is provided for model-transferability validation, together with the corresponding energy- and force-error metrics. This independent dataset was not included in the training, validation, or test subsets used during model development. Additional diffusion-analysis data, including the time-origin-averaged in-plane Li mean-squared-displacement curves, are provided in the Electronic Supplementary Information associated with the manuscript.