Search

Machine learning interatomic potentials (MLIPs) have introduced a new paradigm for atomic simulations. Recent advancements have led to universal MLIPs (uMLIPs) that are pre-trained on diverse datasets, providing opportunities for universal force fields and foundational machine learning models. However, their performance in extrapolating to out-of-distribution complex atomic environments remains unclear. In this study, we highlight a consistent potential energy surface (PES) softening effect in three uMLIPs: M3GNet, CHGNet, and MACE-MP-0, which is characterized by energy and force underprediction in atomic-modeling benchmarks including surfaces, defects, solid-solution energetics, ion migration barriers, phonon vibration modes, and general high-energy states. The PES softening behavior originates primarily from the systematically underpredicted PES curvature, which derives from the biased sampling of near-equilibrium atomic arrangements in uMLIP pre-training datasets. Our findings suggest that a considerable fraction of uMLIP errors are highly systematic, and can therefore be efficiently corrected. We argue for the importance of a comprehensive materials dataset with improved PES sampling for next-generation foundational MLIPs. [ABSTRACT FROM AUTHOR]

Li‐excess disordered rocksalts (DRXs) are emerging as promising cathode materials for Li‐ion batteries due to their ability to use earth‐abundant transition metals. In this work, a new strategy based on partial Li deficiency engineering is introduced to optimize the overall electrochemical performance of DRX cathodes. Specifically, by using Mn‐based DRX as a proof‐of‐concept, it is demonstrated that the introduction of cation vacancies during synthesis (e.g., Li1.3‐xMn2+0.4‐xMn3+xNb0.3O1.6F0.4, x = 0, 0.2, and 0.4) improves both the discharge capacity and rate performance due to the more favored short‐range order in the presence of Mn3+. Density functional theory calculations and Monte Carlo simulations, in combination with spectroscopic tools, reveal that introducing 10% vacancies (Li1.1Mn2+0.2Mn3+0.2Nb0.3O1.6F0.4) enables both Mn2+/Mn3+ redox and excellent Li percolation. However, a more aggressive vacancy doping (e.g., 20% vacancies in Li0.9Mn3+0.4Nb0.3O1.6F0.4) impairs performance because it induces phase separation between an Mn‐rich and a Li‐rich phase. [ABSTRACT FROM AUTHOR]