Search

Artificial intelligence (AI) has emerged as a tool for discovering and optimizing novel battery materials. However, the adoption of AI in battery cathode representation and discovery is still limited due to the complexity of optimizing multiple performance properties and the scarcity of high-fidelity data. We present a machine learning model (DRXNet) for battery informatics and demonstrate the application in the discovery and optimization of disordered rocksalt (DRX) cathode materials. We have compiled the electrochemistry data of DRX cathodes over the past 5 years, resulting in a dataset of more than 19,000 discharge voltage profiles on diverse chemistries spanning 14 different metal species. Learning from this extensive dataset, our DRXNet model can capture critical features in the cycling curves of DRX cathodes under various conditions. Our approach offers a data-driven solution to facilitate the rapid identification of novel cathode materials, accelerating the development of next-generation batteries for carbon neutralization.

Disordered rocksalt oxide (DRX) cathodes are promising candidates for next-generation Co- and Ni-free Li-ion batteries. While fluorine substitution for oxygen has been explored as an avenue to enhance their performance, the amount of fluorine incorporated into the DRX structure is particularly challenging to quantify and impedes our ability to relate fluorination to electrochemical performance. Herein, an experimental-computational method combining 7Li and 19F solid-state nuclear magnetic resonance, and ab initio cluster expansion Monte Carlo simulations, is developed to determine the composition of DRX oxyfluorides. Using this method, the synthesis of Mn- and Ti-containing DRX via standard high temperature sintering and microwave heating is optimized. Further, the upper fluorination limit attainable using each of these two synthesis routes is established for various Mn-rich DRX compounds. A comparison of their electrochemical performance reveals that the capacity and capacity retention mostly depend on the Mn content, while fluorination plays a secondary role.

Machine learning interatomic potentials (MLIPs) have introduced a new paradigm for atomic simulations. Recent advancements have seen the emergence of universal MLIPs (uMLIPs) that are pre-trained on diverse materials datasets, providing opportunities for both ready-to-use universal force fields and robust foundations for downstream machine learning refinements. 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 under-prediction in a series of atomic-modeling benchmarks including surfaces, defects, solid-solution energetics, phonon vibration modes, ion migration barriers, and general high-energy states. We find that the PES softening behavior originates from a systematic underprediction error of the PES curvature, which derives from the biased sampling of near-equilibrium atomic arrangements in uMLIP pre-training datasets. We demonstrate that the PES softening issue can be effectively rectified by fine-tuning with a single additional data point. Our findings suggest that a considerable fraction of uMLIP errors are highly systematic, and can therefore be efficiently corrected. This result rationalizes the data-efficient fine-tuning performance boost commonly observed with foundational MLIPs. We argue for the importance of a comprehensive materials dataset with improved PES sampling for next-generation foundational MLIPs.

The simulation of large-scale systems with complex electron interactions remains one of the greatest challenges for the atomistic modeling of materials. Although classical force fields often fail to describe the coupling between electronic states and ionic rearrangements, the more accurate \textit{ab-initio} molecular dynamics suffers from computational complexity that prevents long-time and large-scale simulations, which are essential to study many technologically relevant phenomena, such as reactions, ion migrations, phase transformations, and degradation. In this work, we present the Crystal Hamiltonian Graph neural Network (CHGNet) as a novel machine-learning interatomic potential (MLIP), using a graph-neural-network-based force field to model a universal potential energy surface. CHGNet is pretrained on the energies, forces, stresses, and magnetic moments from the Materials Project Trajectory Dataset, which consists of over 10 years of density functional theory static and relaxation trajectories of $\sim 1.5$ million inorganic structures. The explicit inclusion of magnetic moments enables CHGNet to learn and accurately represent the orbital occupancy of electrons, enhancing its capability to describe both atomic and electronic degrees of freedom. We demonstrate several applications of CHGNet in solid-state materials, including charge-informed molecular dynamics in Li$_x$MnO$_2$, the finite temperature phase diagram for Li$_x$FePO$_4$ and Li diffusion in garnet conductors. We critically analyze the significance of including charge information for capturing appropriate chemistry, and we provide new insights into ionic systems with additional electronic degrees of freedom that can not be observed by previous MLIPs.

The cluster expansion (CE) method has seen continuous and increasing use in the study of configuration-dependent properties of crystalline materials. The original development of the CE method along with the underlying mathematical formalism and assumptions was focused on the study of metallic alloys. Since then the methodology has been actively and successfully used in the study of ionic materials as well. In this work, we present a cohesive reformulation of the mathematical formalism underlying the CE method based on a synthesis of its original formulation, several additions and extensions that have been proposed since, and a revised representation of its constituent mathematical objects. We then proceed to describe some of the formal implications of using the methodology for charge-neutral configurations in ionic systems. In particular, we discuss the reduction of the size of configuration spaces and the resulting linear dependencies that arise among correlation functions that span the larger unconstrained configuration space. Additionally, we explore the effects of long-range electrostatic interactions. We also demonstrate how the previously proposed use of a point electrostatic term successfully accounts for the majority of the longer-range electrostatic interactions, and leaves the cluster expansion terms to capture mostly short-range interactions. Finally, we present and discuss a variety of recently developed methodologies, including training structure selection, oxidation state assignment, structure mapping, and regression algorithms, that are necessary to address these formal mathematical notions for a practical implementation of the CE method in the study of multicomponent ionic materials.