High-throughput computational discovery of multi-component Ni-rich cathodes with suppressed lattice breathing and delayed H2–H3 transition

We propose a high-throughput computational framework for the discovery of multi-element layered oxide (MELO) cathodes for lithium-ion batteries, combining machine-learning interatomic potentials with multi-objective structural screening. Starting from a Ni-rich layered LiNiO 2 -derived R 3 ¯ m framework, a chemically diverse compositional space is constructed through multi-element substitution in the transition-metal layer, together with explicit sampling of antisite defects and Li/vacancy configurations. Large-scale configurational sampling and delithiation pathway generation are performed, followed by structural relaxation and energy evaluation using a universal neural-network interatomic potential, enabling efficient exploration of more than 2400 delithiated structures across 220 MELO compounds. We introduce physically interpretable descriptors that capture key structural responses during electrochemical cycling, including c-axis evolution, volume change, and the H2–H3 phase transition. These descriptors are combined in a multi-objective Pareto screening scheme to identify compositions that simultaneously optimize structural stability and phase-transition behavior. Validation against density functional theory confirms the reliability of the proposed workflow in predicting lattice deformation and transition trends. The screening identifies a set of Pareto-optimal MELO compounds, including LiNi 0.8 Co 0.08 Mn 0.08 (AlMoZn) 0.04 O 2 , LiNi 0.8 Co 0.08 Mn 0.08 (CrMoZn) 0.04 O 2 , and LiNi 0.8 Co 0.08 Mn 0.08 (AlMoV) 0.04 O 2 , exhibiting reduced lattice breathing, moderated volume change, and delayed H2–H3 phase transition compared with conventional NCM811. Thermodynamic analysis further reveals favorable Li/vacancy mixing energetics and reduced phase-separation tendencies in the selected compounds, while precursor reaction and elemental cohesive energies indicate their potential thermodynamic accessibility. This work establishes a generalizable computational strategy for accelerated discovery of structurally robust multi-component cathode materials.