Abstract

Electrode materials are critical components of rechargeable batteries which determine a series of device-level performances including energy density, rate capability, cyclic stability, and cost. Conventionally, the search of electrode materials relies on experimental trial-and-error and traversing existing computational databases. While these methods have led to the discovery of several commercially-viable cathodes materials, the chemical space being explored is limited and many hidden phases have been overlooked, especially those with meta-stability. We explore the possibility of using first-principles crystal structure prediction methods, in particular, ab-initio random structural searching (AIRSS), to search for novel electrode materials. We show that, by limiting the search space using a number of constraints including interatomic distances, cell volumes and spin states, AIRSS can efficiently predict both thermodynamically stable and meta-stable electrode materials. Specifically, we use LiCoO2, LiFePO4, and LixMnyOz to showcase the efficiency of the current method by re-identifying known cathodes. The effect of tunable parameters such as interatomic distances, cell volumes, and spin states on efficiency of sampling is discussed in detail. A minimal “learning” process is introduced by adapting interatomic distance constraints from low-energy random structures which efficiently captures the local coordination environment of atoms. We also use this method to make predictions. A family of cathode materials based on transition-metal oxalates is proposed which demonstrates superb energy density, oxygen-redox stability, and lithium diffusion properties.