Metaheuristics Computation-Assisted Multi-Element Doping Design in Electrode Materials for Rechargeable Batteries

Abstract

Low-level doping of electrode materials is known to be a common and simple method that can be used to improve electrode performance. However, multi-element doping compositions have generally been confined to the empirical intuition of researchers via trial-and-error. Here we propose a more systematic approach to designing multi-element doping compositions for electrode materials via a non-dominated sorting genetic algorithm (NSGA-II)-based computation. LiMnPO4 was selected to demonstrate our strategy not only because of its promising features such as a high operating voltage, comparable capacity, and structural stability, but also because it is known for insufficient electrochemical performance with poor reversibility and rate capability. In this study, a NSGA-II was employed to determine the optimum multi-element doping composition at the Mn site of olivine-structured LiMnPO4 through six consecutive generations, each of which contained 25 dopant sets that finally led to the pinpointing of two potential candidates for a multi-element doped LiMnPO4. Both the doped LiMnPO4/C with selective compositions delivered a reversible capacity of about 150 mA h g˗1, and also exhibited a high rate capability at the 2C rate and a decent cyclability with 91% capacity retention during 30 cycles. This approach can be widely applied to explore local optimization of multi-element substitutions for high performance electrode materials, and further it would provide a broader view for investigating other functional materials with multiple compositions that could be locally optimized.