High-throughput screening of high energy density LiMn1-xFexPO4 via active learning

Abstract

Lithium-ion batteries (LiBs) with high energy density have gained significant popularity in smart grids and portable electronics. LiMn_1-xFe_xPO₄ (LMFP) is considered a leading candidate for the cathode, with the potential to combine the low cost of LiFePO₄ (LFP) with the high theoretical energy density of LiMnPO₄ (LMP). However, quantitative investigation of the intricate coupling between the Fe/Mn ratio and the resulting energy density is challenging due to the parametric complexity. It is crucial to develop a universal approach for the rapid construction of multi-parameter mapping. In this work, we propose an active learning-guided high-throughput workflow for quantitatively predicting the Fe/Mn ratio and the energy density mapping of LMFP. An optimal composition (LiMn_0.66Fe_0.34PO₄) was effectively screened from 81 cathode materials via only 5 samples. Model-guided electrochemical analysis revealed a nonlinear relationship between the Fe/Mn ratio and electrochemical properties, including ion mobility and impedance, elucidating the quantitative chemical composition-energy density map of LMFP. The results demonstrated the efficacy of the method in high-throughput screening of LiBs cathode materials.