Machine Learning Approach In Exploring The Electrolyte Additives Effect On Cycling Performance Of Lini0.5mn1.5o4 Cathode And Graphite Anode-Based Lithium-Ion Cell

LiNi0.5Mn1.5O4 (LNMO), a high-voltage spinel, has attracted great attention owing to its low cost and high operating voltage. Many great efforts have been devoted to developing full-cell of LNMO/graphite because of electrolyte oxidation issues at such high voltage. In this work, the effect of additives including vinylene carbonate (VC) and lithium bis(oxalate)borate (LiBOB) in carbonate-based electrolytes was investigated on LNMO/Li and graphite/Li to find out the optimized electrolyte composition. The use of additives in 1.0 M LiPF6 in EC:DMC (1:1 v/v) seemed not to give any improvement on the electrochemical behavior of the cell. In the case of 1.2 M LiPF6 in EC:EMC (3:7 v/v), however, both half-cells of LNMO and graphite exhibited stability of discharge capacity during long cycling with the presence of LiBOB or VC additives. Based on the artificial neural network (ANN) simulation, the best electrochemical performance would obtain for LNMO/Li in the electrolyte of 0.67 M LiPF6 in EC:DMC:EMC (2.9:3.5:3.6 (v/v) with 0.41 wt% LiBOB and 1.17 wt% VC additives cell, which delivered 141.1 mAh.g(-1) and remained 90% capacity after 100 cycles when using. Also, it predicted that the graphite/Li in the electrolyte of 1.34 M LiPF6 in EC:DMC:EMC (6.2:1.2:2.6) (v/v) with a 0.08 wt% LiBOB additive would achieve 342.2 mAh.g(-1) and 85% of capacity retention after 100 cycles. The machine learning approach is efficient in exploring the effect of additive and simultaneously searching an optimization of many design parameters and thus saving the significant cost of time-consuming experiments.