Suppression of shuttle effect in Li-S battery: a combined machine-learning and DFT approach for high throughput screening of cathode host materials

Abstract

The shuttle effect is a major issue in Lithium Sulfur battery that impedes practical implementation due to rapid capacity loss. The discovery of novel cathode host materials through complex experimental techniques is inefficient to find suitable cathode anchoring materials. Here, we propose a combined approach of machine learning (ML) and density functional theory (DFT) to discover appropriate sulfur host cathode materials that can effectively suppress the shuttle effect in Li-S batteries. This method aims to improve the search for suitable materials and enhance the efficiency of the discovery process. We applied a classification model to investigate the adsorption of polysulfides (Li2S, Li2S2, Li2S4, Li2S6, Li2S8, S8) on various layered double hydroxide (LDH) materials. We have found that Gradient Boosting model is suitable for predicting cathode host materials with optimum adsorption energy, while the perfectly fitted Adaboost model predicts stable cathode host materials. By combining two classification models 22 materials have been screened out through ML having high potential to be suitable sulfur host cathode materials. Finally, we cross validated with DFT and proposed 16 cathode host materials out of 74 LDH materials are highly viable for the suppression of the shuttle effect. The combined ML-DFT method delivers high-precision and quick solutions for high-throughput screening based on adsorption energy.