Machine learning-guided construction of MoS/MoO heterostructures on hollow carbon shells for polysulfide mitigation in lithium-sulfur batteries.

The commercialization of lithium‑sulfur (LiS) batteries faces fundamental issues from polysulfide shuttling to inefficient redox kinetics, compounded by the absence of systematic methods for catalyst design. Herein, we present a machine learning (ML)-driven strategy to design MoS/MoO heterostructures anchored on nitrogen-doped hollow carbon shells (NCS) via gradient boosting decision trees modeling. The ML-guided optimization identifies critical synthesis parameters (e.g., carbonization temperature, oxidation duration) to balance adsorption capacity and catalytic activity. The resulting heterostructure exhibits superior polysulfide confinement and accelerated conversion kinetics, enabled by MoO-induced anchoring sites and NCS-accelerated electron/ion transport. The LiS batteries integrated with the MoS/MoO-NCS-modified separators deliver a remarkable initial capacity of 1002 mAh g at 1C and remain 661 mAh g after 500 cycles, with a low capacity decay rate of 0.068 %. Even at a high sulfur loading of 8.4 mg cm, the pouch cell maintains 7.25 mAh cm areal capacity with 90.1 % retention after 100 cycles. This work establishes a paradigm for ML-accelerated electrocatalyst design, addressing critical challenges in LiS batteries and advancing scalable synthesis strategies for next-generation energy storage.