Machine-learning-assisted design of a binary descriptor to decipher electronic and structural effects on sulfur reduction kinetics

Abstract

The catalytic conversion of lithium polysulfides is a promising way to inhibit the shuttling effect in Li–S batteries. However, the mechanism of such catalytic systems remains unclear, which prevents the rational design of cathode catalysts. Here we propose the machine-learning-assisted design of a binary descriptor for Li-S battery performance composed of a band match (IBand) and a lattice mismatch (ILatt) indexes, which captures the electronic and structural contributions of cathode materials. Among our Ni-based catalysts, NiSe2 exhibits a moderate IBand and the smallest ILatt and is predicted and subsequently verified to improve the sulfur reduction kinetics and cycling stability, even with a high sulfur loading of 15.0 mg cm−2 or at low temperature (−20 °C). A pouch cell with NiSe2 delivers a gravimetric specific energy of 402 Wh kg−1 under high sulfur loading and lean-electrolyte operation. Such a fundamental understanding of the catalytic activity from electronic and structural aspects offers a rational viewpoint to design Li–S battery catalysts. The sluggish conversion of lithium polysulfides in Li–S batteries can be overcome by the use of catalysts, but their design is typically done via trial and error. Now, a binary descriptor is proposed by machine learning to capture electronic and structural effects for the design of Li–S battery cathode catalysts.