As a zero-emission, eco-friendly fuel, hydrogen gas can be generated via electrochemical (EC) water splitting. Achieving high-efficiency water splitting requires the use of a catalyst to minimize the overpotential to drive the hydrogen evolution reaction (HER). Noble metals such as platinum (Pt) can provide an excellent catalytic activity for HER but are too expensive and scarce for broad applications. Therefore, the development of active HER catalysts made from low-cost materials constitutes a crucial challenge in the utilization of hydrogen energy.

Earth-abundant transition metal dichalcogenides (TMDs), such as molybdenum disulfide (MoS₂), have been discovered recently, which exhibit good activity and stability for electrocatalytic reactions. In order to fully explore the untapped potential of MoS₂, the synthesis recipe for MoS₂needs to be optimized. Such an optimization process needs scientists to search through a combinatorically large space of experimental parameters, which will be time-consuming and costly if using conventional trial-and-error approaches.

In this report, MoS₂ HER catalytic activity optimization is performed by examining different combinations of synthesis parameters during the hydrothermal process. To investigate the structure-activity relationship, scanning electron microscope (SEM), X-ray diffraction (XRD), Raman spectroscopy and various electrochemical characterizations have been conducted. A strong correlation between hydrothermal conditions and HER performance matrix has been observed. In order toaccelerate the search for the best synthesis recipe, machine-learning (ML) techniques have been introduced to help identify the optimal parameter combinations for producing MoS₂. The hydrothermal parameters with the corresponding onset potentials and Tafel slopes are adopted as prior knowledge and are incorporated into the Bayesian Optimization model. The model will be able to guide the wet chemical synthesis of MoS₂and yield the most effective HER catalyst eventually.