Deep learning accelerated study of transition metal sulfide (TMS) catalyst for electrochemical nitrogen reduction reaction (ENRR)

• Main Author: Chen, Xiaofei
• Other Authors: Alex Yan Qingyu
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2023
• Subjects: Engineering::Materials
• Online Access: https://hdl.handle.net/10356/166820
• Institution: Nanyang Technological University

The global supply chain for food and chemicals depends heavily on the production of ammonia (NH3), but the Haber-Bosch process currently used for ammonia production is energy-intensive and has negative impacts on the environment due to carbon emissions. Electrochemical nitrogen reduction reaction (ENRR), a sustainable and cost-effective alternatives to the Haber-Bosch process, has gained increased attention in recent years. However, challenge is presented due to the need of effective catalysts for this process. In this study, the deep learning model directional message passing neural network plus plus (DimeNet++) is employed to screen various transition metal sulfide surfaces for their potential as efficient electrochemical catalysts for ammonia synthesis at ambient temperature. Using the adsorption energies of crucial intermediates calculated by the model, a mechanistic study can be conducted to evaluate catalytic activity. The study identifies several promising catalyst candidates for the production of ammonia through both the associative and dissociative mechanisms, with tungsten disulfide (WS2) showing the lowest overpotential for electrochemical formation of ammonia through the associative reaction pathway. Additionally, iron sulfide (FeS) and molybdenum disulfide (MoS2) show promise in reducing nitrogen to ammonia via the dissociative mechanism. Experimental verification of the prediction results by other researchers further supports the effectiveness of the identified catalyst candidates. Furthermore, the study investigates the effect of heteroatom doping by creating pyrite alloys. Among all the sulfides studied, iron-doped manganese sulfide (MnS2-Fe), and specifically Mn0.9S2Fe0.1, is the most promising catalyst in electrochemical ammonia production. This study demonstrates the potential of deep learning in identifying efficient catalysts for sustainable and cost-effective ENRR and offers valuable insights into transition metal sulfide catalysts for ammonia synthesis via ENRR.