Machine Learning Speeds Up the Discovery of Efficient Porphyrinoid Electrocatalysts for Ammonia Synthesis

Abstract

Two-dimensional transition metal porphyrinoid materials (2DTMPoidMats), due to their unique electronic structure and tunable metal active sites, have the potential to enhance interactions with nitrogen molecules and promote the protonation process, making them promising electrochemical nitrogen reduction reaction (eNRR) electrocatalysts. Experimentally screening a large number of catalysts for eNRR catalytic performance would consume considerable time and economic resources. First-principles calculations and machine learning (ML) algorithms could greatly improve the efficiency of catalyst screening. Using this approach, we selected 86 candidates capable of catalyzing eNRR from 1290 types of 2DTMPoidMats, and verified the results with density functional theory (DFT) computations. Analysis of the full reaction pathway shows that MoPp-meso-F-β-Py, MoPp-β-Cl-meso-Diyne, MoPp-meso-Ethinyl, and WPp-β-Pz exhibit the best catalytic performance with the onset potential of −0.22, −0.19, −0.23, and −0.35 V, respectively. This work provides valuable insights into efficient design and screening of eNRR catalysts and promotes the application of ML algorithmic models in the field of catalysis.