Bilayer MN4-O-MN4 by bridge-bonded oxygen ligands: Machine learning to accelerate the design of bifunctional electrocatalysts

We designed and screened bifunctional catalysts with good oxygen reduction reaction (ORR) and oxygen evolution reaction (OER) performance on bilayer MN4-O-MN4 structures with bridge-bonded oxygen ligands. The ORR and OER catalytic activities of 225 bilayer MN4-O-MN4 structures were explored in an accelerated manner by combining machine learning (ML) and density functional theory (DFT) calculations (DFT-ML). Based on the gradient boosted regression (GBR) algorithm, a series of efficient monofunctional and bifunctional electrocatalysts were successfully predicted with an average prediction error of only 0.04 V and 0.06 V for ORR and OER overpotential (η). ML successfully predicted that the overpotential of the monofunctional catalysts CoN4–O–RhN4 (ORR) and RhN4–O–AgN4 (OER) reached 0.34 V and 0.29 V, respectively; CoN4–O–AgN4 was considered the best bifunctional catalyst due to its overpotential of ηORR = 0.35 V and ηOER = 0.33 V on the bifunctional catalysts. Compared with DFT calculations, the DFT-ML accelerated calculation method resulted in a 9.4-fold improvement in catalyst screening speed. The performance prediction of 225 bilayer MN4-O-MN4 structures was used to screen out the potential bifunctional catalysts, thus providing guidance for the experimental synthesis of better performing bridge-bonded oxygen ligand catalysts.