(180aa) Machine Learning and Prediction of NO3- Reduction Catalysts

The use of fertilizers has led to the accumulation of nitrate in ground and surface water, exceeding safe levels for humans, and in some cases causing eutrophication and dead zones in bodies of water. Electrocatalytic degradation of nitrate via Single Atom Catalysts (SAC) is a promising technology for removing nitrate from water with low energy input while closing the nitrate cycle. SAC design is challenging due to the large number of potential combinations between host and SAC metals. In this work, we use DFT to examine 8 host transition metals (Ag, Au, Cu, Ni, Pd, Pt, Ti, Y) and 12 SAC metals (host metals plus Fe, Mo, In, W), spanning the breadth of transition metals and with a wide range of D-band properties, for their performance in nitrate reduction to N₂ and NH₃. In addition, unsupervised machine learning is used to examine the relationship between the key reaction steps with easily accessible d-band properties of the SAC/host neat surface (d-band center, filling, kurtosis, skewness, and fermi energy). Key reaction energies are estimated to within ~0.3 eV by this method, demonstrating that key steps may be predicted, and thermodynamically inactive surfaces may be ruled out at a low cost before close examination. Additionally, we propose design principles for nitrate reduction SACs and make recommendations of SACs for further study.