(437a) Discovery of Catalysts for the Application of Nitrate Reduction and Water Purification Using Machine Learning Techniques

Nitrate (NO₃) runoff has become a leading pollutant to nearby water supplies in areas with intense agricultural and industrial processes. Redox reactions facilitated by ion exchange is currently the dominant means of converting NO₃ into more benign components such as N₂. However, such processes require the continuous replenishment of ions and produces brine as a by-product that must be subsequently removed. Electrocatalysis can provide an alternative means of reducing NO₃ to N₂ without requiring the continuous regeneration of the material. Furthermore, the absence of ions and other chemicals provides a brine-free product. However, the majority of known electrocatalysts that can select N₂ as a by-product is limited to costly precious metals and the overpotentials required to facilitate these reactions are often very high. The discovery of more cost- and energy-efficient electrocatalysts is essential to making NO₃ reduction with electrocatalysts viable. In this abstract, we screened for potential cost efficient electrocatalysts by querying the Materials Project and Aflow databases for bimetallic alloys. We predicted the oxygen and nitrogen adsorption energies for these alloys using a Graph Neural Network machine learning model previously developed by the Open Catalyst Project (OC20). Using BEP scaling relationships and microkinetic models developed by Singh and Goldsmith group, we estimated the turnover frequency for NO₃ consumption with the predicted adsorption energies. We then performed DFT calculations facilitated by a high-throughput framework on alloy surfaces with an estimated turnover frequency above 0.1/s to accurately assess the selectivity, overpotential and turnover frequencies of each candidate alloy in order to propose a novel set of materials for NO₃ reduction reaction.