A script for generating Nitrogen-Doped structures by changing concentration of Nitrogen in the structure.
Abstract: "The oxygen reduction reaction (ORR) is of great importance because it is the cathodic reaction of a fuel cell. It is a four-electron-transfer process that generates a high energy barrier and limits the reaction efficiency. Although Platinum (Pt) is well known as the most efficient catalyst for ORR, its high cost and scarcity has limited its wide application. In addition, platinum-based catalysts suffer from problems such as time-dependent drift, CO deactivation, etc. A number of more abundant and cheaper materials have been explored to replace platinum. Among them, nitrogen (N) doped carbons have been demonstrated to be promising materials for ORR owing to its low cost, high abundance, and strong resistance to the poisoning of CO. However, the activity of N doped carbon still underperforms that of Pt. In this research, we studied the effect of N doping concentration and surface functionalization structure on the activity of N doped carbon. Machine learning approach proved to be useful to analyze huge amount of possibilities of N doping concentrations and surface functionalization structures. Moreover, machine learning approach reduces the costs required to perform first-principles DFT (Density Functional Theory) calculations by selectively performing calcuations on only promising structures. Different algorithms such as neural network, gradient-boosted regression are tested. By combining machine learning with first-principles DFT calculations, we aim to better understand how tuning N doping concentration and surface functionalization could help increase the activity of N doped carbon."