Understanding the CO2/CH4/N2 Separation Performance of Nanoporous Amorphous N-Doped Carbon Combined Hybrid Monte Carlo with Machine Learning

Amorphous carbon (aC) is widely used as the adsorbent in the purification of industrial gas. Introducing nitrogen dopant can regulate the morphology and improve the adsorption capacity of specific species. Due to the amorphous structure, it is difficult to understand the relationship between structural features and adsorption performance through atom-based simulation. Here, a series of nitrogen-doped amorphous carbon (N-aC) models is built through reverse Monte Carlo method. The uptakes of three common gases, i.e., CO2, CH4, and N-2 are estimated in each constructed framework by using grand canonical Monte Carlo (GCMC). Deep neural network is trained based on the simulated adsorption capacity with nitrogen content, surface area, pore size, atomic charge, and other factors. Through the data-driven approaches, the adsorption capacity and the selectivity of three gases are predicted. The simulation in this study shows that the nitrogen content has less influence on the capacity and selectivity than the structural parameters, while nitrogen doping may improve CO2 loading and separation selectivity in the nanopores with pore size close to gas molecules. This work is helpful in constructing amorphous carbon structures for further simulation and understanding the influence of various features on gas separation.