UC Riverside

Membrane-based materials are an important branch in the field of gas separation. There are increasing number of membrane materials which are synthesized in recent years. Among them, the novel ultrathin membranes with uniform pores like nanoporous graphene attracted attention of many researchers, because their excellent gas separation ability based on molecular sieving effect. Several factors might influence the performance of ultrathin membranes, including pore size, pore shape, pore density, etc. However, the theoretical understanding of these factors has not been clearly addressed. In this dissertation, we firstly investigated the optimal pore size by first-principles density functional theory and simulated the pore-size effect for post-combustion carbon capture by grand canonical Monte Carlo. Next, we used molecular dynamics simulations to study the effect of pore density on the nanoporous graphene membrane. Then, we proposed a bilayer design of nanoporous graphene membrane with continuously tunable effective pore size for gas separations, such as CO2/CH4, N2/CH4, O2/N2. Meanwhile, we studied the effect of entropic selectivity by tracking the trajectories of gas-permeance events. After that, we proposed other design of graphene/ionic-liquid composites with tunable slit pore size. Finally, to satisfy the requirement of exploring huge material database for gas separation, we applied machine learning to predict the selectivity of porous carbon materials. We further used convolutional neural networks to search the optimal porous carbon material from an ultimate input feature. The same approach was also used for prediction of hydrogen locations in copper clusters, which are potential materials for hydrogen storage. The works in this dissertation aims to find and design ultrathin nanoporous materials for gas separation by different computational approaches.