Predicting Gas Adsorption without the Knowledge of Pore Structures: A Machine Learning Method Based on Classical Density Functional Theory

Predicting gas adsorption from the pore structure is an intuitive and widely used methodology in adsorption. However, in real-world systems, the structural information is usually very complicated and can only be approximately obtained from the characterization data. In this work, we developed a machine learning (ML) method to predict gas adsorption form the raw characterization data of N2 adsorption. The ML method is modeled by a convolutional neural network and trained by a large number of data that are generated from a classical density functional theory, and the model gives a very accurate prediction of Ar adsorption. Though the training set is limited to modeling slit pores, the model can be applied to three-dimensional structured pores and real-world materials. The good agreements suggest that there is a universal relationship among adsorption isotherms of different adsorbates, which could be captured by the ML model.

Connecting theory and simulation with experiment for the study of diffusion in nanoporous solids

Nanoporous solids are ubiquitous in chemical, energy, and environmental processes, where controlled transport of molecules through the pores plays a crucial role. They are used as sorbents, chromatographic or membrane materials for separations, and as catalysts and catalyst supports. Defined as materials where confinement effects lead to substantial deviations from bulk diffusion, nanoporous materials include crystalline microporous zeotypes and metal–organic frameworks (MOFs), and a number of semi-crystalline and amorphous mesoporous solids, as well as hierarchically structured materials, containing both nanopores and wider meso- or macropores to facilitate transport over macroscopic distances. The ranges of pore sizes, shapes, and topologies spanned by these materials represent a considerable challenge for predicting molecular diffusivities, but fundamental understanding also provides an opportunity to guide the design of new nanoporous materials to increase the performance of transport limited processes. Remarkable progress in synthesis increasingly allows these designs to be put into practice. Molecular simulation techniques have been used in conjunction with experimental measurements to examine in detail the fundamental diffusion processes within nanoporous solids, to provide insight into the free energy landscape navigated by adsorbates, and to better understand nano-confinement effects. Pore network models, discrete particle models and synthesis-mimicking atomistic models allow to tackle diffusion in mesoporous and hierarchically structured porous materials, where multiscale approaches benefit from ever cheaper parallel computing and higher resolution imaging. Here, we discuss synergistic combinations of simulation and experiment to showcase theoretical progress and computational techniques that have been successful in predicting guest diffusion and providing insights. We also outline where new fundamental developments and experimental techniques are needed to enable more accurate predictions for complex systems.

Fast screening of porous materials for noble gas adsorption and separation: a classical density functional approach

The design and screening of porous materials for noble gas adsorption and separation are an important issue in the production and utilization of gases. The conventional method to do this is via molecular simulation. In this work, we introduced a classical density functional theory (CDFT) to replace molecular simulation because CDFT is more efficient. A molecular dynamics (MD)/CDFT combined method was proposed to consider the flexibility of the adsorbent. The theory was first examined by comparing it to reported experiments and simulations. Then, the theory was applied to determine the most favorable adsorbents for noble gas adsorption/separation from 4764 real adsorbents and 1200 hypothetical adsorbents. A series of favorable adsorbents was identified, and some of them seemed promising. The macroscopic adsorption isotherms and microscopic density profiles of the most favorable adsorbents were examined, and the adsorption mechanisms were revealed. The specific separation of Kr/Xe was examined, and two of the adsorbents showed higher adsorption efficiency than shown in previously reported data.