Computational Characterization of Nanoporous Materials – Zeolites

Porous materials have many important applications in the chemical industry.  Zeolites, for example, are used for gas separation as well as cracking catalysts in oil refining, alkylation and isomerization reactions catalysts.  Although the number of possible zeolite structures has been estimated to be more than 2.5 million, only about 180 structures have been synthesized to date.  Our long-term goal is to develop methods that will allow the identification of structures with optimal performance for specific applications, with a focus on predicting new materials for CO2 separation.  Although current state-of-the-art molecular simulations allow for accurate prediction of zeolite properties, the characterization of an entire database of hypothetical structures would be out of the question even with today’s supercomputers. 

We have proposed an approach that will allow the screening of large databases of structures.  Only structures that are predicted to exhibit properties of interest are subjected to a follow-on characterization using the more accurate but also more expensive molecular simulation methodology. In our approach the selection of structures to undergo characterization is done using knowledge-based techniques. The latter are derived by analyzing a sample set of zeolite structures to identify the correlations between predicted properties and a number of relatively easy to compute structural descriptors. The later predictions of which structure should undergo characterization are done solely on these structural descriptors.

The provided computer-time allocation will be spent on both performing accurate molecular simulations to predict the adsorption and diffusion properties of the nanoporous materials as well as calculating structural descriptors e.g. ones representing channel and cage topologies. Obtaining these results is the first step towards building the presented knowledge-based screening technique.

Berend Smit, University of California, Berkeley
Maciej Haranczyk, Lawrence Berkeley National Lab