Porous Pseudomaterials for Studying Structure-PropertyRelationships of Gas Adsorption

The discovery in 1995 of metal-organic frameworks (MOFs) – with record-breaking surface areas – sparked exponential growth in research efforts dedicated to the development of new porous adsorbents, particularly for energy related gas storage applications. However, despite their promise, decades of research have yet to yield MOFs that perform well enough for many of these applications, particularly high-pressure vehicular natural gas storage and post-combustion carbon capture. To understand why, I developed a novel computational methodology for generating large (100,000+) libraries of randomly configured Lennard-Jones (LJ) crystals, or “pseudomaterials”, with the intention of calculating various adsorption-related properties of interest en masse using grand canonical Monte Carlo (GCMC) simulations. These libraries were used to map an n-dimensional structure-property space, where n refers to the number different structure- and property-parameters. One approach for generating these libraries of pseudomaterials is random sampling, where each structure is generated algorithmically at random; however, we attempt to improve the overall computational efficiency using alternative methods. These alternative methods include mutation algorithms and augmenting random sampling with property prescreening. Mutation algorithms identify pseudomaterials with unique structure-property combinations and selectively perturb their structural characteristics to sample the sparsely populated regions of the structure-property space. Property prescreening uses machine learning models to predict the properties of a pseudomaterial to justify whether it is a candidate worthy of more computationally expensive GCMC simulations; it is an attempt at reducing the computational expense associated with running simulations on pseudomaterials with redundant properties. The overall goal of this new computational methodology was to observe structure-property relationships for porous materials in general (i.e., not limited to any particular sub-class). I showed that understanding these structure-property relationships provides insights into the design of better adsorbents for a wide range of gas storage and separations applications. In the future, this methodology could potentially be extended to better understanding porous materials for catalysis, sensing, and more.