(346k) Leveraging Quantum-Chemical Screening Methods to Guide the Discovery of Promising Metal?Organic Frameworks

One of the major advances in computational materials science over the last decade has been the ability to use high-throughput quantum-chemical calculations to accelerate the materials discovery process. Automated density functional theory (DFT) workflows in particular show great promise in identifying new material design rules and narrowing down regions of chemical space in ways not possible via traditional small-scale experimental or computational approaches. This is particularly appealing for applications involving metal−organic frameworks (MOFs), as their highly modular design leads to a nearly unlimited number of structures that could potentially be realized by combing the vast array of inorganic nodes, organic linkers, and geometric topologies.

Recently, we developed a high-throughput periodic DFT workflow to computationally investigate large numbers of MOFs in a robust and automated manner, with a specific focus on catalytic and gas separation applications involving the binding and activation of small molecule adsorbates. As one representative example, we highlight how this workflow has been used to identify a simple ligand-exchange procedure that can turn an otherwise unselective cobalt MOF into one that strongly and selectively binds O₂ over N₂, which is appealing for industrial air separation applications. Experimentally obtained adsorption isotherms confirm this finding and indicate that the identified material has one of the highest room-temperature O₂/N₂ selectivities of a MOF to date. More broadly, we demonstrate how our automated periodic DFT workflow has been used to construct the largest database of MOF quantum-chemical properties and the implications that this may have for materials design in the future. We briefly discuss our ongoing efforts to leverage this resource to predict the quantum-chemical properties of tens of thousands of MOFs in a matter of seconds via recent advances in machine learning.