Metal-organic frameworks (MOFs) are crystalline microporous materials composed of transition metal nodes connected by organic linkers.1 The diversity of MOF applications is owed to the vast number of node and linker combinations yielding a wide variety of crystal structures with different topologies and, ultimately, properties. This structural diversity, on one hand, makes MOFs an excellent platform for building diverse functional materials, but on the other hand presents a challenge in selecting the optimal combination of nodes and linkers for a particular application. With MOFs being crystalline materials, their behavior is inherently linked to their crystal structures. Yet, until recently, the MOF field has suffered from the lack of a proven methodology for ab initio crystal structure prediction (CSP), a procedure for predicting structures of MOFs, starting from molecular diagrams of the constituent nodes and linkers. Our approach, based on ab initio random structure search (AIRSS) and Wyckoff alignment of molecules (WAM), has finally enabled CSP for MOFs, with proven generality with respect to metal coordination type and topological architecture.2 This presentation will describe our advances in developing MOF CSP, since the original publication. A family of energetic MOFs, initially designed via CSP, followed by experimental synthesis of structures matching the predictions will be presented. To date, this makes the first example of MOF structures being designed by means of ab initio CSP. This will be followed by the presentation of our investigation of the solvent effects on the outcome of MOF synthesis. Since MOF structures are inherently porous, their lattices can be stabilized by the incorporation of solvent molecules from the synthetic medium. This has a profound effect on the underlying crystal energy landscape, since predicted non-porous polymorphs of MOFs cannot benefit from this kind of solvent-mediated stabilization. An example of how such effects can be accounted for with the help of molecular dynamics (MD) simulations will be presented. Finally, our use of machine learning (ML) to train atomistic potentials, as a computationally- more efficient alternative to periodic density-functional theory (DFT) calculations will be discussed.

Ab initio crystal structure prediction of metal-organic frameworks

Fall 2022 abstracts, posters, and presentations will launch in mid-September.

ACS Fall 2022

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.