What makes a plausible metal–organic framework linker? Discovery from a general chemical database

Abstract

Metal–organic frameworks (MOFs) have attracted significant interest for their diverse applications and tunable properties derived from their organic linkers and inorganic secondary building units. However, the organic linkers historically used in MOFs constitute only a narrow subset of the synthetically realizable small- to medium-sized organic molecules. In this work, we examine the typical characteristics of linkers in experimentally characterized MOFs, and we investigate the potential of the largest open-access chemical database, PubChem, as a source of organic molecules that could be incorporated into MOFs as linkers. Based on the analysis of connectivity, rigidity, and geometry of linkers present in the CoRE MOF (Computation-Ready, Experimental Metal–Organic Framework) database, we apply a series of chemical and geometric filters to PubChem molecules to identify linker candidates that we incorporate into hypothetical MOFs. We then predict the stabilities of a representative subset of hypothetical MOFs with machine learning models. From the most stable hypothetical MOFs, we identify commonly occurring PubChem-derived candidate linkers, and we identify design rules for multivariate stability such as the importance of moderate steric bulk. By presenting analyses of linkers in experimentally reported MOFs, demonstrating a workflow for identifying potential MOF linkers from chemical databases, and disseminating our candidate linkers and hypothetical MOFs, we expect this work will facilitate the discovery of new linkers and the creation of new MOFs with tailored properties.