(360ac) Using Text-Mining and Community Knowledge to Quantify and Engineer Stability in MOFs

For over two decades, metal-organic frameworks (MOFs) have been developed for various applications in gas separations, sensing, and catalysis. In practice, we must activate a MOF and remove solvent from its pores to render it porous and usable. Simultaneously, the MOF must also be stable under the thermal conditions. Although the tailored metal active sites and porous architectures of MOFs are promising for separations, sensing, and catalysis applications, a lack of understanding of how to improve their stability limits their use. MOFs vary in their coordination geometries, pore sizes, coordination chemistry, metal identity, and oxidation states, which challenge the development of general structure-activity relationships that generalize over various families of MOFs.

We use natural language processing (NLP) to extract stability information over thousands of MOF manuscripts, to obtain insights on solvent removal and thermal stability of these materials. Since computation cannot readily predict thermal or solvent removal stability, publicly available experimental data provides an avenue to construct machine-learning (ML) models. We use our graph-based representation, revised autocorrelations (RACs) for MOFs, to make predictions over a diverse set of MOFs. For the first time, we curate an extensive set of experimental data on MOFs that enable us to map MOF structures to their corresponding experimental stabilities.

We use our NLP derived labels on MOF stability to train ML models that predict MOF solvent-removal and thermal stabilities. From feature insights from our interpretable representation, we find that MOF stability is primarily governed by linker chemistry. Simultaneously, these feature analyses indicate the possibility of orthogonal tuning of stability, showing that solvent removal and thermal stability can be separately engineered. We show how hypothetical changes to MOF structures can be tested for their effects on stability using our ML models. We then demonstrate how we can use our ML models to propose alterations to 3d transition metal MOFs that can make them stable under catalytic conditions for methane oxidation. We also show how our strategy can lead to improved hyper-stable materials for MOF screening.