The building-block approach to the synthesis of metal-organic frameworks (MOFs) has opened the possibility to synthesise a virtually infinite number. This creates exciting opportunities, but also raise the question of how to identify and classify MOFs among the plethora of existing crystal structures – this is particularly difficult as the definition of MOFs is still under debate. At the same time experimental trial-and-error discovery of MOFs is not fast enough and therefore new methods accessible not only to computational researchers but mainly to experimentalists need to be developed.<br/><br/>In collaboration with the Cambridge Crystallographic Data Centre, we have developed a curated database containing all the MOFs deposited in the Cambridge Structural Database (CSD). This initiative provides the MOF community with tools to extract their desired structures from the pool of CSD crystalline structures and to visualise their data of interest. In order to extract the desired MOF structures, we developed a number of “look-for-MOF” criteria. We also developed new capabilities to enable researchers to browse and look for MOF families based on metal-clusters, chirality, surface chemistry (functional groups) and pore and network dimensionality. This has resulted in a regularly updated CSD-MOF subset of ca. 90,000 structures to date. This subset can also further evolve depending on changes to the definition of a MOF, allowing users to match the criteria relevant to their area of interest. We also offer the possibility of visualizing the MOF landscape of properties on an interactive webpage (https://tinyurl.com/CSD-MOF-vis).<br/><br/>We also have demonstrated the power of the CSD-MOF subset for computational high-throughput screening (HTS), where we analyzed their performance in different applications. We have completed the full cycle from the screening of MOFs to the identification and synthesis of optimal materials. Our studies delimit the relationships between structural properties and gas adsorption performance in dynamic 5D representations, allowing the final user to select any material analyzed. Furthermore, as the computational cost of the molecular simulations needed in HTS is still too high, we show how the integration of machine learning algorithms allows predicting adsorption performance not only of existing materials but also future ones. All in all, we believe the combination of developed tools will greatly enhance the HTS MOFs discovery for multiple applications.

Building a collection of metal–organic frameworks in the Cambridge Structural Database for materials discovery

2019 Fall National Meeting

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