Amorphous Metal-Organic Framework Database for Amorphization Prediction and CO Direct Air Capture Screening.

The amorphous phases of crystalline metal-organic frameworks (MOFs), known as amorphous metal-organic frameworks (AMOFs), offer a vast yet underexplored search space for numerous applications. To efficiently navigate this chemical space, data-driven methodologies are essential for elucidating structure-property relationships. In this study, a novel approach to explore this space is introduced by utilizing chemically accurate reactive force field (ReaxFF) to first create the AMOF database. Through melt-quenching simulations performed using ReaxFF, a database consisting of 487 AMOF structures was constructed. From this database, machine learning models trained on the generated AMOF database accurately predicted which MOFs are more likely to go through thermal amorphization based on key structural features. Furthermore, promising candidates for CO direct air capture (DAC) applications were identified by running Widom insertion calculation, discovering three AMOF structures exhibiting DAC-favorable features after thermal amorphization. These findings highlight the effectiveness of this new data-driven approach for exploring the amorphous phases of MOFs beyond existing databases.