Predicting hydrogen storage in MOFs via machine learning

• Main Authors: Alauddin Ahmed, Donald J. Siegel
• Format: Article
• Language: English
• Published: Elsevier 2021-07-01
• Series: Patterns
• Subjects: energy storage; fuel cells; metal-organic frameworks; hydrogen storage; machine learning; materials discovery
• Online Access: http://www.sciencedirect.com/science/article/pii/S2666389921001240

Summary: The H2 capacities of a diverse set of 918,734 metal-organic frameworks (MOFs) sourced from 19 databases is predicted via machine learning (ML). Using only 7 structural features as input, ML identifies 8,282 MOFs with the potential to exceed the capacities of state-of-the-art materials. The identified MOFs are predominantly hypothetical compounds having low densities (<0.31 g cm−3) in combination with high surface areas (>5,300 m2 g−1), void fractions (∼0.90), and pore volumes (>3.3 cm3 g−1). The relative importance of the input features are characterized, and dependencies on the ML algorithm and training set size are quantified. The most important features for predicting H2 uptake are pore volume (for gravimetric capacity) and void fraction (for volumetric capacity). The ML models are available on the web, allowing for rapid and accurate predictions of the hydrogen capacities of MOFs from limited structural data; the simplest models require only a single crystallographic feature. The bigger picture: The efficient storage of hydrogen fuel remains a barrier to the adoption of fuel cell vehicles. Although many storage technologies have been proposed, adsorptive storage in metal-organic frameworks (MOFs) holds promise due to the low operating pressures, fast kinetics, reversibility, and high gravimetric densities typical of MOFs. Nevertheless, the volumetric storage densities of known MOFs are generally low; hence, new MOFs with improved volumetric performance are desired. Identifying optimal MOFs remains a challenge, however, because relatively few MOFs have been characterized experimentally, and the building-block structure of MOFs suggests that the number of possible materials is limitless. To accelerate the discovery process, this study develops machine learning models that predict the hydrogen capacity of MOFs. The models identify promising materials, clarify structure-property relations, and can be used—on the web or through an API—to predict the performance of new MOFs.