Recent advances in computational modeling of MOFs: From molecular simulations to machine learning

Probing Complex Chemical Processes at the Molecular Level with Vibrational Spectroscopy and X-ray Tools

“…Metal–organic frameworks (MOFs) are becoming increasingly attractive for scientists due to their chemical tunability and high internal surface area . These crystalline materials are comprised of metal ions connected by organic linkers to form porous structures that can be optimized for a particular purpose, promising a world of molecular-level “crystal engineering.” However, the synthetic principles behind MOF formation are still in the tinker-and-see stage, and more often than not they rely generally on brute force strategies such as high-throughput screening and trial-and-error . Active kinetic control of emergent behavior within these systems requires a detailed understanding of structural transformation, reaction, and diffusion kinetics on the system-relevant length and time scales.…”

Artificial intelligence assisted with designing metal‐organic frameworks (MOFs)

“…Furthermore, many MOF properties, such as their mechanical and thermal properties, are difficult to measure experimentally and are not well understood. [17][18][19] Despite these challenges, there have been several successful examples of AI being used to design new MOFs. One example is the work of Li and colleagues, who used a deep neural network to predict the band gap of double perovskites based on their crystal structure.…”

“…While there are databases of known MOFs, such as the Cambridge Structural Database, these databases do not contain information on all the properties of MOFs. Furthermore, many MOF properties, such as their mechanical and thermal properties, are difficult to measure experimentally and are not well understood 17–19 …”

Advancing CH4/H2 separation with covalent organic frameworks by combining molecular simulations and machine learning

“…Machine learning (ML) methods have been useful to analyse the huge amount of materials' data obtained from HTCS for establishing the relations between the structural and chemical properties of materials and their performances in different applications. 31–34 ML methods have been adapted to MOFs, 35–37 and very recently to COFs for gas storage and separation. 38,39 For example, Pardakhti et al 40 used ML algorithms to predict the CH 4 storage capacities of 69 839 hypoCOFs together with 17 846 porous polymer networks (PPNs), and showed that using chemical and structural properties as inputs of an ML algorithm leads to accurate CH 4 uptake predictions.…”