Computational Screening of Trillions of Metal–Organic Frameworks for High-Performance Methane Storage

TLDR

Computational screening of MOFs has expanded, yet most studies are confined to initial material sets and rely on brute‑force methods. The study aims to develop a systematic strategy to identify MOFs with desired properties from over 100 trillion candidates using machine learning and evolutionary algorithms. The authors employ machine learning and an evolutionary algorithm to search an enormous MOF space. The algorithm identified 964 MOFs with methane working capacity above 200 cm³ cm⁻³ and 96 MOFs surpassing the current world record of 208 cm³ cm⁻³.

Abstract

In the past decade, there has been an increasing number of computational screening works to facilitate finding optimal materials for a variety of different applications. Unfortunately, most of these screening studies are limited to their initial set of materials and result in a brute-force type of screening approach. In this work, we present a systematic strategy that can find metal–organic frameworks (MOFs) with the desired properties from an extremely diverse and large set of over 100 trillion possible MOFs using machine learning and evolutionary algorithm. It is demonstrated that our algorithm can discover 964 MOFs with methane working capacity over 200 cm3 cm–3 and 96 MOFs with methane working capacity over the current world record of 208 cm3 cm–3. We believe that this methodology can take advantage of the modular nature of MOFs and can readily be extended to other important applications as well.