Targeted metal–organic framework discovery goes digital: machine learning’s quest from algorithms to atom arrangements

Abstract

The integration of metal nodes with organic linkers in structured architectures offers the prospect of creating an extensive array of metal–organic frameworks (MOFs). Although this vast pool of materials has exciting possibilities, it also presents formidable challenges. Conventional techniques are ill-equipped to handle the sheer volume of materials. Consequently, over the past few decades, researchers have devised a range of empirical, semiempirical, and purely theoretical prediction models. Despite these efforts, these models have grappled with limited universality and accuracy. The advent of machine learning (ML) driven by big data has ushered in a new era impacting various scientific domains, including chemistry and materials science. As a new field of research, MOFs have reaped substantial benefits from ML. The approach not only unravels the intricate relationships between MOF structures and their performance but also sheds light on their diverse applications. In this comprehensive review, we delve into the scientific advancements that have propelled the computational modeling of MOFs, offering readers a fresh perspective on the transformative impact of ML in reshaping the research and development of reticular chemistry. Our exploration spanned from molecular simulations to the implementation of cutting-edge ML algorithms. As we explore this new domain, we enhance our comprehension of the fundamental principles governing MOF synthesis and enable applications across various engineering disciplines. Finally, we offer a forward-looking perspective on the potential opportunities and hurdles awaiting MOF design and discovery, based on the power of big data-driven approaches.