Abstract

The success of the research on metal-organic frameworks (MOFs) and related porous materials over the past two decades makes it now possible to believe we can tailor-make materials with desired properties for several key environmental-related applications, such as carbon capture and energy storage. To fully realize the potential of this development, we need to find materials that perform optimally in multi-scale processes, considering scales from the molecular level to the chemical plant. However, this requires a holistic perspective over the full design and discovery process, which involves exploring immense materials spaces, various material properties, their synthesis, as well as process design and engineering. The complexity of exploring all potential options by conventional scientific approaches seems intractable. Therefore, we are now developing tools from the field of machine learning and artificial intelligence that will enable us to pursue our aim of understanding and designing materials in a new way.

In my talk, I will discuss some steps toward this approach, presenting machine learning methods in the context of the rational design of MOFs for adsorption applications, starting from the molecular scale with a view toward the next steps. We will discuss why and how to quantify the structural diversity of MOF material databases, assess newly-reported materials’ novelty, design MOFs with desired thermal properties, automate force field generation, and learn from failed experiments.