This project will deliver MOFA, an exascale code for the discovery of new materials for carbon capture. MOFA will provide unique capabilities to enable accelerated in silico design of metal-organic frameworks (MOFs). Running at scale on Aurora and Frontier, the team will create and release a new MOF database through the Materials Data Facility, which includes MOF exemplars that are resilient to humid environments and exhibit enhanced affinity and selectivity for carbon dioxide. 

MOFA will address the following grand challenges: 1) it will increase cross-platform GPU compatibility and performance by running at scale on Polaris (NVIDIA GPUs), Aurora (Intel GPUs),and Frontier (AMD GPUs); 2) it will combine generative AI, graph modeling, online learning, and Bayesian optimization to assemble MOFs with competitive properties for carbon capture, whose properties will be validated with state-of-the-art atomistic simulations; and 3) it will increase applicability for addressing materials development by rapidly converging to chemical design space regions where MOFs can be selected in terms of cost-effectiveness, synthesizability, and other manufacturing constraints. Ultimately, MOFA will enable scientists and industry partners to reduce time to solution and costs in the modeling of new materials for carbon capture by using cutting-edge generative AI and optimization methods, and robust HPC simulations in modern exascale computing environments.