Accelerating Solutions to Optimize Separations of Rare Earth Elements with High Performance Computing

D.A. Penchoff, C.C. Peterson, G.M. Bosilca, R.J. Harrison, E.F. Valeev, C.B. Sims
University of Tennessee,
United States

Keywords: REE, separations, computational, artificial intelligence


Rare earth elements (REEs) are essential in critical materials needed for clean energy, computers, satellites, medical devices, cell phones, hybrid vehicles, batteries, lasers, steel production, and many other instruments used in every-day applications and defense. China has become the largest producer of REEs and maintains a strong monopoly in the global marketplace. Environmental regulations across most of the globe cause the REE separation process to be costlier than importing REEs from China where regulations are less restrictive. Currently, simply increasing mining activity to break this monopoly is not a feasible solution since the mined material would have to be shipped to China for processing. To combat this reliance on foreign sources, nations could diversify their supply of REEs and become less reliant on a single country. This presentation will address promising paths to diversify supply of REEs, multidisciplinary solutions to optimize separations of REEs through the utilization of high performance computing (HPC) and artificial intelligence (AI), and efforts in the ‘Carbon, Ore, Rare Earth and Critical Materials (CORE-CM) Project at the University of Tennessee. The ultimate objective of the Southern Appalachia (SoApp) CORE-CM project is to develop and deploy new technologies for manufacturing REEs, critical minerals (CM), and valuable non-fuel, carbon-based products (CBP) from coal and/or coal waste in the SoApp Basin, thus revitalizing distressed SoApp coal communities and reducing reliance on foreign imports of REEs and CM. A diverse and experienced coalition team is identifying the critical technologies, infrastructure, supply chains, human capital, and policy gaps that must be addressed to manufacture REE, CM, and CBP from coal resources and wastes in SoApp. This region and team leverages existing assets to build an incomparable REE innovation ecosystem. Advances in HPC enable accelerated predictive capabilities in many areas. As we enter the exascale era, accurate predictions of REE separations require multidisciplinary perspectives to design software that accurately models the physical descriptions of REE-containing systems while taking advantage of novel heterogenous HPC architectures. Chemical descriptions utilizing electronic structure (ES) theory require implementation of advanced models of ES including effects of special relativity and faithfully describing the complex interplay of weak and strong correlation effects in systems with hundreds of atoms for predictive description of REE separation platforms. Current HPC capabilities allow applications of artificial intelligence to various areas. Separations of REEs require building reliable data sets and algorithmic evaluations in the prediction of binding characteristics. This presentation will address current efforts in CORE-CM focused on recovering REEs from coal ash in the Appalachian region, and efforts in HPC-enabling capabilities in molecular modeling and AI applications in separations of REEs. Methodologies discussed will include HPC development to advance REE-modeling at exascale levels, and applications of AI in separations of REEs. In particular, performance of algorithms including machine learning and neural networks applied to separations of REEs will be addressed. Additionally, current development of electronic structure theory methods implementation in MPQC and MADNESS, and results from surveying electronic structure theory methods accuracy in predicting selective binding to REEs will be discussed.