Researchers from Oak Ridge National Laboratory, Cleveland Clinic, and IBM have used a quantum computer to study the atomic-level behavior of a material proposed for producing and recovering tritium fuel in fusion reactors. The work represents the first known use of heterogeneous quantum-classical computing to calculate tritium binding in fluorine-lithium-beryllium molten salt, known as FLiBe. The results were recently posted on arXiv.
Studying tritium interactions in molten salt
Many proposed fusion reactors would use tritium, a scarce hydrogen isotope, as fuel. Because natural supplies are limited, future reactors would need to produce, or breed, tritium inside a material surrounding the fusion plasma. FLiBe is a leading candidate for this blanket material. Lithium in the salt can produce tritium when exposed to neutrons generated by fusion reactions, but researchers must understand how the tritium binds to and moves through the material before they can optimize its recovery.
Accurately modeling those interactions is difficult because the number of possible electronic configurations grows rapidly as molecular systems become larger. The researchers addressed this challenge by combining quantum processing units with classical high-performance computing resources. Their workflow divided simulated FLiBe clusters into smaller atomic fragments. Classical computers solved the less complex fragments, while an IBM Heron r3 quantum processor analyzed larger fragments using extended sample-based quantum diagonalization. Classical systems then reconstructed the fragments into total energy calculations.
Evaluating nine molecular configurations
The researchers evaluated nine FLiBe configurations and compared the hybrid workflow’s calculations with classical reference methods. For individual fragments, the quantum-assisted method reproduced reference ground-state energies with a mean absolute deviation of 0.3 kilocalories per mole. The largest fragment calculations used as many as 66 qubits.
The study also exposed a major limitation. Differences between the fragmented models and full-molecule calculations reached an average of 110 kilocalories per mole for tritium binding energies. The researchers attributed that discrepancy primarily to how the workflow divided the system into fragments, rather than to the quantum hardware’s solution of those fragments. “This study provides an initial assessment of heterogeneous quantum-classical computing for investigating the interaction between tritium and molten salts,” the authors wrote.
Scaling the workflow for materials discovery
The current calculations covered small clusters and individual fixed configurations. Predictive simulations of tritium behavior in realistic molten salts would require larger systems, more configurations, improved fragment construction, and calculations of binding free energies. The research team plans to scale the workflow beyond the 21-ion clusters used in the study. Its longer-term goal is to develop a computational approach that fusion researchers could use to evaluate candidate salt compositions before conducting difficult and costly high-temperature experiments.
For lab managers, the study illustrates how emerging quantum systems could enter research environments as components of integrated computing workflows rather than as replacements for classical infrastructure. Supporting this type of work would require coordination among specialized hardware, high-performance computing resources, software platforms, data-transfer systems, and multidisciplinary research teams.
This article was created with the assistance of Generative AI and has undergone editorial review before publishing.
