IBM, Oak Ridge National Laboratory, and Cleveland Clinic have for the first time calculated nine molecular configurations of FLiBe — a molten salt composed of fluorine, lithium, and beryllium — using a quantum computer. This is the first documented instance of such simulations applied to nuclear fusion research, published on arXiv on June 29, 2026.
FLiBe is one of the most promising candidate materials for extracting tritium from fusion reactors. Tritium is a rare hydrogen isotope essential as fuel for most fusion machines currently under development. The challenge is straightforward to describe but enormously difficult to solve: securing adequate tritium supplies has been a major obstacle to large-scale fusion power for decades. Modeling FLiBe chemistry on classical computers becomes computationally prohibitive as system complexity scales up. The chemistry involved in extracting tritium from molten salt is so intricate that conventional methods cannot describe it accurately, while physical experiments with FLiBe demand enormous energy inputs and highly specialized equipment.
The team used IBM’s 156-qubit Heron quantum processor integrated with Oak Ridge’s supercomputing infrastructure, in an approach known as quantum-centric supercomputing: quantum circuits handle the portions of the problem best suited to that paradigm, while classical systems manage the rest. This hybrid architecture enabled more precise determination of FLiBe’s electronic structure and the way its atoms bond with tritium at a fundamental molecular level. In practice, the researchers were able to identify tritium binding properties across all nine configurations — information that would have remained out of reach using classical computing alone.
The work is part of the U.S. Department of Energy’s Genesis Mission, which aims to integrate supercomputing, artificial intelligence, and quantum computing to accelerate strategic scientific research. IBM is one of the mission’s industrial partners. The result adds to a series of milestones reached in 2026 by IBM quantum systems: simulation of real magnetic materials, creation of a half-Möbius molecule never previously observed, and modeling of biological proteins containing up to 12,635 atoms. Fusion represents the next step in this progression toward concrete scientific applications.
The collaboration is set to continue. The near-term goal is to reduce data transfer latency between quantum and classical resources and to scale simulations to larger molecular systems. The broader ambition is that the entire fusion energy ecosystem could use this class of simulations directly — to design and validate materials without relying on costly physical experiments. If this trajectory holds, quantum computing becomes a design tool for the fusion reactors of the future: not merely a supplement to laboratory testing, but a partial replacement for it.
