IBM, Oak Ridge National Laboratory and Cleveland Clinic used IBM quantum hardware to model how a fusion plant would produce and recover tritium, described as the first known instance of fusion-material computations run on quantum computers. The work was reported on 12 August.
Tritium is the constraint. It is a hydrogen isotope with a half-life of about 12 years, which means it cannot be stockpiled for long, and there is very little of it in the world. A fusion plant is therefore expected to breed its own by exposing lithium to neutrons, and to capture what it breeds using FLiBe, a molten salt of fluorine, lithium and beryllium that wraps the reaction like a blanket. The team modelled nine molecular configurations of that salt, with and without tritium present, to work out which arrangements recover it most efficiently.
This matters because tritium breeding is a genuine blocker on fusion power and a separate problem from the plasma physics that attracts the attention. A plant that confines its reaction beautifully but cannot recover more tritium than it consumes does not produce electricity for long.
The caveat is the category of result. This is a simulation milestone rather than an engineering one, and the accounts of it do not report how the output was validated against conventional simulation, which is the comparison that would establish whether the quantum machine did something a classical computer could not already do. Nine molecular configurations is also a long way from a materials programme. Quantum chemistry remains the most credible near-term application for this hardware, which is exactly why claims made in it deserve the closest reading.
