Tungsten — the metal chosen to line the most punishing parts of future fusion reactors — may suffer more radiation damage at extreme energies than scientists expected, according to new simulations reported this week.
In a study published in Physical Review Letters, researchers used machine-learning interatomic potentials to simulate radiation damage cascades in tungsten at an unprecedented scale: up to 1 billion atoms. The results reveal four distinct regimes of primary damage as a function of damage energy, and crucially, a transition to a high-energy regime that deviates from all earlier models.
Rather than following the assumed smooth trend, the damage produced per unit energy shifts from sublinear to superlinear before eventually becoming linear. In plain terms, in the energy range relevant to fusion neutrons, the metal accumulates damage faster than previous models predicted — which could shorten the estimated lifetime of plasma-facing components.
Tungsten is the leading candidate for the divertor and other plasma-facing components of experimental reactors, prized for its extremely high melting point and low erosion under heat. But those same components must absorb the full force of 14 MeV neutrons produced by deuterium-tritium fusion reactions, making radiation damage the key constraint on how long reactor parts can survive between replacements.
The new findings matter because component lifetime estimates feed directly into reactor economics: in a power plant, the availability of the machine — and its cost of electricity — depends on how often the most damaged parts must be swapped out.
The simulations also demonstrate a methodological leap: modeling a billion atoms is only possible thanks to machine-learned interatomic potentials, which capture quantum-accurate forces at a fraction of the computational cost of traditional first-principles methods. That scale lets researchers observe the full trajectory of primary damage events, from the first atomic collisions to the complex, extended defect structures that follow.
For fusion engineers, the message is to keep margins in mind: the metal may be harder-hit than the textbooks say, and validating damage models against experimental irradiation data is now more important than ever.
