Votta et al. (2602.22177) simulate what happens when ITER — the world's most expensive machine — has a disruption, and the news is that the standard mitigation strategy probably doesn't work in the mode that matters most.
Disruptions are the nightmare scenario for tokamak fusion. The plasma loses confinement suddenly, dumping its energy into the walls. To prevent this, ITER plans to inject shattered pellets — frozen deuterium and neon fragments — that cool the plasma in a controlled way. The problem: if cooling is too abrupt, thermal electrons can be accelerated to relativistic speeds by the collapsing electric field, creating a “runaway electron” beam that carries megaamps of current and can melt tungsten walls.
The paper adds four physics models to the simulation framework: vertical motion scrape-off, plasmoid drift for pellet deposition, hyper-resistive transport to prevent unphysical current channels, and updated Compton scattering seeds for ITER's tungsten walls. The results:
In non-nuclear H26 scenarios (the initial test campaigns), staggered pellet injection can avoid multi-megaamp runaway beams. But in DT H-mode — the actual fusion power operation — the conditions for complete avoidance are “typically violated when nuclear seeds are present.” The nuclear reactions themselves create seed electrons that amplify through avalanche processes.
There exists a narrow path: long pre-thermal-quench duration to thermalize hot-tail electrons, high deuterium assimilation with limited neon, and an initial seed current as low as a single relativistic electron. All three must be met simultaneously. Missing any one produces a runaway beam.
This is the engineering version of a phase transition: there's no gradual degradation. Either you're below the threshold on every axis simultaneously, or you get a catastrophic runaway beam. The margin for error in ITER's most important operating mode is, at best, thin.