Tungsten's damage curve has a bump in it, and fusion lifetime models miss it

Researchers at the University of Helsinki have published molecular dynamics simulations in Physical Review Letters indicating that tungsten accumulates primary radiation damage in a pattern current models do not describe: the defect count rises sublinearly with recoil energy, then superlinearly, then linearly ^[1][3]. Since tungsten is among the leading candidates for the components that face the plasma, a scaling law that is wrong in the middle of the range is a component-lifetime question rather than an academic one ^[4].

The quantity in dispute is narrow and specific. When a fast neutron strikes a lattice atom, it knocks that atom off its site; the displaced atom collides with others and displaces them in turn, and the resulting cascade is what the field calls primary radiation damage ^[5][6]. One standard way to measure it is simply to count how many atoms end up out of position after a single recoil ^[7]. In metals, according to first author Jesper Byggmästar, existing models assume that count grows sublinearly with recoil energy and then linearly ^[2]. The Helsinki runs report a third regime in between ^[3]. The practical consequence follows arithmetically: extrapolations built on a sublinear-then-linear law will undercount defects across the superlinear window ^[14].

The result is downstream of a tooling decision rather than a physics hypothesis. Byggmästar says the project started as an effort to port the group's machine-learning simulation model to GPUs, after which the team realised it could run larger and more accurate simulations than before and turned to tungsten under extremely high-energy ion irradiation ^[8][9]. Resolving all three transitions, he says, required atomistic simulation at unprecedented scale for the field, reaching one billion atoms at once ^[3]. The group frames that as two separate claims: that accurate billion-atom molecular dynamics is both feasible and capable of surfacing new physics, and that the damage model extracted from the runs can feed predictions at longer, reactor-relevant timescales ^[10][11].

Two limits are worth holding onto. The paper covers primary damage only, meaning the damage from a single atomic recoil, while an operating reactor or an irradiation experiment subjects the material to far more irradiation over much longer times and length scales, which Byggmästar describes as an active research problem in its own right ^[12]. And the published account states that tungsten would deteriorate more than originally anticipated at fusion-relevant energies without putting a number on how much more ^[13]. The direction is asserted; the magnitude that a designer would need is not, at least not in this account.

What to watch: whether the superlinear regime survives independent replication with different interatomic potentials, and whether the extracted damage function is actually adopted by the longer-timescale codes it is meant to feed ^[11]. The Helsinki group says it is now working on higher accuracy and still larger systems, and that the same approach could be pointed at other materials in high-energy environments ^[15][16]. Anyone sizing divertor replacement intervals on a sublinear-then-linear assumption has a reason to check where their recoil spectrum sits relative to the bump.