A CoNiV hydrogen potential says ordering, not just composition, sets uptake

Chen and co-authors have built a machine-learning interatomic potential for the Co-Ni-V-H system and used it to ask how chemical short-range order changes hydrogen behaviour in CoNiV, an alloy already reported to resist hydrogen embrittlement ^[1][2]. The finding that matters for anyone writing a hydrogen-service specification is that two cells with the same composition do not have the same hydrogen energetics: the ordered state has higher average hydrogen solution energies and fewer strongly binding sites than a chemically random alloy, which the authors say suggests lower bulk hydrogen uptake ^[5][12].

The order itself is specific and stubborn. The simulations show strong V-centered ordering that suppresses V-V clustering and survives up to 1000 K ^[3]. Adding 1.0 at.% hydrogen changes it negligibly, so in this model hydrogen is a passenger in the chemical arrangement rather than a driver of it ^[4].

The dislocation result is the part that constrains mechanism talk. At partial dislocations, hydrogen segregates preferentially to tensile core regions, but the core behaves as a shallow, reversible trap whose effect is much weaker than the chemical trapping states ^[6][7]. In other words, within this potential the local chemical environment, not the dislocation core, is where hydrogen is held; the authors conclude that robust local chemical order plays a critical role in uptake, partitioning and dislocation trapping ^[8].

Two cautions before this becomes a processing rule. First, these are computed energetics from a fitted potential, and the abstract frames reduced uptake as a suggestion rather than a measurement; it reports no experimental hydrogen uptake or desorption data ^[11]. Second, the comparison state is a chemically random alloy, which is a computational reference rather than a demonstrated product state. The paper's own result that the ordering holds to 1000 K cuts both ways: it argues the effect is not a fragile laboratory artefact, and it implies you would need aggressive quenching or a strongly non-equilibrium route to approach the random limit ^[3]. Nobody has published, in this abstract, a map from anneal-and-quench schedule to degree of order to measured uptake for CoNiV.

That map is the practical gap. If order is what regulates site energetics ^[8], then a chemistry-only specification for a concentrated alloy in hydrogen service is underdetermined, because it does not pin the variable doing the work. The same logic applies in reverse to additive manufacturing and other rapid-solidification routes, where the as-built state is the least likely to have equilibrated order and where post-build heat treatment would be the lever. None of that is demonstrated here; it is what follows if the mechanism holds outside the simulation.

The work was supported by the US Department of Energy, Office of Science, Basic Energy Sciences, Materials Sciences and Engineering Division under award DE-SC0025170, with computing at NERSC under contract DE-AC02-05CH11231 and award BES-ERCAP0036165 ^[9]. It is published open access in npj Computational Materials, with the authors declaring no competing interests ^[10][13].

What to watch: whether the Co-Ni-V-H potential is released in a form other groups can run, since the whole argument rests on its site energetics ^[1]; whether anyone measures hydrogen content or desorption spectra on CoNiV samples heat-treated to different degrees of order, which is the direct test of the uptake claim ^[5]; and whether the same treatment is extended to grain boundaries and vacancy-hydrogen clusters, which the abstract does not cover but which dominate real hydrogen inventories ^[8].