Skip to content

Molecular dynamics simulations of the effects of vacancies on nickel self-diffusion, oxygen diffusion and oxidation initiation in nickel, using the ReaxFF reactive force field

Summary

A Ni–O ReaxFF description is developed from QM training data spanning Ni and NiO equations of state, Ni vacancy formation and self-diffusion, and O insertion and diffusion barriers in Ni. MD validation shows agreement with published Ni self-diffusivity and O interstitial diffusivity/activation energy. The work then examines how vacancies alter O transport and internal oxidation: a vacancy-pair oxygen migration mechanism is proposed, and simulations suggest voids at grain boundaries can promote local O segregation via strong O–vacancy binding, nucleating NiO in voids—linking atomistic transport to early oxidation microstructures. The study is framed as a mechanistic bridge between bulk diffusion coefficients and microstructural sites where oxide nucleates internally.

Methods

Force-field training (Ni–O ReaxFF). Parameters are fit to QM-derived data that include equations of state for several Ni crystal phases and for NiO, Ni vacancy formation energy, the vacancy-mediated Ni self-diffusion barrier in fcc Ni, and—because interstitial oxygen is central to oxidation kinetics—oxygen insertion energies and oxygen diffusion barriers used as additional training targets (abstract; full tables in papers/Zou_ActaMater_2014.pdf).

MD application (validation and oxidation scenarios in LAMMPS). After parameterization, MD uses the LAMMPS ReaxFF package with a uniform 0.25 fs timestep on three-dimensional periodic boundary conditions (PBC) supercells (bulk Ni, interstitial O models, and larger grain-boundary constructs as defined in the article). When temperature and pressure controls are active, the authors use a Berendsen thermostat (100 fs damping) together with a Berendsen barostat (3000 fs damping). Diffusion coefficients are extracted after switching to the NVE ensemble and accumulating mean-square displacements over ≥1 ns trajectories as described in the article. Additional NPT segments (e.g., zero-pressure relaxation of GB models) and high-temperature NVT stages (e.g., 1500 K for 200 ps in the void/oxidation discussion) appear in the Results/Methods narrative for specific structural setups; supercell sizes, GB construction details, and stage-by-stage run lengths are tabulated in the PDF rather than duplicated here. Pressure control: NPT segments use the cited Berendsen barostat to relax cell volume toward zero external pressure where applicable, whereas NVE diffusion analysis is strictly constant-volume with stress tensor-derived diagnostics only as reported in the article.

Static QM / DFT-only: N/A — the peer-reviewed contribution centers on ReaxFF training and ReaxFF MD; refer to cited QM references within the article for original DFT settings.

Electric field / shock / enhanced sampling: N/A — not part of the reported Ni–O diffusion and internal-oxidation initiation studies.

Findings

The ReaxFF parametrization reproduces Ni self-diffusion and interstitial O diffusion benchmarks, including diffusivities and activation energies that the authors report in quantitative agreement with published values. Beyond dilute interstitial hopping, they propose an oxygen–vacancy pair migration mechanism for O transport when vacancies are present. In grain-boundary models containing void space, simulations suggest strong O–vacancy binding can drive local oxygen segregation and NiO nucleation inside voids, offering an atomistic picture for internal oxidation ahead of uniform bulk scale growth. Validation is staged explicitly against literature Ni and O diffusion data; GB/void results are interpreted relative to experimental hints that GB oxygen short circuits may promote oxide inside GB voids. Vacancy concentration and GB topology shift oxygen partitioning between bulk-like interstitial paths versus vacancy-assisted channels and void-trapped oxide precursors (quantitative trends in the article’s figures). ReaxFF inherits QM-training approximations; GB models are finite and high-temperature, so extrapolation to long-time engineering oxidation requires the same caveats stated in the discussion.

Limitations

  • Complex real alloys (alloying elements, long-range stress) are not fully captured in a binary Ni–O model.
  • Short MD windows can miss rare long-range diffusion events; reported diffusivities should be interpreted as consistent with the sampling protocol in the article rather than as exhaustive high-temperature kinetics.

Relevance to group

Foundational metal oxidation + ReaxFF paper from the group with clear validation metrics—often cited for Ni/NiO reactive simulations.

Citations and evidence anchors

Reader notes (navigation)