A ReaXFF carbon potential for radiation damage studies
Summary¶
Molecular dynamics of collision cascades in graphite has relied for decades on empirical bond-order potentials, but comparisons to density functional theory showed mismatches in predicted defect populations and transition pathways. This article refits a ReaxFF carbon model to reproduce LDA reference formation energies and pathways among interstitial, vacancy, di-interstitial, and divacancy defects in graphite, including the spiro interstitial and Stone–Wales (Dienes) motifs emphasized in prior DFT compilations. Training augments existing hydrocarbon and fullerene datasets used for the broader ReaxFF carbon line. Elastic constants are acknowledged as less accurate than point-defect properties: \(c_{33}\) is too high and \(c_{44}\) too low versus experiment. Preliminary low-energy cascade simulations produce point-defect motifs consistent with ab initio expectations.
Methods¶
1 — MD application (atomistic dynamics)¶
ReaxFF molecular dynamics is used for preliminary low-energy collision cascades in graphite to check that defect populations after impact trend toward LDA-consistent interstitial and vacancy motifs following the refit (Nucl. Instrum. Methods Phys. Res. B 393, 49–53). The article frames these cascade segments as scoping studies rather than exhaustive high-energy damage statistics.
- Engine / code: ReaxFF reactive MD as implemented in the article (confirm package and input decks in
pdf_path). - System size & composition: Graphite supercells for cascade tests; explicit atom counts and PKA energies are N/A on the indexed excerpt.
- Boundaries / periodicity: N/A — PBC vs fixed boundaries for cascades not stated in the indexed excerpt.
- Ensemble: N/A — NVE/NVT staging for cascade segments not stated in the indexed excerpt.
- Timestep: N/A — Δt (fs) not stated in the indexed excerpt.
- Duration / stages: N/A — thermalization vs cascade segment lengths not stated in the indexed excerpt.
- Thermostat: N/A — not stated in the indexed excerpt (cascade protocols often use NVE + boundary thermostats; confirm in PDF).
- Barostat: N/A — NPT not indicated for cascade checks on indexed pages.
- Temperature: N/A — initial lattice T / thermostat targets not stated in the indexed excerpt.
- Pressure: N/A — not stated for cascade cells in the indexed excerpt.
- Electric field: N/A — not used.
- Replica / enhanced sampling: N/A — direct cascade MD.
2 — Force-field training¶
ReaxFF parameters for carbon are reoptimized against LDA formation energies, barriers, and reaction pathways for graphite defects—including spiro, Stone–Wales (Dienes), di-interstitial, and divacancy sequences—using the Latham et al. QM database cited in the paper, merged with the existing hydrocarbon/fullerene ReaxFF training lineage so radiolytic and organic chemistry can coexist in one parameter set.
3 — Static QM / DFT-only¶
LDA QM supplies static reference energies and pathway data for the fit; on-the-fly AIMD is not the reported production tool for cascades.
4 — Review / non-simulation framing¶
N/A — primary research article.
Findings¶
Outcomes and mechanisms¶
The refit ReaxFF tracks many defect formation energies and migration/barrier sequences from LDA more closely than empirical bond-order alternatives highlighted in the introduction, while retaining compatibility with the broader hydrocarbon reactive carbon line.
Comparisons¶
Elastic constants remain less accurate than point-defect properties: \(c_{33}\) is too high and \(c_{44}\) too low versus experiment as acknowledged in the article—so the potential should not be used as a high-fidelity phonon model without further work.
Sensitivity / design levers¶
Because training extends rather than replaces hydrocarbon/fullerene datasets, users can combine radiation damage with chemical reactivity in one framework—contrasting with potentials fit only to elastic data.
Limitations, outlook, and corpus honesty¶
High-energy cascades and cumulative dose statistics need broader validation than the preliminary runs summarized here. Numerical cascade settings and defect counts should be taken from pdf_path, not inferred from this short wiki note.
Limitations¶
High-energy cascade regimes and cumulative dose effects may require additional validation; elastic property errors may couple into thermal transport if misused.
Corpus notes¶
Because this potential extends hydrocarbon datasets, workflows that combine radiolytic chemistry with hydrocarbon adsorption should document which parameter subsets dominate energy errors—especially when migrating from older Tersoff or AIREBO baselines.
The publisher PDF is open access under CC BY, which simplifies redistribution of figures in internal presentations provided attribution is preserved.
Relevance to group¶
RxFF Consulting / van Duin co-authorship connects the reactive carbon line to nuclear materials and radiation damage applications.