Development of a Charge-Implicit ReaxFF Potential for Hydrocarbon Systems
Summary¶
This Journal of Physical Chemistry Letters communication introduces charge-implicit ReaxFF (ci-ReaxFF) for C/H systems by removing on-the-fly charge equilibration (no EEM/QEq-like updates each step). The motivation is computational: conventional ReaxFF’s explicit charge dynamics can dominate cost in very large simulations and can produce unphysical charged fragments in high-energy collision regimes where electronic equilibration is not the right physical picture. The authors reoptimize bonded and nonbonded parameters against an expanded training corpus that includes condensed-phase configurations and Ziegler–Biersack–Littmark (ZBL)-style short-range repulsive pair terms for keV-scale nuclear collisions, aiming to retain ReaxFF-2008-class accuracy on the training metrics while improving throughput and stability for organic solid and particle bombardment use cases.
Methods¶
Training follows the letter’s Development section: start from ReaxFF-2008 datasets, augment with condensed-phase benchmarks and ZBL high-energy pairs using distance-dependent weighting so that close encounters are dominated by the nuclear repulsive wall. The authors also revisit van der Waals cutoff selection using β-carotene NPT equilibration tests described in the text. Implementation targets LAMMPS compatibility; potential files and usage examples are referenced to Supporting Information.
MD application (throughput benchmarks)¶
Engine / code: LAMMPS molecular dynamics with the proposed charge-implicit ReaxFF for C/H systems (abstract). System sizes: benchmarked from ~13k to ~900k atoms in the reported speed tests. PBC: bulk organic cells implied by β-carotene NPT tests and keV bombardment examples. Ensemble: NPT referenced for β-carotene equilibration; NVE/NVT stages for cascades—N/A — full staging not transcribed from the excerpt. Timestep / thermostat / barostat / run length: N/A — import from letter + SI (papers/Kanski_Ashraf_JPC_letters_2018.pdf). Temperature: cascade tests emphasize keV energies rather than a single thermostat K in the abstract; thermal benchmarks use the letter’s stated conditions. Pressure: NPT targets appear for β-carotene cutoff tests (exact pressure in Methods). Electric field: N/A — not used. Enhanced sampling: N/A — not indicated beyond standard MD for the throughput comparisons summarized in the abstract.
Findings¶
The manuscript reports that ci-ReaxFF matches ReaxFF-2008 on the quoted training-quality metric while delivering roughly 2–5× faster MD in reported tests spanning 13k–900k atoms. Removing explicit charge dynamics is argued to eliminate spurious charged species in high-energy collision cascades that motivated the development. Applicability is explicitly hydrocarbon-focused as parameterized; polar chemistry and strong electrostatic heterogeneity are outside the intended scope.
Comparisons. Speed and accuracy comparisons are internal to ReaxFF-2008 vs ci-ReaxFF under the letter’s benchmark protocols.
Sensitivity / outlook. Performance depends on system size and collision energy regime; users should mirror the letter’s ZBL blending and cutoff tests before claiming portability.
Corpus honesty. Numeric speedup factors and atom ranges come from the J. Phys. Chem. Lett. abstract; reproduce from the PDF/SI when auditing.
Limitations¶
Implicit charges cannot reproduce full electrostatic chemistry of full ReaxFF; systems requiring accurate ionic solvation or strong charge separation need the explicit charge model or alternative parametrizations.
Reproducibility notes¶
Operators should record ZBL blending weights, cutoffs, and timestep choices when running ci-ReaxFF cascade simulations, because keV-range collisions are sensitive to timestep-dependent energy deposition if integration is too aggressive. Benchmarks against full ReaxFF should use identical geometry and ensemble settings to isolate charge-model effects from protocol differences.
For organic β-carotene benchmarks used in cutoff selection, archive NPT target pressure, temperature, and equilibration length, because reported density sensitivities are only meaningful relative to those reference conditions. When publishing throughput comparisons, include hardware, MPI decomposition (if any), and neighbor list rebuild intervals so speedups are not tied to a single machine configuration.
Relevance to group¶
Adri C. T. van Duin-linked ReaxFF variant aimed at fast reactive simulations of organic solids and particle bombardment.