Skip to content

ReaxFF+ — a new reactive force field method for ionic systems and its application to the hydrolysis of aluminosilicates

Summary

ReaxFF+ extends ReaxFF with a modified charge-equilibration / bond-order coupling aimed at ionic liquids and gases, where standard QEq-style treatments can struggle to represent ionic vs covalent character of the same element types simultaneously. The authors motivate aluminosilicate hydrolysis in acidic/alkaline aqueous environments, benchmark cluster and reaction energetics against DFT, and illustrate ReaxFF+ on large reactive MD models—including >20,000-atom trajectories—using an efficient MD implementation. The corpus PDF is a Just Accepted J. Phys. Chem. C posting; cite the version-of-record for pagination-sensitive numbers when available.

Methods

MD application (atomistic dynamics)

Engine / implementation: reactive MD uses ReaxFF+ in the Tremolo-X solver within Atomistix ToolKit (ATK) (as stated in the Molecular dynamics simulations subsection of pdf_path), motivated by access to >20,000 atoms and >50 ps trajectories for hydrolysis-in-solvent models.

Integration / thermo–barostat: Δt = 0.2 fs is used for time integration (“small enough to account for water molecule vibrations and rotations” in the article). Nosé–Hoover thermostat and barostat are used for temperature/pressure control (references 35–37 in the article).

Illustrative aqueous supercell (liquid water benchmark): a ~6.2 nm cubic cell with 8000 H₂O molecules is built on a grid with small random displacements/orientations, pre-relaxed with conjugate gradients, then cycled through short NPT segments (300 K, 1 bar, 200 fs) repeated with CG until a final NPT equilibration (300 K, 1 bar, 60 ps) yields stable averages (pressure, temperature, cell volume, hydrogen-bond energy). The authors report an equilibrium density ~1378 kg m⁻³, noting overdensity relative to experiment as a known ReaxFF-family limitation, while showing reasonable O–H covalent and hydrogen-bond lengths and O–O RDF peak positions vs experiment.

Hydrolysis MERPs in solvent: for selected aluminosilicate hydrolysis reactions, minimum-energy paths are computed with ReaxFF+ inside a cavity carved from the equilibrated water supercell (van-der-Waals surface of the solute), yielding >24,000 atoms in three-dimensional periodic boundary conditions treated only with ReaxFF+ in the reported large-scale path calculations.

Electric field: N/A — not used as a controlled external bias in the documented workflows. Replica / enhanced sampling: N/A — not used beyond standard NPT equilibration and MERPs.

Force-field training (ReaxFF+ development)

ReaxFF+ introduces a revised charge-equilibrium scheme combined with the bond-order framework so covalent vs ionic bonds can be distinguished more consistently for problematic cases in standard ReaxFF/QEq (see Sections II–III). Training/validation lean on DFT reference data (including cluster ionics, water clusters, and aluminosilicate reaction energetics), with CP2K/Quickstep PBE-GGA GAPW setups and TZVP-class bases described for the DFT portions, and climbing-image NEB for paths. The article also describes a partially automated parameter-generation workflow aimed at reducing manual ReaxFF fitting effort.

Static QM / DFT

QM appears as the reference generator for energies/barriers and NEB benchmarks used to build and test ReaxFF+; it is not a separate “application DFT-only” study independent of the force-field development narrative.

Findings

  • Ionic cluster / water-ion benchmarks: the authors report that ReaxFF+ reproduces qualitatively correct ionic character for illustrative Na⁺·(H₂O)ₙ and OH⁻·(H₂O)ₙ clusters (charge localization on Na vs O of OH⁻) and captures key autoionization/deprotonation charge distributions better than their baseline ReaxFF description in the same tests.
  • Energetics vs experiment/QM: they note H₂O···H₂O stabilization is underestimated vs a cited experimental value (-3.9 kcal mol⁻¹ vs -5.57 kcal mol⁻¹) but still “reasonably” reproduced for a reactive FF water model, and they emphasize simultaneous accuracy for both Al(OH)₄⁻ → Al(OH)₃ + OH• and Al(OH)₄⁻ → Al(OH)₃ + OH⁻ dissociation energetics as a differentiator enabled by the new charge scheme.
  • Liquid-environment hydrolysis: Table 4 in the article compares reaction and activation energies for a representative aluminosilicate hydrolysis reaction in vacuum, with Na⁺, in liquid water, and in water + Na⁺, showing ReaxFF+ vs original ReaxFF alongside QM references; qualitatively, solvent can shift exothermicity and barrier ordering relative to vacuum, and Na⁺ effects differ between vacuum-like vs fully solvated setups.
  • Corpus honesty: some PDF pages in this ingest show stream decompression warnings in automated text extraction; treat tabular energies and MERP details as authoritative only when read from a clean VOR PDF, and keep extraction_quality: partial until extracts are regenerated.

Limitations

The corpus PDF is a Just Accepted manuscript (ACS disclaimer on the deposited PDF): figures, pagination, and numbers may change at technical editing. Prefer the version-of-record J. Phys. Chem. C PDF for authoritative barriers and elastic/bulk tables, then refresh pdf_sha256 / extraction_quality / normalized/papers/2016bohm-venue-paper.json accordingly. extraction_quality: partial reflects incomplete first-pass text extraction for this slug.

Relevance to group

Methodological extension of ReaxFF toward geochemistry-relevant silicate reactivity in electrolyte-like environments.

Citations and evidence anchors