Skip to content

New ReaxFF Reactive Force Field Optimized for Vibrational and Thermal Properties of Molybdenum Disulfide

Summary

Phonon dispersions and lattice thermal conductivity of two-dimensional transition metal dichalcogenides are expensive at full first-principles fidelity in large, defective cells. This letter introduces ReaxFF₂₀₂₄ for monolayer MoS₂, fit using a comparatively small DFT training set, with the explicit goal of reproducing harmonic and anharmonic lattice dynamics while retaining ReaxFF’s ability to model bond rearrangements. Relative to the earlier ReaxFF₂₀₁₇ MoS₂ parameterization, the new force field shifts zone-center optical phonon frequencies much closer to DFT references and predicts an in-plane thermal conductivity κ∞ ≈ 73 W K⁻¹ m⁻¹, described as falling within the spread of experimental measurements quoted in the paper.

Methods

  • Baseline and target: Prior ReaxFF₂₀₁₇ for 2D MoS₂ (Ostadhossein et al.) reproduces vacancy energetics but overestimates optical phonon frequencies; ReaxFF₂₀₂₄ is reoptimized with emphasis on small-displacement dynamics near equilibrium as well as reactive degrees of freedom.
  • Training data: Density functional theory calculations—described as a modest number of DFT evaluations compared with typical machine-learning potential datasets—used to fit bond-order and related ReaxFF terms for Mo–S interactions relevant to phonons and thermal transport.
  • Validation: Comparison of Γ-point and dispersion-related phonon features to DFT (e.g., A₂″, E′ modes cited in the abstract); thermal conductivity from equilibrium MD using Green–Kubo heat-flux autocorrelation (as standard for κ in the letter’s framing).
  • Context: Discussion contrasts against ML potentials (SNAP, GAP, MTP) and nonreactive fits (e.g., Stillinger–Weber) that target phonons but omit reactivity.

1 — MD application (atomistic dynamics). LAMMPS molecular dynamics with the letter’s ReaxFF on monolayer MoS₂ (PBC 3D supercell; see VOR for atom count). NPT equilibration: Berendsen thermostat (damping 100 fs) and Berendsen barostat (damping 5000 fs), 0.25 fs time step, 25 ps duration, target near 300 K as in the main text. Melt/anneal stages in the SI use 0.5 fs in places. κ from NVT equilibrium Green–Kubo with isothermal temperature sets per the letter (e.g. ~300 K and higher for thermal transport). N/A — no static electric field in these MD sets; N/A — no replica or metadynamics; NPT barostat only on the equilibration leg; κ production is NVT as stated.

2 — Force-field training. Parent: ReaxFF ReaxFF₂₀₂₄ for 2D MoS₂ built from ReaxFF₂₀₁₇-class parameters; DFT reference for formation energies, harmonic/anharmonic phonon targets, and small-displacement dynamics; QM training set described as small versus MLIP corpora. Optimization and parameter table in the letter and SI; validation against DFT phonon frequencies and experiment-anchored κ quotes.

3 — Static QM / DFT-only — N/A (DFT is reference for ReaxFF training, not a separate DFT “application” block for this page).

4 — Review — N/A.

Findings

  • Optical modes at Γ: ReaxFF₂₀₂₄ moves the zone-center A₂″ and E′ frequencies from 1438 cm⁻¹ and 822 cm⁻¹ (ReaxFF₂₀₁₇) to 456 cm⁻¹ and 358 cm⁻¹, near DFT values 461 cm⁻¹ and 367 cm⁻¹ quoted in the abstract.
  • Thermal conductivity: κ∞ ≈ 73 W K⁻¹ m⁻¹ for the monolayer model studied, consistent with the experimental range cited by the authors.

  • Positioning: The authors highlight ReaxFF as simultaneously trainable for near-equilibrium phonon dynamics and reactive chemistry, enabling defect and heat transport studies in TMDs in one classical stack; outlook in the letter points to future transfer tests (see ## Limitations).

  • Corpus honesty: This note tracks a galley pdf_path; any erratum-level number change belongs to the VOR.

Limitations

Galley PDF in corpus; check the final journal pages for any numerical updates. Classical MD thermal conductivity depends on simulation size, thermostat, and duration; the letter should be consulted for convergence tests. Transferability to other chalcogenides is suggested but requires separate validation.

Relevance to group

Core van Duin-group ReaxFF development for 2D MoS₂ with Vincent H. Crespi and colleagues; supports large-scale reactive + thermal transport modeling.

Citations and evidence anchors

Reader notes (navigation)

  • Corpus filename includes galley; when a clean JPCL version-of-record PDF is ingested, update pdf_path / pdf_sha256 via manifest sync scripts.