Skip to content

Atomic level insight into the nucleation of SnSe thin films using a graphene mask in molecular beam epitaxy: ReaxFF molecular dynamics simulations

ReaxFF reactive MD is used to study selective-area nucleation of SnSe on MgO(001) with graphene masks, mimicking aspects of molecular beam epitaxy (MBE). An Sn/Se/Mg/O/C parameterization is trained against DFT (VASP, PBE, 500 eV cutoff, PAW, Γ-centered k-grids, DFT-D3, 20 Å vacuum) and prior ReaxFF data for Sn–Se phases and Sn/Se adsorption on MgO. The corpus PDF is a publisher galley; when a version-of-record PDF is available locally, prefer it for pagination and figures.

Summary

The work reports atomistic ReaxFF MD of SnSe nucleation on MgO(001) with perforated graphene masks, motivated by selective-area thin-film growth. The authors extend a Sn/Se/Mg/O/C ReaxFF (building on prior Sn/Se condensed-phase and cluster training), fit Sn–O, Se–O, Sn–Mg, and Se–Mg interactions to DFT adsorption data on MgO, and keep Mg/O from a Na/Ca/Mg/C/O/H parameterization; graphene edge atoms interact only through nonbonded terms with Sn, Se, and Mg. MD uses the AMS platform, orthorhombic cells (about 21 × 63 × 90 Å and 21 × 63 × 120 Å), periodic boundaries, 1–4-layer graphene masks with a central hole (~3.2 Å mask–substrate gap), Sn:Se 1:4 precursor supply, energy minimization, heating to 500 K in NVT, velocity Verlet integration with time steps of 0.25 fs or smaller, total run lengths of 50–1000 ps, and Berendsen thermostats. Two thermostat schemes are compared: a single strict thermostat (100 fs damping) versus a “mixed” scheme (100 fs on the bottom 12 MgO layers, 107 fs on precursors, mask, and the top four substrate layers) to approximate MBE heating from the substrate back side.

Methods

  • Reactive force field: ReaxFF bond-order formulation; training extends Chin et al. Sn/Se/Mg/O/C data (heats of formation, equations of state for α-Sn, Pnma SnSe, cubic SnSe, trigonal P3̅m1 SnSe₂, Sn\(_x\)Se\(_y\) clusters) with added VASP-PBE DFT binding energies for Sn and Se on three MgO adsorption sites; optimization of Sn–O, Se–O, Sn–Mg, Se–Mg bonds, off-diagonals, angles, and torsions. Mg/O from Dasgupta et al.; C–C/C–O covalent terms from literature; C–Se, C–Sn, C–Mg treated as nonbonded only so mask edges do not form covalent bonds to Sn, Se, or Mg.
  • DFT reference (training): VASP, PAW PBE, 500 eV cutoff, 0.1 meV electronic convergence, 0.01 eV/Å force threshold, dipole correction normal to the surface, Gaussian smearing 0.02 eV, DFT-D3, Γ-centered k-mesh equivalent to 4×4×1 for a 1×1 MgO c(0001) cell, 20 Å vacuum.
  • MD protocol: AMS; cells ~21 × 63 × 90 Å and 21 × 63 × 120 Å; periodic boundaries; graphene masks 1–4 layers with a central hole (24 C atoms removed); Sn:Se 1:4; minimization then NVT at 500 K; velocity Verlet; Δt ≤ 0.25 fs; Berendsen thermostat(s) as above; run lengths 50, 125, 250, and 1000 ps depending on the case.

1 — MD application (ReaxFF, MBE-mimic). The authors use the AMS platform to run molecular dynamics (MD) with ReaxFF; details match the MD bullet list below. NVT at 500 K, Berendsen thermostat (~100 vs ~107 fs damping in the split-substrate case), velocity Verlet, Δt ≤ 0.25 fs, PBC ~21×63×90/120 Å cells, 1–4-layer graphene mask with a ~3.2 Å gap to MgO, Sn:Se 1:4 supply, heating to 500 K, 50–1000 ps case-dependent runs. N/A — no NPT/barostat or external E-field in the summarized protocol. N/A — no metadynamics or replica methods in the summary here.

2 — Force-field training — the three bullet lists in Methods (ReaxFF construction + VASP PBE+DFT-D3 training data). 3 — Static QM — VASP PBE reaction/adsorption data for the fit; not a DFT-only application paper.

Findings

  1. A single-layer graphene mask promotes formation of the crystalline P3̅m1 SnSe₂ phase during nucleation relative to scenarios explored in the study.
  2. Using multiple thermostats (hot substrate bulk, weak coupling to gas and mask) reduces spurious gas-phase Sn\(_x\)Se\(_y\) clustering and is presented as better mimicking MBE-like conditions than a single thermostat for the same system size and duration.
  3. Mask thickness (1–4 graphene layers) and precursor loading modulate SnSe morphology (see article figures / SI).

Comparisons, limitations, and outlook. The ReaxFF trends are intended as relative kinetic readouts of nucleation morphology under thermostat schemes that mimic MBE-like heating; they do not replace continuum reactor CFD or long-time coarsening experiments without further calibration (see ## Limitations).

Limitations

Finite system sizes and short nanosecond-scale trajectories limit direct comparison to experimental coarsening timescales; galley PDF pagination may differ from the journal version.

Relevance to group

Develops and applies group ReaxFF methodology to 2D materials nucleation on oxide substrates with explicit processing-inspired thermostat design.

Citations and evidence anchors

Reader notes (navigation)