Skip to content

Atomistic insights into chemical vapor deposition process of preparing silicon carbide materials using ReaxFF-MD simulation

ReaxFF MD combined with DFT verification and an energy-based Monte Carlo (EMC)–MD workflow models methyltrichlorosilane (MTS) CVD on β-SiC, with emphasis on deposition efficiency, film stoichiometry versus temperature, and post-deposit crystallinity (OVITO IDS).

Summary

The study addresses SiC chemical vapor deposition (CVD) relevant to coatings (e.g., TRISO fuel particles) using ReaxFF MD for the C–Si–Cl–H system. DFT (Gaussian, B3LYP/6-31G**) validates bond/angle curves and lattice constants against ReaxFF. Production MD uses LAMMPS and OVITO, NVT with Nosé–Hoover thermostat, 100 MTS molecules above a β-SiC (100)/(010)/(001) slab, minimization, 100 ps equilibration at 300 K, ramp 0.05 K/fs to target temperature, 1000 ps hold, Δt = 0.2 fs, periodic x/y and fixed z, substrate sizes 5a₀×5a₀×3a₀ and 10a₀×10a₀×3a₀ (a₀ = 0.4358 nm). Temperatures 1600–3800 K appear in the parameter table. An EMC–MD–CVD layer uses DFT-derived relative bond-strength ordering for C, Si, Cl, and H at the surface to preferentially remove H and Cl during deposition passes, implemented with Gaussian and LAMMPS; crystallinity uses a defined metric combining cubic diamond and neighbor-shell fractions (OVITO).

Methods

  • ReaxFF and DFT checks: Established C–Si–Cl–H ReaxFF; comparison to B3LYP/6-31G** for Si–C stretch and Cl–Si–C / Si–C–H bends; lattice constant of β-SiC and MTS internal coordinates versus experiment (article Table 2).
  • Direct MD: LAMMPS, OVITO, NVT, Nosé–Hoover, Δt = 0.2 fs, 100 MTS precursors, thermal protocol above, β-SiC substrate orientation noted in text.
  • EMC–MD–CVD: DFT bond-energy curves for Si–Si, Si–C, C–C, Cl–Si, Cl–C, H–Si, H–C combined into surface-atom energy scales; periodic removal of weakly bound H/Cl by Monte Carlo rules coupled to MD segments (Fig. 5–6 in article); temperature-dependent deposition efficiency and film composition (Figs. 7–10).
  • Crystallinity analysis: OVITO Identify Diamond Structure (IDS) and a defined crystallinity metric from cubic diamond and first/second neighbor shell populations.

1 — MD application (ReaxFF CVD, direct path). Engine: LAMMPS + ReaxFF; NVT with Nosé–Hoover thermostat; Δt = 0.2 fs; 100 MTS molecules over β-SiC (100)/(010)/(001) slabs; 5a₀×5a₀×3a₀ and 10a₀×10a₀×3a₀ cells (\(a_0=0.4358\) nm); PBC in x,y; fixed z; minimization; 100 ps equilibration at 300 K; ramp 0.05 K/fs; 1000 ps hold; T 1600–3800 K in table; OVITO post. N/A — no NPT/barostat or external E-field in the direct MD path summarized here. N/A — no replica/metadynamics in that leg; EMC is a separate kinetic layer (below).

2 — Force-field training — N/A; uses established C–Si–Cl–H ReaxFF with DFT B3LYP checks (not a new FF fit paper in the sense of AGENTS block 2—see article for provenance). 3 — Static QM (validation): Gaussian B3LYP/6-31G** for bonds/ angles and lattice β-SiC vs ReaxFF; EMC uses DFT bond-strength orderings for surface H/Cl removal rules ( § of EMC in article). 4 — Hybrid EMC–MD — Monte Carlo H/Cl stripping with MD segments; N/A for a single NPT barostat in that subprotocol.

Findings

  1. The ReaxFF parameter set reproduces the validated DFT curves and β-SiC lattice data within small reported errors versus experiment.
  2. The EMC–MD–CVD model yields deposition efficiency versus temperature with a maximum at intermediate temperatures and reduced efficiency at very high \(T\) (e.g., 3800 K vs 1600 K), interpreted as difficulty forming C–Si bonds when chemistry is too hot—consistent with trends described alongside experimental literature cited in the paper.
  3. Film composition shifts from Si-rich at lower temperature to more balanced to C-rich at higher temperature, traced to gas-phase MTS pyrolysis products (e.g., CH₄ vs carbonaceous fragments) and differential removal of Si vs C at the surface in the EMC criterion—reported as aligning with prior experimental results from the authors’ group.
  4. Larger-scale trajectories show an initially amorphous deposit that orders toward β-SiC-like character; crystallinity metrics from OVITO IDS increase strongly with T in the range studied.

Corpus honesty — not a host-group paper; see ## Limitations.

Limitations

Highly reactive high-temperature chemistry and simplified gas-phase chemistry remain approximate; EMC removal rules are a kinetic shortcut rather than full gas-phase reaction networks.

Relevance to group

External collaboration context (Tsinghua); demonstrates ReaxFF + workflow coupling for industrial CVD-style SiC growth relevant to reactive MD practice in the broader community.

Citations and evidence anchors