Skip to content

Reactive force-field molecular dynamics simulation for the surface reaction of SiH x ( x = 2–4) species on Si(1 0 0)-(2 × 1):H surfaces in chemical vapor deposition processes

Summary

Reactive molecular dynamics with a ReaxFF-style bond-order potential is used to study plasma-enhanced CVD–relevant surface chemistry: SiHx (x = 2–4) impinging on a hydrogen-terminated Si(100)-(2×1) surface. The work targets PECVD contexts where gas-phase radicals reach the substrate and participate in reflection, desorption, chemisorption, and physisorption. An existing parameter set is adjusted so that gas-phase dissociation energies of the SiHx species are better reproduced before surface collision studies. The authors motivate the work by the need to treat radical–surface outcomes as reactive events rather than prescribed sticking coefficients in reactor-scale models.

Methods

Force-field training and targets (A)

  • Framework: ReaxFF-type bond-order reactive MD for Si/H chemistry relevant to SiH\(_x\) (x = 2–4) interacting with hydrogen-terminated silicon.
  • Adjustment described in the article: an existing parameter set is refit so gas-phase dissociation energies of SiH\(_x\) species match reference energetics more closely before surface collision studies (see primary text for training targets and weighting).
  • QM reference level: follow the article’s statement of reference energies or benchmarks used to judge SiH\(_x\) dissociation—not restated here.

Molecular dynamics and collision protocol (B)

  • Engine / code: Reactive MD consistent with LAMMPS-style ReaxFF integration (as stated in Computational Materials Science; confirm package and version in the PDF).
  • Surface: Si(100)-(2×1):H; substrate temperature and adsorbed H coverage are swept as control variables.
  • Impacts: SiH\(_x\) radicals strike the surface one at a time (sequential impingement); each event is binned as reflection, desorption, chemisorption, or physisorption.
  • Reproducibility (numerics not duplicated here): Ensemble (NVT with surface thermostat where used), timestep, 3D PBC slab dimensions, QEq or fixed-charge electrostatics, Coulomb cutoff, and per-impact and cumulative trajectory length for sequential impingement of SiH\(_x\) are in the Comput. Mater. Sci. article; confirm there before reruns.

MD application (CVD, single-impact)

Engine / code: ReaxFF-based RMD (typically LAMMPS). Surface / composition: Si(100)-(2×1):H; incident SiH\(_2\)–SiH\(_4\) one radical per impact with outcome bins (reflection/desorption/chemisorption/physisorption). Substrate temperature and H coverage are control variables. N/A — no NPT barostat or hydrostatic pressure run called out in this summary (constant-volume or fixed lattice patterns per article). N/A — no shock piston or static interfacial electric field; N/A — no replica or metadynamics beyond RMD unless the SI says otherwise. Equilibration and per-impact/aggregate production duration (ps/ns): N/A — not restated in this wiki summary; see VOR/SI. Coulomb and QEq details per primary text.

Findings

Role of temperature

Higher substrate temperature promotes bond breaking in incident SiH\(_x\) and increases desorption of adsorbed species from the surface, shifting the balance among outcome channels in the authors’ classification.

Role of hydrogen coverage

Adsorbed H passivates dangling bonds, blocking chemisorption sites, while thermal motion of surface H facilitates reflection and desorption of incoming radicals.

Multiscale utility

Outcome statistics from single-impact trajectories supply discrete sticking/reflection weights usable as boundary conditions in reactor-scale models, with the caveat that sequential impacts omit concurrent flux and coverage transients.

Outlook (paper)

The authors suggest extending the same reactive framework to SiC, SiGe, and related CVD chemistries.

Limitations

Parameterization is focused on gas-phase dissociation energetics for selected species; transferability to other feedstocks or full reactor-scale behavior is not established in the excerpted material. One-radical-at-a-time impingement omits flux overlap and surface coverage transients present in high-rate CVD, so outcome statistics should be mapped to continuum boundary conditions with care. Software timestep and thermostat choices in the full PDF govern energy drift in long production segments; mirror those settings when reproducing sticking distributions.

Relevance to group

Illustrates ReaxFF surface chemistry in semiconductor CVD-like settings (Si/H), adjacent to broader reactive MD practice in the corpus.

Citations and evidence anchors