Skip to content

Early stage oxynitridation process of Si(001) surface by NO gas: Reactive molecular dynamics simulation study

Summary

Ultrathin gate oxynitrides on silicon are produced by exposing clean surfaces to nitric oxide so nitrogen and oxygen incorporate within a few angstroms of the interface. Cao et al. apply ReaxFF reactive molecular dynamics to NO chemisorption and reaction on reconstructed Si(001) between 300 K and 1000 K, aggregating 1120 independent single-molecule NO events to extract temperature-dependent kinetic information for early oxynitridation. Complementary density-functional nudged elastic band calculations compare activation barriers for intra-dimer versus inter-dimer NO bridge precursors prior to dissociation, while sequential NO dosing simulations probe self-limiting film growth and nitrogen/oxygen depth profiles.

The study is motivated by advanced metal–oxide–semiconductor field-effect transistors, where silicon oxynitride films buffer the silicon channel from high-\(k\) dielectrics, suppress boron penetration from doped gates, and tune interfacial defect densities relative to pure oxide stacks. Nitric oxide is treated as a primary nitrogen precursor compared with nitrous oxide routes often used in furnace chemistry.

Methods

MD application (atomistic dynamics)

  • Engine / code: LAMMPS is used for the reactive molecular dynamics simulations (J. Appl. Phys. 119, 125305; pdf_path).
  • Interaction model: ReaxFF for N / O / Si / H as referenced in the article (including the benchmark note pointing to SI-1).
  • System / composition: Si(001) slab model (1476 Si atoms) with lateral dimensions 2.256 nm × 2.256 nm and 5.282 nm total thickness; ~1.1 nm vacuum above the surface; dimer-row orientation aligned with the simulation box axes as in Fig. 1.
  • Boundaries / frozen regions: Periodic boundary conditions in the two lateral directions; bottom two Si layers (0.27 nm) fixed; 29 layers (3.85 nm) below the reactive region act as a heat bath with temperature rescaling every timestep; top 10 layers (1.33 nm) unconstrained for chemistry.
  • Timestep: 0.5 fs (stated as the rescaling interval for the bath layer control).
  • Thermostat / thermal control: Berendsen thermostat couples the heat-bath region and the topmost free layer during substrate equilibration at the reaction temperatures (300, 500, 700, 900 K are explicitly mentioned in the Methods excerpt).
  • Temperature / statistics: 1120 independent single-NO events sampled for kinetics across 300–1000 K (abstract); representative trajectory lengths for classification snapshots are on the order of ~1–5 ps at stated temperatures (see figure captions in pdf_path).
  • Sequential dosing: After 10 ps equilibration at the reaction temperature, NO is added every 20 ps in the consecutive-reaction growth studies (Methods text).
  • Initial NO placement / energy: NO starts 0.35 nm above the surface; initial NO translational energy corresponds to 300 K thermal energy (~38 meV as written).
  • Ensemble: The manuscript describes constant-temperature control via rescaling/thermostatting of bath regions while keeping dynamical Newtonian evolution in the reactive zone—read pdf_path for the precise NVT-vs-NVE phrasing used by the authors.
  • Barostat / pressure: N/A — fixed-volume slab workflow; no NPT pressure targeting is described in the excerpted Methods text.
  • Electric field: N/A — not used.
  • Replica / enhanced sampling: N/A — not used.

Force-field training

N/A — not a parameterization paper; the manuscript applies an established ReaxFF description for NO + Si(001) chemistry.

Static QM / DFT (NEB benchmarks)

  • Method: Nudged elastic band (NEB) calculations compare intra-dimer vs inter-dimer NO bridge precursors on the way to dissociation.
  • Authored kinetic conclusion: Even if an intra-dimer complex can be lower in energy, the inter-dimer bridge pathway is argued to dominate because its activation barrier is lower in the NEB results (abstract-level statement).

Findings

  • Validation: The authors state that the statistical MD kinetics are consistent with prior experimental and theoretical literature on NO chemistry on Si, supporting use of the reactive MD approach for this interface problem.
  • Pathway selection: NEB results motivate favoring an inter-dimer bridge intermediate for NO dissociation despite thermodynamic preferences for some intra-dimer geometries—i.e., kinetic control of the dominant channel in their model.
  • Growth phenomenology: Low-temperature sequential dosing shows self-limiting incorporation consistent with STM-oriented expectations discussed in the paper, while higher temperature broadens N/O depth distributions—framed as guidance for oxynitridation thermal budgets in gate-stack processing.

Limitations

Numerical barriers and prefactors must be taken from the article and SI; ReaxFF omits explicit electronic excitations beyond the empirical bond-order form.

Relevance to group

Silicon oxynitridation and gate-stack processing are classic ReaxFF application areas for semiconductor oxidation.

Citations and evidence anchors