Kinetics of Silane Decomposition in High-Pressure Confined Chemical Vapor Deposition of Hydrogenated Amorphous Silicon

Abstract

Confined HPCVD measures silane pyrolysis kinetics (20–33 MPa, 5.9 µm capillary); ReaxFF NVT MD on clean vs H-terminated c-Si shows heterogeneous decomposition and supports H-desorption-limited growth.

Summary¶

Experiments: Silane pyrolysis in confined high-pressure CVD at 20–33 MPa in a 5.9 µm inner-diameter microcapillary yields first-order kinetics in silane with Ea = 53.7 ± 2.9 kcal/mol and A = 1.5×10¹⁰ m/s, aligning activation energies with H desorption from crystalline silicon surfaces—suggesting H desorption limits film growth. Simulations: ReaxFF MD of silane interacting with clean vs H-passivated c-Si slabs (15 SiH₄ molecules, 7 MPa, 400 °C equivalent conditions in a 22×22×50 Å cell with a 256-atom Si slab; minimization and slab prep per SI). NVT production runs monitor unreacted silane. No homogeneous gas-phase reactions appear in the ReaxFF trajectories; clean surfaces consume silane into silicon hydrides within ~10 ps, while H-terminated surfaces show no reaction in the same protocol—supporting heterogeneous control and dangling-bond availability as rate-limiting in confined HPCVD.

Microcapillary confinement raises pressure without large reactor volume, letting kinetics be measured where gas-phase chain growth is suppressed—matching the simulation emphasis on surface-limited chemistry.

Methods¶

Force-field training / fitting. ReaxFF parametrizations for Si/H chemistry (including silane and silicon hydrides) are used as published; additional force-field testing notes appear in the Supporting Information. No new QM refit is reported in the main article text summarized here.

MD application (atomistic dynamics). Engine / code: ReaxFF molecular dynamics as implemented in the authors’ workflow; specific MD package N/A — not named in the main-text proof PDF excerpt used here (confirm software/version in Supporting Information if required for reproduction). System size & composition: 15 silane molecules with a 256-atom c-Si slab in a 22 × 22 × 50 Å periodic cell, comparing clean vs hydrogen-passivated surfaces (Figure 4). Boundaries / periodicity: Three-dimensional periodic supercell (standard for this slab–gas setup). Target conditions: ~400 °C (~673 K) and ~7 MPa silane pressure as stated for the reactive cell. Ensemble: canonical ensemble (NVT-class) production trajectories after separate NVT minimization/pretreatment of the slab and gas regions described in the SI. Timestep / femtosecond integration: explicit Δt N/A — not printed in the main proof excerpt; SI lists minimization and slab-preparation settings. Duration / stages: separate NVT pretreatments of surface and gas regions (SI), then combined runs long enough to resolve ~10 ps clean-surface consumption kinetics reported in Figure 4; total production nanoseconds N/A — confirm in SI tables. Thermostat: coupling details for the canonical ensemble runs are N/A — beyond the ensemble label in the main proof excerpt—see SI. Barostat: N/A — constant-volume canonical runs without hydrostatic barostat control in the excerpted protocol. Pressure: initial/target silane pressure ~7 MPa (as stated); thermodynamic pressure evolves during reactive events but is not tabulated further on this page. Electric field: N/A — not applied. Enhanced sampling: N/A — brute-force MD.

Static QM / DFT. N/A — DFT is not the time integrator for the ReaxFF decomposition trajectories summarized here.

Review / non-simulation framing. Confined high-pressure CVD experiments (20–33 MPa, 5.9 µm capillary) provide Arrhenius parameters (Ea = 53.7 ± 2.9 kcal mol⁻¹, A = 1.5×10¹⁰ m s⁻¹ as reported) compared against the heterogeneous mechanism suggested by simulation. Corpus note: pdf_path points to a proof PDF.

Findings¶

Outcomes and mechanisms. Confined HPCVD measurements give first-order silane kinetics with Ea = 53.7 ± 2.9 kcal/mol and A = 1.5×10¹⁰ m/s, with Ea in the range reported for H desorption from c-Si surfaces—supporting H-desorption-limited growth. ReaxFF trajectories on clean c-Si show rapid heterogeneous consumption of SiH₄ into silicon hydrides (~10 ps scale in Figure 4), whereas H-terminated surfaces show no reaction under the same protocol, implicating dangling-bond availability after H removal as rate-limiting. No homogeneous gas-phase reactions are observed in those simulations.

Comparisons. The authors relate the activation energy story to prior CVD literature on heterogeneous vs homogeneous pathways under conditions where silane adsorption outcompetes oligomer adsorption.

Sensitivity and design levers. Pressure (20–33 MPa experiments vs ~7 MPa simulation cell), temperature (400 °C-class conditions in simulation), and surface termination (clean vs H-passivated) are the primary levers discussed.

Limitations and outlook (as authored). Proof PDF; full reactor dimensions, flow, and SI MD settings exceed this short summary—use the article/SI for reproduction.

Corpus honesty. Simulation cell parameters are abbreviated here; pdf_path is a proof duplicate—confirm final values on the version-of-record when available.

Limitations¶

Proof PDF; ReaxFF system size and timescales are illustrative of mechanism, not full reactor chemistry.

Relevance to group¶

Adri C. T. van Duin is a co-author; connects ReaxFF to semiconductor CVD kinetics.

Citations and evidence anchors¶

DOI: 10.1021/acs.iecr.7b03515

reaxff-family

Corpus file is a proof PDF.