Integrated atomistic chemical imaging and reactive force field molecular dynamic simulations on silicon oxidation
Summary¶
Silicon microelectronics depend on thermally grown SiO₂, but the Si/SiO₂ interface is not atomically abrupt: a suboxide (SiOₓ) transition region carries gradients in stoichiometry and bonding that influence transport, reliability, and variability. This Applied Physics Letters article couples atom probe tomography (APT)—a chemical imaging modality with near-atomic depth resolution in favorable cases—with ReaxFF reactive molecular dynamics to compare experimental and simulated evolution of suboxide character as oxidation temperature changes. The motivating claim is that integrated experiment and atomistic oxidation modeling can converge on the same trends for interfacial SiOₓ fraction and oxygen ingress, where simpler continuum oxidation models may wash out chemical detail.
Methods¶
Experiment (APT)¶
Si needles are thermally oxidized at 1–4 Torr at 383 K and 548 K. Laser-assisted APT (LEAP 3000 X Si, 532 nm, ~10 ps pulses, ~0.9–1.0 nJ, 150 kHz) reconstructs Si/SiO₂ interfacial chemistry and intermixing.
MD application (atomistic dynamics)¶
Reactive molecular dynamics with ReaxFF (Si/O parameterization attributed to Buehler et al.; extended tables in SI, Appl. Phys. Lett. 106, 011602 (2015)) starts from a Si(100) 2×1 slab cell 21.7 × 21.7 × 27.1 ų (LAMMPS-style MD setup as standard in the cited ReaxFF Si/O literature). NVT Berendsen thermostat equilibration at 300 K and 700 K (20 ps, 0.1 ps damping) precedes 10 ps NVE relaxation. 3D PBC applies to the oxidation supercell. Integrator timestep: N/A — not stated in the short APL paragraph here; see SI if a numeric dt is required. NPT stress control: N/A — not reported as a primary knob. Profiles of suboxide/diffusion are compared to APT proximity histograms; the article stresses qualitative agreement given experiment vs nanosecond MD time-scale mismatch.
Force-field training¶
N/A — applies an existing ReaxFF Si/O parameterization with literature precedent; not a new fit paper.
Static QM / DFT¶
N/A — DFT is cited comparatively in the surrounding literature discussion, not as an on-the-fly engine in the reported MD.
Findings¶
APT and ReaxFF both show higher SiOₓ relative to SiO₂ and broader oxygen ingress at 548 K than at 383 K, in qualitative agreement though absolute widths and times differ—i.e., oxidation and intermixing trends align between experiment and simulation despite different accessible thickness scales. Proximity histograms highlight suboxide in an interfacial region II, matching simulation-resolved mixed-valence O environments in the figures/text. Sensitivity to oxidation temperature is therefore captured at the qualitative level the authors claim. Limitations include APT reconstruction biases and MD timescales far shorter than furnace oxidation; PDF/SI remain authoritative for numerical interface widths.
Limitations¶
APT reconstruction can bias interface chemistry; MD captures only short-time oxidation kinetics, so absolute interface widths should be treated as qualitatively comparable between experiment and simulation (as the authors discuss).
Relevance to group¶
Direct van Duin-authored ReaxFF interface to APT, emphasizing validation at atomic chemical resolution.