Development of the ReaxFF Reactive Force Field for Inherent Point Defects in the Si/Silica System
Evidence and attribution¶
Authority of statements
Prose sections below (Summary, Methods, Findings, etc.) are curated summaries of the publication identified by doi, title, and pdf_path in the front matter above. They are not new primary claims by this wiki.
For definitive numerical values, reaction schemes, and interpretations, use the peer-reviewed article (and optional records under normalized/papers/ when present)—not this page alone.
Summary¶
Si/O/H ReaxFF parameters are redeveloped to treat O interstitials and migration in bulk Si and oxidation chemistry at a-SiO2/Si interfaces, emphasizing point-defect behavior omitted by earlier fits. The new “ReaxFF\(_\text{present}\)” reproduces the bond-centered hopping pathway for O in Si with a barrier near 65 kcal/mol, closer to experiment/DFT than prior SiOH(2010) behavior that spuriously allowed deep O diffusion at 300 K. Annealing simulations of prepared a-SiO2 on Si recover expected interfacial O transport relative to literature, including fixing the low-temperature spurious diffusion issue.
Methods¶
2 — Force-field training. Starting from Fogarty et al. ReaxFF\(_\text{SiOH(2010)}\), the authors add DFT-based training on O-related defects in a-SiO\(_2\)/Si and the O-migration path in bulk Si, then re-fit selected Si–Si, Si–O, and valence angle parameters with a single-parameter least-squares-style search (eq. 1 in the J. Phys. Chem. A paper) until ReaxFF and QM/experiment targets compromise under σ-weighted errors.
1 — MD application. Molecular dynamics is performed with ReaxFF\(_\text{present}\) (reactive MD; N/A for a unique public code string in the short p1–2 excerpt—see the J. Phys. Chem. A PDF / SI for the MD engine if required). ReaxFF\(_\text{present}\) trajectories use PBC and the velocity Verlet integrator with a 0.5 fs time step in the a-SiO\(_2\) annealing protocol quoted in §2 of the PDF; Berendsen-style thermal and pressure damping are set to 100 fs and 5000 fs in that glass-preparation block. The a-SiO\(_2\) melt/quench sequence alternates high-T NVT and NPT ramps at P = 0 (see the 7000 K → 300 K two- step anneal with 50–500 ps holds and final NPT relaxation to a ~2.21 g cm⁻³ glass in the Methods). O diffusion in bulk Si is studied with NVT MD on 800–2400 K targets in the Results text; interface anneals use the prepared a-SiO\(_2\)/Si cells described in the same article (Figure 6 ff.). Two-phase melting- point MD for validation uses NVT/NPT stages with 2744- atom phases as narrated in the text. Shear / shock: N/A. External uniform electric field: N/A. Replica / metadynamics: N/A — direct NVT/NPT sampling in the reported stages. Barostat: NPT at zero mean pressure in the glass-formation legs; NVT in selected high-T holds without stress servocontrol (per the protocol paragraphs). MD engine (package name): N/A in the short indexed excerpt; treat as ReaxFF MD with the stated ensembles and fs-level controls in the VOR PDF.
3 — Static DFT as a sole Method block: N/A — DFT trains the ReaxFF; the reported evidence is largely ReaxFF MD plus QM targets**.
Findings¶
O in bulk Si hops BC→BC in the (110) plane with a split-type saddle; ReaxFF\(_\text{present}\) predicts a ~64.8 kcal mol⁻¹ barrier in line with cited DFT/experiment. NVT diffusion runs (800–2400 K) show onset of appreciable O motion above ~1400 K with BC-only hops at high T and D(T) trends consistent with published data in the broad sense of the article. a-SiO\(_2\) density and RDF/coordination metrics for the new FF match experiments more closely than SiOH(2010) on the comparisons plotted in Figs. 10–11/Table 4. a-SiO\(_2\)/Si stack anneals on ReaxFF\(_\text{present}\) reproduce O transport at the interface without the unphysical room-temperature deep O penetration seen for SiOH(2010) in the same class of setup (the low-T spurious-diffusion regression the authors target).
Limitations¶
Focus on Si/SiO2 defect and interface phenomena; parameters are not automatically transferable to unrelated chemistries (e.g., alkali-containing silicates) without testing. Alkali contamination, strain from epitaxial Si, and high-κ dielectric adjacency are named in the broader literature as contexts requiring additional validation beyond the pure Si/O/H scope emphasized in the J. Phys. Chem. A article. Stress-assisted diffusion, H charging, and radiation damage in power devices may require additional training data not central to the bulk Si / a-SiO\(_2\) defect benchmarks highlighted there.
Relevance to group¶
Foundational ReaxFF for silicon oxidation—central to much group and field work on gate stacks and silica.
Citations and evidence anchors¶
papers/Nayir_JPC_C_SiOx_2019.pdf — main text (ReaxFF\(_\text{present}\) training, a-SiO\(_2\) anneal protocol, O diffusion and interface tests). https://doi.org/10.1021/acs.jpca.9b01481
Supporting information PDF: 2019nayir-venue-paper.