Skip to content

Structural features of sodium silicate glasses from reactive force field–based molecular dynamics simulations

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

This Journal of the American Ceramic Society (JACS in the sense of the American Ceramic Society, not the American Chemical Society) article benchmarks reactive molecular dynamics of sodium silicate glasses using ReaxFF against a widely used partial-charge pairwise potential and against experimental structure probes. The motivation is twofold: sodium silicate glass is an industrially important archetype for multicomponent silicates, and reactive potentials are increasingly needed to study corrosion and environmental interactions where bond rearrangement with water matters. Prior work with nonreactive potentials produced detailed short- and medium-range structural comparisons for many compositions, but capturing reaction chemistry with water requires potentials that treat bond breaking and formation self-consistently. Lu Deng, Adri C. T. van Duin, Jincheng Du, and AGC-affiliated coauthors therefore evaluate whether refined ReaxFF parameters can reproduce key glass structural signatures before deploying them in water–glass studies.

Methods

1 — MD application (melt, quench, annealing; LAMMPS). Engine: LAMMPS runs classical molecular dynamics for Na\(_2\)O–SiO\(_2\) supercells (thousands of atoms; 3D PBC / periodic boundaries in the bulk glass setup as standard for these packings). Nosé–Hoover thermostat and barostat for NPT where applicable; Ewald summation; 10 Å cutoffs for short-range pairs (as stated). Teter (rigid-ion) melt–quench (initial structures): minimize at 0 K, relax at 300 K, heat to 3500 K and hold 100 ps (NPT), then cool to 300 K at 5 K ps⁻¹ (NPT); time step 1 fs for this Teter preparation block. ReaxFF (Yu17 / Hahn18 families) uses Teter-prepared glasses as inputs; NPT anneal over Tₐ = 300, 600, …, 3000 K (11 temperatures), 100 ps at each Tₐ, then 5 K ps⁻¹ cool to 300 K; ~80 ps NPT at 300 K; 10 ps NVT; 10 ps NVE with structural output from the last 2 ps (50 fs output interval). Barostat: NPT through anneal and equilibration; NVE for the final output segment. Replica / umbrella / metadynamics: N/A. Electric field: N/A. Discussion compares cost: ReaxFF ~35× slower than Teter for the same simulated time under their setup, with ReaxFF using 0.1 fs vs 1 fs (Teter) per step. Pure ReaxFF melt–quench is noted as feasible but more expensive and can raise coordination defects (SI Table S3).

2 — Force-field training in this work. N/A — the study adopts published Na–Si–O–H ReaxFF parameter lines (Yu17 and Hahn18 / Hahn1831 in the text) and does not report a new in-paper CMA/MCFF fit.

3 — Static DFT in this work. N/A — validation is ReaxFF and Teter MD against diffraction, NMR motifs, and experiment.

Findings

  • ReaxFF lines vs Teter. The Hahn18-family ReaxFF sets (including (OO)-related treatments, labeled as Hahn18-OO in the text) better match Qₙ, PDFs, and F(Q) against the Teter nonreactive baseline in their protocol than the Yu17 set; a pure ReaxFF melt–quench leaves more coordination defects than a Teter-seeded glass followed by the reported Hahn-line re-anneal (Table S3).
  • Experiment. Neutron- and X-ray-broadened F(Q), PDFs, and coordination/Qₙ are compared to experimental and literature data; F(Q) peak positions are largely reproduced, while the Discussion notes amplitudes and some secondary features (shoulder/split peaks) where the rigid-ion or ReaxFF models differ.
  • Practical route. The authors recommend a mixed workflow: efficient Teter melt–quench to build initial glasses, then Hahn-ReaxFF re-annealing near the glassy regime to obtain models suitable for subsequent water–glass ReaxFF studies, citing ~35× wall-time cost and 0.1 fs vs 1 fs timesteps.
  • Outlook. Polarizable-ion models and further ReaxFF improvements are mentioned where F(Q), Qₙ, and related metrics remain imperfect in their tests (Discussion/Conclusions).

Limitations

Specific compositions studied; ReaxFF refinements evolve—later parameter sets may supersede details here.

Relevance to group

Collaboration with UNT/AGC on industrially relevant silicate glasses; van Duin co-authorship on validation-centric ReaxFF for soda–silica networks.

Citations and evidence anchors

papers/Deng_NaSiOx_JaCERS_2019.pdf — abstract and structural comparison sections. https://doi.org/10.1111/jace.16837

Reader notes (navigation)