Skip to content

Mechanical size effects of amorphous polymer-derived ceramics at the nanoscale: experiments and ReaxFF simulations

Summary

Silicon oxycarbide (SiOC) polymer-derived ceramic (PDC) micro- and nanofibers are fabricated without fillers; reducing fiber diameter from about 1.1 um to about 630 nm is reported to increase tensile strength from about 1 GPa to about 3.3 GPa, beyond what the Griffith scaling picture alone would suggest, motivating a defect-size picture tied to pyrolysis degassing. A Si/O/C/H/N ReaxFF parameterization is developed and used to model "flawless" PDC structures so that simulated elastic response can be compared with experiment and failure can be tied to bond-breaking chemistry (C-C failure as a limiting mechanism in the simulations).

Methods

Experiments (SiOC PDC fibers). Polymer-derived ceramic (PDC) SiOC micro- and nanofibers are fabricated without fillers and tensile-tested on MEMS fixtures. The manuscript reports a strong diameter dependence of strength in the examined window (abstract/introduction cites testing down to about 610 nm diameter and compares against Griffith-type scaling expectations). Nominal strain rate in fiber tests is on the order of ~60 με s⁻¹ as stated in the article.

Reactive force-field development (Si/O/C/H/N). The authors develop a new ReaxFF parameter set for Si/O/C/H/N by combining and re-optimizing the CHON-2017_weak parameterization with training sets for Si/O/H (Pitman et al.) and C/O/Si (Newsome et al.), targeting both polymer densities/morphology and ceramic stress response (see Table S1/ESI discussion in the paper). QM reference data used in the fit are summarized via energy-profile comparisons in the ESI (e.g., Fig. S1 referenced in the main text).

MD application (virtual mechanical testing, LAMMPS + ReaxFF). Engine / code: LAMMPS is used for virtual mechanical testing with the developed ReaxFF potential (Nanoscale computational details). System size & composition: simulations start from a cross-linked SiOC geometry adapted from prior work (Gao et al.), energy-minimized with the new field to yield a cubic cell about ~2 nm per side. Boundaries / periodicity: three-dimensional periodic boundaries are used in the MD workflow described in the article. Ensemble / pressure: tensile modulus runs use an NPT ensemble with a cuboidal constraint so Poisson contraction is represented (four faces parallel to the loading axis are free to move accordingly). Timestep: 0.1 fs for the reported modulus / Poisson ratio virtual tests. Thermostat: thermostat damping parameter reported as 10 fs alongside the NPT modulus protocol. Strain rate / loading: tensile modulus is evaluated across strain rates from about 10¹⁰ to 10¹² s⁻¹; modulus and Poisson ratio fits use the linear regime up to about ~2% strain at 300 K, with additional temperature-dependent tests reported at 500 K, 800 K, and 1000 K at 10¹² s⁻¹. Electric fields / enhanced sampling: N/A — not part of the mechanical testing narrative summarized above. Duration / stages: the workflow includes energy minimization, NPT equilibration, and strain-controlled production segments for modulus and temperature sweeps; cumulative trajectory lengths in ps/ns are tabulated in the Nanoscale Methods/SI rather than duplicated line-by-line here.

Findings

Outcomes / mechanisms (experiments). Decreasing fiber diameter from about 1.1 µm to about 630 nm is reported to nearly triple tensile strength (from about ~1 GPa to about ~3.3 GPa in the abstract framing), with a stronger-than-Griffith diameter scaling (the article discusses a scaling closer to σ_f ∝ d^−1.68 vs a d^−0.5 Griffith expectation when flaws are capped by diameter). The interpretation emphasized in the paper is size-dependent flaw populations during pyrolysis/degassing (voids/microcracks), where thinner fibers provide shorter diffusion paths for volatile escape.

Outcomes / mechanisms (ReaxFF “flawless” models). MD reproduces an experimental elastic modulus near 107 ± 16 GPa with simulation averages near 102 ± 16 GPa (with per-realization spreads quoted in the article) and reports Poisson’s ratio near 0.338 ± 0.039 and shear modulus near 38 ± 8 GPa from the dense models. The simulations quote an upper-bound strength near ~8.5 GPa for an idealized void-free structure, with C–C bond failure highlighted as the limiting rupture channel in the RDF/bond-break analysis.

Comparisons. Simulated moduli are described as insignificantly dependent on the ultrahigh strain rates used (10¹¹–10¹³ s⁻¹ window in the modulus discussion), while still matching the experimental modulus within quoted uncertainty.

Sensitivity / design levers. Temperature is explored in MD for mechanical softening: a modest modulus drop when heating from 300 K toward 500 K, but a large drop by 800 K (article quotes about 45% reduction by 800 K), with yield strain also shifting with temperature.

Limitations / corpus honesty. Real fibers contain disorder and voids that are deliberately removed in the “flawless” limit; ReaxFF accuracy is bounded by training scope. Ultra-high strain rates in MD are not equivalent to MEMS fiber strain rates, so strength numbers are model-relative upper bounds unless mapped via the authors’ extrapolation discussion.

Limitations

Real PDCs contain disorder, voids, and processing variability not captured by idealized cells; ReaxFF remains a classical reactive approximation with accuracy bounded by its training scope and by the ultrahigh strain-rate MD protocol used for modulus/strength probing.

Relevance to group

Co-authored by Adri C. T. van Duin; couples experimental PDC processing and mechanics with ReaxFF development for Si-O-C-H-N ceramics.

Citations and evidence anchors

https://doi.org/10.1039/c9nr00958b