Effects of water on the mechanical properties of silica glass using molecular dynamics
Summary¶
Silicate glasses can contain water either as molecular H₂O trapped in the network or as hydroxyls formed by hydrolysis of Si–O–Si bridges; experiments show both species alter mechanical response, but isolating their roles is challenging when both are present. Mei et al. use ReaxFF molecular dynamics in LAMMPS with the Yeon et al. Si/O/H parameterization—developed for load-driven silica hydrolysis—to prepare bulk silica glass samples with controlled in-network molecular water versus hydroxyl content, then subject them to mechanical tests that resolve elastic moduli, strength, fracture toughness, and network metrics (Si–O distances, Si–O–Si angles). The Acta Materialia study is explicitly about decoupling speciation: molecular water can plasticize without fully depolymerizing the network, whereas hydroxyls cut bridging bonds and reduce connectivity more directly.
Methods¶
1 — MD application (atomistic dynamics). Molecular dynamics is run in LAMMPS with the Yeon et al. ReaxFF Si/O/H parameterization developed for load-driven silica hydrolysis; structures are visualized in OVITO (per §2 of Acta Materialia). System / composition: the compact-tension (CT) silica block in Fig. 1a has in-plane extents ~22 nm and ~2.3 nm thickness after replication, with O(10⁴)–O(10⁵) Si/O atoms depending on defect/crack content and replication factors detailed in the Acta article. Pure silica glass is prepared and densified with a 0.5 fs integration timestep; subsequent mechanical and reactive segments use 0.25 fs as stated in the Simulation details section. Thermostat (NVT): the paper uses NVT integration for the quoted anneal and deformation stages; the family and damping (Berendsen-style or Nosé–Hoover) are given in the VOR PDF (see §2; not repeated here in full). After controlled introduction of molecular H₂O vs hydroxyl speciation in the network, samples are relaxed at 600 K in NVT for 1 ns (see the 600 K NVT anneal and glass density ~2.07 g cm⁻³ in the paper). Mechanical testing includes compact tension (CT) and uniaxial-tension-derived quantities (e.g. Young’s modulus from ~1% true strain linear fits; critical tensile stresses at a 10⁹ s⁻¹ strain rate in the CT geometry as tabulated in the article). PBC and plane strain / loading directions follow the sketches in Figs. 1–3; see the PDF for sample dimensions and water content labels (samples 1–3 and variants in Tables). NVT-centric aging and NVT mechanical pulls are used for the quoted β-relaxed / deformed states; any NPT holds (if present for preparation) should be read from the full Methods in the PDF—barostat-led mean pressure targets are not the focus of the abstract-level summary here. Static external electric field: N/A. Replica / metadynamics: N/A**.
2 — Force-field training: N/A in this paper’s core—the authors adopt the published Yeon ReaxFF fit to Si/O/H load-dependent chemistry rather than re-optimizing parameters de novo.
3 — Static QM / DFT: N/A as a first-principles mechanical protocol; ReaxFF is the active potential in the reported trajectories.
Findings¶
Molecular water can increase Young’s modulus at low water content, whereas hydroxyls decrease modulus under the surveyed compositions. Hydroxyls reduce strength and fracture toughness by breaking network connectivity; molecular water weakens the glass by driving the silica network toward strained configurations before external stress is applied. In tension, stress–strain plateaus in molecular-water-containing samples coincide with Si–O bond rupture followed by silanol formation, and molecular water lowers the critical stress where the plateau begins relative to drier baselines in the model. The authors relate these trends to microstructural indicators—Si–O bond lengths and Si–O–Si angles—so that mechanical softening is tied to measurable network distortions rather than to a black-box “water effect.”
Readers building retrieval clusters should pair this page with fracture-focused silica entries (for example Rimsza JGR subcritical fracture) when questions span bulk modulus versus crack-tip chemistry.
Limitations¶
System sizes and strain rates follow MD constraints; quantitative agreement with macroscopic fracture tests requires upscaling. Operators curating brittle fracture benchmarks should pair this entry with continuum corrosion models when translating to component lifetimes. Water contents explored in the article should be interpreted relative to the glass synthesis pathway assumed in the simulations, not as universal service environments.
Relevance to group¶
Demonstrates ReaxFF-based mechanical testing of hydrated silica with van Duin-parameter lineage tied to prior silica–water reactivity work.