A new transferable interatomic potential for molecular dynamics simulations of borosilicate glasses
Evidence and attribution¶
Authority of statements
Prose below summarizes the borosilicate potential paper identified by doi, title, and pdf_path.
Summary¶
Borosilicate glasses are technologically important for display panels, laboratory glassware, and nuclear waste immobilization, but atomistic modeling requires potentials that remain transferable across boron-rich and silica-rich compositions while reproducing boron coordination (trigonal BO\(_3\) versus tetrahedral BO\(_4\)), mass density, medium-range order, and transport-related properties such as shear viscosity. This article introduces a fixed-parameter empirical interatomic potential intended to span silicate-to-borate chemistries without composition-dependent refitting of energy parameters. The authors validate the model using LAMMPS molecular dynamics across a series of glass compositions, comparing simulated structural and mechanical properties to experimental trends and to the needs of industrial and nuclear-glass design workflows.
Prior classical potentials for borosilicates often struggled because boron speciation is composition- and thermal-history dependent: network modifiers can convert BO\(_3\) to BO\(_4\) or create non-bridging oxygens on silicon, shifting ring statistics and viscosity in ways that are difficult to capture with composition-specific refits. The authors position their model as a single global parameter set that still tracks these crossovers across a broad composition grid.
Methods¶
Empirical potential (non-reactive classical MD)¶
The authors introduce a two-body Buckingham + fixed partial charge form with a single global parameter set spanning borosilicate compositions from borate-rich to silicate-rich end members (pair parameters tabulated in the article; Buckingham catastrophe safeguards discussed where needed). This is a fixed-bond-topology classical model (not ReaxFF): it targets composition transferability without composition-dependent charge refits of the kind criticized for earlier borosilicate potentials.
MD application (LAMMPS, melt–quench)¶
Simulations use LAMMPS with 11 Å cutoffs for both short-range and Coulomb interactions and PPPM electrostatics (accuracy ~10⁻⁵ as stated). The integration timestep is 1.0 fs. ~3000 atoms are placed randomly in a cubic cell without overlaps, melted at 3000 K for 10 ps in NVT followed by 100 ps at zero pressure in NPT, then linearly cooled to 300 K at 1 K/ps under NPT. Each glass is relaxed at 300 K, zero pressure, for 100 ps, followed by 100 ps NVT production from which 100 snapshots spaced 1 ps apart are averaged. Exploding high-temperature NPT trials trigger an alternate path: 100 ps NVT melt at 3000 K, NVT cool to 300 K, then restart the standard melt–quench. Glassy-state RDFs, coordination, density, and Green–Kubo viscosity use those trajectories. Boundaries / periodicity: all melt–quench and production segments use three-dimensional periodic boundary conditions (PBC) on the cubic simulation cell, as standard for bulk glass MD in LAMMPS.
-
Thermostat / barostat brands: N/A — not extracted into plain text by the pypdf pass used for curation; the J. Non-Cryst. Solids PDF (
papers/ReaxFF_others/Wang_Mauro_Bauchy_Borosilicate_pairpotential.pdf) should be consulted for any explicit integrator or thermostat names beyond the NVT/NPT staging above. -
Electric field / metadynamics: N/A — not used (bulk melt–quench MD only).
Observables¶
BO\(_3\)/BO\(_4\) fractions, mass density, pair and ring statistics, and shear viscosity from equilibrium MD (stress autocorrelation / Green–Kubo protocol per the article) are compared to experimental benchmarks from Smedskjaer & Mauro datasets and related literature compositions (Table 1 naming such as 75B … 0B in the PDF).
Findings¶
Across the benchmarked borosilicate set, the potential reproduces compositional trends in boron speciation, density, short- and medium-range structure, and viscosity in line with the experimental and simulation targets discussed in the paper. The work supports using a single transferable parameterization for structure–property exploration in complex borosilicate formulations, subject to the usual caveats of classical fixed-bond-order models.
Practical takeaway. For nuclear waste glass design workflows, having viscosity and speciation trends tracked together across modifier content is often more useful than reproducing a single end-member silicate perfectly; the paper’s validation emphasizes composition sweeps rather than a single benchmark composition.
Limitations¶
The potential does not describe bond-making and bond-breaking chemistry; extrapolation outside the fitted composition grid or to extreme thermal histories should be checked case by case. Oxide and alkali diffusion in highly modified glasses may need additional validation.