Reactive Potential for the Simulation of Active Brazing of a Ceramic–Metal Interface
Summary¶
The paper develops a seven-element Fe/Ni/Co/Ag/Al/Zr/O ReaxFF parameterization aimed at liquid metal–metal and liquid metal–alumina interfaces relevant to active brazing (Ag–Zr filler on Kovar vs α-Al\(_2\)O\(_3\)), emphasizing surface tension, heats of formation, and interfacial reactions connected to run-out in braze joints.
Methods¶
- QM training data: Amsterdam Modeling Suite (AMS) 2021.102 — periodic BAND with PBE-D3, DZ basis, small frozen core, scalar relativity, “good” numerical quality; cluster ADF calculations preferentially PBE-D3, TZ basis, “very good” quality; BLYP-D3 or M06 used where PBE-D3 failed to converge (noted for some Fe/Co angle scans). H/O started from the aqueous ReaxFF branch; Zr–O training augmented with OQMD data; training included metals, oxides, alloys, and mixed-metal oxides. Zr/O parameters transferred from combustion to aqueous branches were refit to better capture suboxide energetics.
- ReaxFF optimization: Standalone ReaxFF trainer (Penn State Materials Computation Center distribution referenced in the article) with parabolic interpolation/extrapolation workflow; initial parameters merged from prior ReaxFF lines as described.
- Melting and surface tension (LAMMPS): Melting temperatures via solid–liquid coexistence workflows (0.5 fs timestep, Langevin thermostat for initial solid equilibration; Nosé–Hoover coexistence segments, PTM structure analysis in OVITO with criteria stated in the paper). Surface tensions from liquid slabs (~4000 atoms, NVT, 1.5 ns production, stress sampling).
- Diffusion: MSD slopes from ~100 ps liquid simulations (~9826-atom cells) for Ag and dilute Al/Zr/O in Ag.
- Interface MD: Molecular dynamics in LAMMPS; Al\(_2\)O\(_3\) vs Ag and Ag–Zr liquids at 1600 K for 500 ps; Kovar(111) vs Ag, Ag–10% Zr, Ag–10% Al slabs (924 liquid on 1728-atom substrate) at 1600 K for 600 ps with 3D PBC periodic slabs; additional 600 ps after Al addition. 1 atm-like NPT may appear in other validation blocks, but the quoted interfacial runs are typically NVT at high T; N/A — isotropic pressure control in every short interfacial segment. N/A — static external electric field; N/A — metadynamics; N/A — umbrella in this brazing interfacial protocol summary.
Findings¶
- The force field reproduces many heats of formation and equations of state reasonably across metals/oxides; liquid Ag/Al alloy surface tensions track experiments with noted trade-offs in parameterization.
- Melting points are qualitatively sensible for the braze-relevant window, but pure Ag is ~350 K above experiment in this parameterization—nonetheless below oxides/Kovar such that molten filler vs solid substrates can still be staged.
- Ag–Al\(_2\)O\(_3\) simulations show Zr scavenging oxygen and forming Zr-rich suboxide clusters; pure Ag also shows O transfer that the authors flag as a limitation of the model.
- Kovar interfaces show Zr-driven intermetallic formation and complex segregation; Ag–10% Al does not show strong Al excess at the Kovar interface in these short trajectories, in tension with some EAM literature but aligned with other cited braze studies.
Limitations¶
The authors highlight elevated Ag melting, problematic Ag–alumina behavior including O scavenging by pure Ag, and Ag–alumina interface instability as weaknesses; quantitative comparison to experiment for suboxide stoichiometry may require grand-canonical or fixed-μ protocols not used here.
Relevance to group¶
Rothchild, Diana M. van Duin, Kowalik, Chandross, van Duin: ReaxFF for active brazing Ag–Zr fillers on alumina/Kovar-class interfaces.