Skip to content

eReaxFF force field development for BaZr0.8Y0.2O3-delta solid oxide electrolysis cells applications

Summary

An eReaxFF parametrization for BaZr0.8Y0.2O3-delta (BZY20) is fitted to DFT and constrained-DFT data for oxygen vacancies, water chemistry, and excess electrons, then applied in zero- and finite-bias-style MD to probe water splitting, proton motion, and hydrogen evolution on BZY20 surfaces.

Methods

QM reference: DFT targets include Y-doping site preference in bulk BZY20, oxygen-vacancy formation energies in bulk and on (100) and (110) slabs with Ba-O versus Zr-Y-O terminations and variable surface vacancy concentrations (e.g. 12.5% and 25% for supercell convenience), equations of state, H2O adsorption and splitting energies, H2 formation energies, and C-DFT energies for an excess electron localized on lattice oxygens in isolated clusters (parameterized electron hopping barriers, e.g. ~16.87 kcal/mol O1 to O10 and ~9.85 kcal/mol in a hydrogenated cluster scenario per SI). eReaxFF optimization targets qualitative trends for vacancy formation energies while emphasizing Y-site energies, bond lengths, and EOS agreement.

Reactive MD: Bond-scan biasing potentials assess barriers for adsorption, splitting, and H2 evolution. Large-scale examples: Zr-Y-O (100) slab with 12.5% surface vacancies, periodic box ~42.95 x 42.90 x 50.00 Angstrom, steam represented by repeated H2O additions (10 H2O every 200 ps), 1000 K, NVT, ~3000 atoms, 200 ps segments run in ~48 h CPU on four CPUs cited for one case; zero-bias runs used Amsterdam Modeling Suite (AMS). ACKS2 charge updates accompany explicit electrons in eReaxFF. Authors apply an H-H bond bias (SI table) to lower H2-formation barriers for observable MD time scales.

1 — MD application. Code: eReaxFF-based dynamics in AMS for cases cited. Size: order ~3000-atom slab supercell; (100) Zr-Y-O with 12.5% surface O vacancies. PBC slab geometry. NVT at 1000 K; timestep and Nose/Berendsen details for all runs: N/A if not in this note—PDF tables. Duration: H2O dosed every 200 ps; cumulative 1.4 ns steam MD cited in the narrative. No NPT hydrostatic pressure or barostat in the long zero-bias example summarized. No external electric field; voltage enters via biasing and explicit electron treatments, not a continuum field table here. Enhanced sampling beyond bond-biased / constrained pathways: N/A in the short summary. 2 — Force-field training. Covered above (QM targets, C-DFT, eReaxFF reoptimization). 3 — Static QM-only: not applicable as the main contribution—DFT and C-DFT serve the eReaxFF fit; standalone static-only claims follow the QM and eReaxFF validation sections in the paper.

Findings

DFT and eReaxFF agree that BZY20 "Type 3" Y-doping is most stable and that EOS near equilibrium volume matches well. Surface chemistry: On Zr-Y-O (100) with vacancies, water adsorbs strongly to vacancies; Path A (H on lattice O between Y and Zr) is kinetically favored over Path B (between two Zr) with lower barrier (~7 vs ~9 kcal/mol) and more exothermic splitting. H2 evolution barriers from surface OH are very high without voltage (~70-90 kcal/mol class), so no H2 forms in zero-bias long MD. Ba-O (100) termination shows stronger vacancy adsorption but less favorable single-water splitting unless water-assisted proton transfer lowers the barrier (~13.4 kcal/mol vs ~37.5 kcal/mol for single-water path in the reported scan). Explicit excess electron on OH lowers one H2 barrier pathway to ~45 kcal/mol from ~67 kcal/mol; with H-H bias, barriers fall to about 28.4 kcal/mol (no excess electron) and ~20 kcal/mol (with excess electron). 1.4 ns cumulative steam MD on the Zr-Y-O surface shows vacancy filling, water dissociation, proton transfer into subsurface layers (down to sixth layer cited), and no H2 without applied bias, consistent with high barriers. Comparisons / parameter sensitivity (e.g. surface termination, bias vs H-H workarounds) are as stated in the PDF; kcal/mol locators in this note are VOR-backed, not the galley duplicate elsewhere in the corpus.

Limitations

eReaxFF and bond-biased / explicit-electron workflows are approximate stand-ins for electrode polarization and electrolyte microenvironments; zero-bias long MD omits steady-state applied voltage boundary conditions. Absolute oxygen-vacancy formation energies are described as qualitatively tracked in places; transferability beyond BZY20-like stoichiometries and surface terminations is not established in one article.

Relevance to group

Co-authored eReaxFF development for proton-conducting perovskite electrochemistry with Adri C. T. van Duin, supporting solid-oxide interface simulation in the corpus.

Citations and evidence anchors