Skip to content

Statistical Analysis of Tri-Cresyl Phosphate Conversion on an Iron Oxide Surface Using Reactive Molecular Dynamics Simulations

The authors fit ReaxFF parameters for Fe–P–O chemistry relevant to tri-cresyl phosphate (TCP) on amorphous iron oxide, then run one hundred independent LAMMPS replicas to build temperature-resolved statistics of which TCP atoms form load-bearing bonds to the surface.

Summary

The paper develops a ReaxFF description of Fe/P/O interactions for TCP chemisorption on passivated amorphous iron oxide and applies it in one hundred parallel NVT molecular dynamics simulations, each containing a single TCP molecule above the same substrate realization. Initial orientations split evenly between P=O-down and P=O-up poses. Simulations follow a temperature ladder from 300 K to 700 K in 100 K steps, with up to one nanosecond per stage until bond populations plateau, higher-temperature runs restarted from lower-temperature final frames, and Nosé–Hoover thermostats acting on non-fixed atoms. The study contextualizes site-resolved bonding statistics with experimental TCP tribofilm formation under oxygen-poor conditions.

Methods

Force-field training (2). Parent / scope: new ReaxFF terms for Fe–P–O chemistry. DFT reference: training data for Fe–O–P angles, Fe–P dissociation, Fe(II)/Fe(III)–O–P bending, and P/PO adsorption on iron facets; Table 1 and Figure 2 give ReaxFF vs DFT errors; parameters in SI. Optimization / fit: standard ReaxFF least-squares optimization as described in the article. Substrate: one passivated amorphous Fe\(_x\)O\(_y\) slab.

MD application (1). Engine: LAMMPS RMD with 0.25 fs timestep; bond order every 1.25 ps (cutoff 0.3). System / composition: one TCP per cell on the oxide slab; 100 independent replicas, P=O-down vs P=O-up 50/50 initial orientations. Boundaries / PBC and fixed substrate atoms as in the PDF (periodic in-plane; bottom layers frozen or constrained per article). Ensemble: NVT; Nose–Hoover thermostat on non-fixed atoms. Barostat / mean pressure / stress control: N/A — NVT only. Duration: temperature ladder 300–700 K in 100 K steps, ~1 ns per stage to plateau; last ~25 ps of each steady window averaged. Electric field: N/A. Replica / parallel ensemble: 100 replica MD (not metadynamics or replica exchange for enhanced sampling — N/A for those).

DFT (3). Training DFT is cited above; this paper is RMD-centered, not a standalone static QM benchmark study.

Findings

Among heteroatom bonds tracked for film-growth metrics, Fe–C contacts are the most probable, while hydrogen-mediated contacts are excluded as non-load-bearing. Raising temperature from 300 K to 700 K increases overall TCP–surface bonding but shifts the dominant attachment sites from aryl carbons (notably C5/C6) toward phosphoryl oxygens at the highest temperature. The replica ensemble produces site-resolved histograms that short single trajectories cannot sample reliably. The abstract notes that parallel replica statistics should preserve relative reaction probabilities compared with one long serial trajectory, echoing prior work cited in the introduction on time-parallel MD for surface chemistry.

Limitations

Only one disordered oxide morphology is modeled, and the simulations omit shear or pressure from sliding contact, so tribological load transfer is represented through static thermal sampling rather than mechanochemical conditions.

Relevance to group

van Duin / Shin ReaxFF development for aviation lubricant additive chemistry on ferrous surfaces.

Citations and evidence anchors

  • papers/Khajeh_FeO_Phosphate_JPCC_2019.pdf

Reader notes (navigation)