- Research Article
- 10.1080/08927022.2026.2649552
Reactive force field (ReaxFF) molecular dynamics simulations and graph neural networks (GNN) prediction on antioxidant mechanisms of phenolic antioxidants in TMPTO lubricating oil
- Mar 31, 2026
- Molecular Simulation
- Wenjing Song + 5 more +5
To investigate the antioxidant mechanisms of phenolic antioxidants in base oils, this study employed the reactive force field (ReaxFF) molecular dynamics (MD) simulations to reveal the reaction pathways and product distributions of trihydroxymethylpropyl trioleate (TMPTO, base oil) and antioxidants AO-80 and XH-245 under high-temperature conditions. A multi-output regression prediction model (MORP) was developed to achieve accurate prediction of carbon atom types and chemical bond types in the reaction systems. Furthermore, a chemical reaction prediction model (CRPM) based on graph neural networks (GNN) was developed to predict the product types and distributions at different times during the antioxidant process of lubricating oils. The results show that the oxidative decomposition of TMPTO is characterised by the cleavage of C–C bonds and C–O bonds. The main small-molecule oxidation products include CO2, C2H4, H2O, CO, formaldehyde, formic acid, and other C2 and C3 compounds. The ArOH groups in phenolic antioxidant molecules can delay the oxidative cleavage process of base oil through hydrogen transfer steps. The system with the two types of antioxidants exhibits better antioxidant effects than the single antioxidant system under the same conditions. The MORP and CRPM models provide good prediction results for carbon atom types, chemical bond types, and product distributions.
Read more