Revealing the Combustion and Fluorination of Aluminum in HF/O2 Atmospheres by Molecular Dynamics Simulations

JH Han and MJ Wen and XY Chang and ZH Zhou and DP Chen and QZ Chu, JOURNAL OF PHYSICAL CHEMISTRY C, 129, 20506-20516 (2025).

DOI: 10.1021/acs.jpcc.5c01242

This study develops a high-precision neural network potential (NNP) model based on machine learning to simulate the heterogeneous combustion reactions between aluminum (Al) and HF/O2 gas molecules. It investigates the corrosion and oxidation mechanisms of Al induced by HF/O2 at the atomic level, focusing on the corrosive effects of HF on the Al surface. The NNP model is validated against an ab initio database, showing high accuracy in predicting atomic energies, forces, crystal parameters, equations of state (EOS), and adsorption energies. The model effectively captures the microscopic mechanisms of surface reactions on Al and performs reliably in various conditions. Molecular dynamics (MD) simulations using the NNP model study Al fluorination by HF molecules, compared to O2-mediated reactions. The results show that HF causes heterogeneous corrosion on the Al2O3 layer, producing AlF3 gas and exposing reactive Al surfaces. In contrast, O2/Al reactions lead to the formation of a dense oxide layer through chemical adsorption. Calculations of diffusion coefficients and energy barriers further confirm that Al atoms exhibit a higher migration rate and lower diffusion energy barrier in HF/Al, whereas the oxidation reaction in O2/Al significantly suppresses Al diffusion. This research offers new theoretical insights into HF/Al combustion and is the first to apply high-precision machine learning potentials to multiphase interface combustion studies, supporting propellant combustion optimization.

Return to Publications page