Exploring Thermal Properties of Al-Mg alloy Using Neural Network Potential

Al-Mg alloys have been identified as having significant potential for application in energetic materials and propulsion systems. The thermophysical properties of these alloys directly influence a number of processes, including propellant preparation, storage safety, and combustion efficiency. In this study, a high-precision NNP for Al-Mg alloys was constructed based on first-principles data covering the phase diagram composition. This potential function demonstrates excellent performance in predicting a number of key properties, including energy, atomic forces, crystal structure, mechanical properties, vibrational characteristics, and dynamic evolution over a wide temperature range. This renders it a dependable instrument for atomic-scale thermal property research. Employing the innovative NNP potential, molecular dynamics simulations were employed to systematically investigate the effects of Mg content on the melting point, thermal expansion coefficient, specific heat capacity, thermal conductivity, radial distribution function, and Warren-Cowley parameters of Al-Mg alloys. The findings suggest that even a minimal amount of Mg doping can substantially reduce the alloy's melting point and thermal conductivity, while concurrently increasing the thermal expansion coefficient and specific heat capacity. As the Mg content rises, the thermal expansion coefficient and thermal conductivity demonstrate non-monotonic trends, with inflection points at 40 wt.% and 30 wt.%, respectively. This phenomenon is intimately linked to the stability of the Al-Mg alloys crystal structure. These findings reveal, at the atomic scale, the regulatory mechanisms of Mg content on the thermal properties of Al-Mg alloys, providing an atomic-scale basis for composition design and performance control in propellants.