Machine Learning Potential Construction and Compressive Mechanical Properties of Al-Cu Intermetallic Compounds

The optimization design of Al-Cu intermetallic compounds is crucial for the mechanical properties of Al-Cu alloys. Molecular dynamics (MD) simulation can provide microscopic processes of the mechanical behavior of Al-Cu alloys, and the interatomic potential is the key physical basis to ensure the reliability of the MD simulation. This work constructed a depth potential (DP) function for the Al-Cu system based on first-principles calculations, and compared the physical properties predicted by DP (crystal structure, energy-volume curve, pressure-volume curve, and phonon spectrum) with density functional theory (DFT) and embedded atom method (EAM) results. The generalization ability and accuracy of the DP model were verified. Based on the DP potential, MD simulations were conducted on the compression process of five Al-Cu intermetallic compounds (θ-Al₂Cu, θ′-Al₂Cu, Al₃Cu, Al₄Cu₉, and AlCu₄ phases). The characteristics and laws of yielding phenomena in structures such as θ-Al₂Cu, θ′-Al₂Cu and AlCu₄ were presented. The yield stress and shear stress of θ-Al₂Cu, θ′-Al₂Cu and AlCu₄ increase with the increase of strain rate, and the yield strain also increases correspondingly. This phenomenon arises from the enhancement of phonon drag obstruction to atomic slip. Among them, θ-Al₂Cu has the best compressive performance, yielding at a strain rate of 4×10⁹ s⁻¹ when compressed to 17.4%, with a yield strength of 51.15 GPa. Screw dislocations are produced, and the atoms slip along the [$ \overline{1} 11$], [111] and [$11 \overline{1} $] directions. θ′-Al₂Cu yields when compressed to 10.0%, and the atoms slip in the plane perpendicular to the compression. Yielding occurs when the AlCu₄ phase is compressed to 13.4% and the atoms slip along the [401] and [$40 \overline{1} $] directions.