Heterogeneous nucleation during metal solidification is a key pathway for achieving grain refinement. However, due to the challenges in experimentally probing the solid-liquid interface, the underlying atomistic mechanisms remain unclear. Although classical lattice matching theories provide a framework for geometric compatibility, the static rigid models inherently neglect the strong atomic dynamics characteristic of high-temperature interfaces. In this study, ab initio molecular dynamics (AIMD) with on-the-fly machine learning force field (MLFF) is applied to evaluate the templating efficiency of different potential substrates for Al alloys. The templating efficiency is defined here as the capability to induce structural ordering within the pre-nucleation layer (PNL). The results show that a deep interfacial potential well (Ew) provides the necessary driving force to localize liquid atoms, whereas excessive atomic vibrations spatially disrupt this ordering process. The effect of temperature prompts the integration of interface dynamics into geometric mismatch. The dynamic stability is the main prerequisite for the initial PNL formation near the liquidus, and the epitaxial growth of PNL into stable nuclei is constrained by the accumulated lattice strain as the temperature decreases. Based on this dynamic-static mechanism and engineering considerations, a (Ta, Ti)B2/Al3Ta substrate is designed, which achieves significant grain refinement in Al-Si alloys. These findings deepen the understanding of heterogeneous nucleation and highlight interfacial dynamics as an important factor for the design of nucleation substrates.
扫码关注我们
求助内容:
应助结果提醒方式:
