An interface-oriented data-driven scheme applying into eutectic patterns evolution

Abstract

Solid-liquid interface dynamics plays a crucial role in microstructure evolution during solidification. However, it is difficult to extract sufficient spatial information to describe the variation in the irregular and non-periodic solid-liquid interface domains, resulting in a challenge remaining in quantifying analysis by a quick and low-dimensionally visualized way. We proposed an interface-oriented two-point statistics method to solve the unbiassed statistics of the spatial information in these irregular and non-periodic interfaces. The challenge was successfully overcome by a carefully designed data-driven scheme, which has the advantages of easy to operate, high accuracy and efficiency. Taking the eutectic microstructure evolution simulated by phase-field method as an example, we also obtained novel insights on tailoring eutectic patterns, that is, the transformation of four patterns (lamellar elimination, regular lamellae, oscillatory eutectics and branching lamellae) depends on initial lamella spacing whether the interface energy is isotropic or anisotropic. Additionally, interfacial energy anisotropy, compared to the isotropic case, not only prolongs the time of patterns transformation, but also narrows the spacing range between elimination and creation of lamellae. Our proposed scheme is also able to be applied into more occasions about microstructure evolution in different materials systems.