Understanding the relationship between the undercooled liquid and stable and metastable phases which form is a long standing challenge in materials synthesis. Knowledge about the local orders in the liquid and glassy states and their relationship to the structure of crystalline phases are vital to navigating the energy landscape necessary to predict and guide materials processing. This is particularly true in intermetallic
compounds containing rare earth (RE) elements which exhibit a plethora of stable and metastable phases. By taking the advantage of deep machine learning (ML) technology, an accurate interatomic potential for Al-Tb system based on an artificial neural network approach was developed. This rapidly developed potential enables efficient molecular dynamics simulations to generate reliable atomistic structure models for the Al90Tb10 metallic glass. ML-assisted analysis based on cluster alignment reveal the short- and medium-range structure order in the metallic glass are consistent with building blocks in the low-energy metastable crystalline structures at similar chemical compositions. This discovery provides atomistic level understanding of phase competition and selection in complex intermetallic systems involving rare earth (RE) elements.
L. Tang, Z. J. Yang, T. Q. Wen, K. M. Ho, M. J. Kramer, and C. Z. Wang. “Short- and medium-range orders in Al90Tb10 glass and their relation to the structures of competing crystalline phases”Acta Materialia, 204 116513 (2021), https://doi.org/10.1016/j.actamat.2020.116513