Cluster and network formation in Fe-Si-B amorphous alloys: a machine learning molecular dynamics study

Local atomic arrangements in amorphous materials, particularly short- and medium-range orders (SRO and MRO, respectively), govern their metastability and distinctive physical properties. In 1979, Gaskell proposed that the structural motifs of transition metal-metalloid glasses reflect the ordering tendencies found in their crystalline counterparts. In this study, we investigate SRO and MRO in amorphous Fe-Si-B alloys, a key class of soft-magnetic transition metal-metalloid glasses used in energy-efficient motor applications. We previously developed a machine-learning interatomic potential based on the Gaussian approximation potential framework for this system. Using this potential, we performed melt-quench molecular dynamics simulations on 2048-atom systems to clarify the structural roles of the minor elements B and Si, focusing on both SRO and MRO. The simulations reveal that B-centred clusters predominantly adopt bi-capped anti-prism (BAP) and tri-capped trigonal prism (TTP) geometries, which are characteristic of crystalline Fe₂B and Fe₃B, respectively. This result provides strong support for Gaskell's structural model. In contrast, Si-centred clusters preferentially form icosahedral motifs that are typically associated with amorphous structures. At the MRO level, Si- and B-centred clusters exhibit distinct modes of cluster-pair connectivity, with B-centred clusters, particularly those based on the BAP geometries, showing a higher tendency towards network formation.