While the Fe-P-C metallic glass (MG) is a classic soft magnetic material, its broader application is hindered by limited glass-forming ability (GFA), whose atomic-scale origins remain inadequately understood. Unraveling these origins necessitates a precise description of atomic interactions and a comprehensive analysis of structural evolution during vitrification. However, developing reliable interatomic potentials for this ternary alloy has been challenging due to its complex hybrid metallic-covalent bonding. In this study, a deep-potential (DP) model for the Fe-P-C system was constructed by iteratively training a neural network within an active learning framework using datasets derived from ab initio molecular dynamics (AIMD) and density functional theory (DFT). The optimized potential achieves near-DFT accuracy, with deviations in lattice constants $\le 0.05\AA$ for crystals and energy-volume equations of state within 1% for amorphous phases, thereby capturing hybrid bonding features inaccessible to empirical potentials. Large-scale molecular dynamics simulations involving 5000 atoms over nanosecond durations reveal hierarchical structural evolution of $\mathrm{F}{\mathrm{e}}_{80}{\mathrm{P}}_{13}{\mathrm{C}}_7$ during the liquid-to-glass transition. As temperature decreases, phosphorus coordination shifts from solute-solute to solute-solvent preference, while weakly covalent C–C dimers ($\sim 1.425\AA$) become kinetically trapped upon cooling. Concurrently, $\langle 0,4,4,x\rangle$ Voronoi polyhedra emerge as structural intermediaries between liquid and glass states. Analysis of medium-range order further reveals distinct aggregation-repulsion patterns among Voronoi clusters and fivefold symmetric units, underscoring a competitive interplay between chemical ordering and topological frustration. These findings confirm the applicability of DP modeling to multicomponent alloys with hybrid bonding nature, provide mechanistic insights into the glass formation process of Fe-P-C MG, and elucidate the cooperative roles of chemical and topological short-range ordering in ternary MGs.

Close

PrevNext

View figure in article

• Interatomic & molecular potentials
• Amorphous materials
• Glassy systems
• Machine learning
• Molecular dynamics