Cooperative game of composition, structure, and properties in rare earth bearing melts: Molecular dynamics simulation and interpretable machine learning study

Z Lyu and C Gu and Z Lyu and YH Liu and Q Hu and YP Bao, MATERIALS & DESIGN, 257, 114489 (2025).

DOI: 10.1016/j.matdes.2025.114489

Understanding rare earth element behavior in slag is crucial for improving their utilization in steel and enhancing slag melt metallurgical properties. However, the large compositional space and reactive chemical nature of these melts pose significant challenges. Molecular dynamics was employed to reconstruct short-range ordered rare earth-bearing slag melts. The viscosity was calculated using the reverse non-equilibrium molecular dynamics method, uncovering the stepwise variation in composition-structure-property relationships. An XGBoost machine learning model was developed, and the Shapley method from cooperative game theory was utilized to measure the contributions of composition, structure, and diffusion features to viscosity. This work offers a theoretical basis for boosting rare earth utilization in steelmaking, provides strategies for precise slag composition design and melt property adjustment, and confirms the principle of composition control-structure evolution-property regulation.

Return to Publications page