Data-driven quantification of grain boundary structures and their correlation with migration kinetics

Understanding the complex relationship between the grain boundary (GB) structure and its migration mechanisms is a critical challenge in materials science. This is due to the difficulty of experimental determination and the immense computational effort required for simulations. To accelerate this determination, we introduce a data-driven Process–Structure–Mechanisms–Properties (PSMP) framework to quantitatively analyze the GB structure during motion and link it to process conditions and kinetic properties. The core of this framework is the GB Structure Transition Function (GBSTF), a low-dimensional structural time-series derived from advanced topological data analysis. We quantify the GB structure using persistent homology (PH) and pair correlation functions (PCF), and then apply principal Component Analysis (PCA) to generate the GBSTF. We applied this framework to approximately 2800 molecular dynamics (MD) simulations of various Aluminum GBs, investigating migration driven by either mechanical stresses ( τ ) or a jump in chemical potential ( ψ ). Our results demonstrate a strong dependence of GB kinetics on the driving force type. It was found that the GB structure as quantified by GBSTFs correlates with the observed kinetic differences, indicating that the driving force type activates distinct migration mechanisms. This finding was validated by calculating the activation energy ( Q ) for the Σ 25 a ( 0 1 7 ) boundary, which showed significantly different values for ψ -driven ( Q ≈ 0 . 126 eV ) versus τ -driven ( Q ≈ 0 . 216 eV ) migration. The PSMP framework provides a unified, quantitative method to discriminate between complex GB mechanisms, offering a powerful tool for predicting and controlling microstructure evolution.