Synthetic driving force simulations of grain boundary migration with adaptive atom classification

Synthetic driving force (SDF) molecular dynamics methods provide an atomistic framework for probing grain boundary migration and underlying mechanisms. Existing formulations nevertheless may be affected by artifacts arising from the use of a fixed number of nearest neighbors, nonconservative forces, or sensitivity to user-defined parameters. In this study, we developed a modified cross-reference order parameter SDF (MCROP-SDF) method that combines a smoothly weighted fourth-neighbor descriptor, an adaptive threshold, and a pair-balanced force derived consistently from a continuous synthetic potential. The method was validated for representative tilt, twist, and mixed boundaries in FCC Ni and BCC Fe. The revised neighborhood restored pair reciprocity and suppressed spurious multilevel order parameter profiles in BCC crystals, while the adaptive threshold remained statistically robust. The force formulation yielded zero net synthetic force and satisfied work–energy consistency within numerical accuracy. For structurally ordered grain boundaries with well-resolved order parameter profiles, the MCROP-SDF method reproduced the temperature-dependent migration and shear-coupling behavior obtained with the energy conserving orientational SDF (ECO-SDF) method. In the low driving bias limit, the predicted mobility agreed closely with the zero-driving-force FAIRWalk benchmark in both magnitude and temperature dependence. For low-angle and strongly distorted boundaries, the ECO-SDF predictions became sensitive to the selected cutoff radius and threshold because of severe atom misclassification, whereas the present method provided stable classification and migration predictions without case-by-case retuning. The MCROP-SDF method therefore provides a robust and reproducible framework for atomistic GB migration simulations.