[P] P24~P35 Computational Science

[P30] Grain Boundary Energy-Conditioned Inverse Design via symmetry constrained diffusion model

○Kazuya Miyamoto^1,2, Ibuki Okuda^1,2, Teruyasu Mizoguchi¹ (1. IIS, UTokyo, 2. School of Engineering, UTokyo)

Keywords:

Materials Informatics,Machine Learning,Deep Learning

We extend the diffusion model MatterGen to grain boundaries via a mirror-symmetry constraint that keeps two boundaries in the periodic cell equivalent, enabling inverse design of Cu Σ5 twist boundaries whose relaxed energies monotonically follow the target grain-boundary energy.