| Abstract Scope |
Photoplasticity, the light-induced change in semiconductor mechanical properties, is critical to optoelectronic reliability, yet its atomic origins remain unclear. Here, we combine constrained density functional theory (CDFT) with machine-learning potentials (MLPs) to investigate photomechanical deformation in ZnS and CdTe, two II–VI semiconductors with contrasting responses. In ZnS, photoexcitation increases the Peierls stress of 30° and 90° partial dislocations by 9–61% without changing their core structures. Electron–hole pairs reduce charge density at dislocation cores, redistribute strain into the surrounding matrix, and suppress dislocation glide, explaining the illumination-induced ductile-to-brittle transition. In CdTe, MLPs for the ground state and two excited states reveal a non-monotonic response: the minimum Peierls stress first increases by ~21% and then decreases by ~24%, consistent with experiments. This behavior arises from coupled electronic–lattice interactions, dislocation core reconstruction, and FCC-to-BCC martensitic transformations. This CDFT-MLP framework enables efficient prediction of photo-induced mechanical behavior in semiconductors. |