Multiphysics-induced nonlinear dynamic analysis of perovskite optoelectronic devices via a machine learning-assisted constitutive model

Metal halide perovskites have garnered extensive research attention in the fields of optoelectronic devices and solar cells. However, their mechanical properties under multi-physics fields, especially the nonlinear effects induced by complex physical environments, remain largely unexplored. This paper establishes a machine learning-assisted constitutive model that embeds supervised polynomial regression surrogates to map light intensity, crystal thickness, illumination time, temperature and AC frequency to the photostrictive, photothermal, and dielectric responses of lead halide perovskites under multi-physics fields. The nonlinear governing equations of perovskite-based optoelectronic devices are formulated based on Timoshenko beam theory and von Kármán nonlinearity and are subsequently solved using the differential quadrature method, direct iteration, and the incremental harmonic balance method. Comprehensive parametric studies are conducted to explore the nonlinear frequency characteristics and dynamic responses of perovskite beams under multi-physics fields. The numerical results reveal that increasing light intensity and voltage, and decreasing crystal thickness, lower the natural frequency, shift the time history curve rightward, contract the phase trajectory, and increase the nonlinear frequency ratio, peak amplitude, and nonlinear hardening behavior. These findings contribute to the practical applications of perovskite-based devices in optoelectronics and related fields.