ANALYSIS OF PIEZOELECTRIC SEMICONDUCTORS VIA DATA-DRIVEN MACHINE-LEARNING TECHNIQUES

Piezoelectric semiconductors (PSs) have great potential applications in future multifunctional devices. The interaction among deformation, polarization, and carrier in PSs under an external load leads to a variety of innovative electronic, mechanical, and optical behaviors. Studying multi-field coupling mechanical behaviors of PSs plays a fundamental role in PS devices. However, the partial differential equations (PDEs) describing PSs are nonlinear. This brings rather a mathematical challenge to analysis of PSs. In this work, we propose a machine-learning-based approach to solving multi-field coupling problems of piezoelectric semiconductors. Based on the linear analytical solution, the physics-informed neural networks (PINNs) is trained to obtain the data-driven solutions for a static axial extensional PS rod.