Metalenses, as the typical diffractive optical device, provide a platform for the research of the optical imaging and light field modulation, and have received massive attraction. However, the traditional complex design process relies on the numerical solutions of Maxwell’s equations and requires high computational and time resources. In recent years, the inverse design based on the nerual networks provides an effective method for the design of the diffractive optical devices, and improve the the design freedom and efficiency. In this paper, an inverse design method for multifocal metalens based on the tandem neural network is proposed, and achieves efficient forward prediction of the phases, and the design of structural parameters. Based on this network, a single-focus metalens designed with the target focal length of 9 μm has been achieved with the design error controlled below 2.8%. On the foundation, this method can also be used to realize the design of the multifocal metalens. The dual-wavelength metalens is first designed to realize the recognition of specific wavelength p and multifocal imaging, and then expanded to four wavelengths. The study has potential applications in the field of optical communication, imaging, detection, etc.
Song et al. (Wed,) studied this question.