Artificial intelligence-generated photonics: mapping optical properties to subwavelength structures directly via a diffusion model

Subwavelength photonic structures and metamaterials provide revolutionary approaches for controlling light.The inverse design methods proposed for fabricable subwavelength structures are vital for the development of new photonic devices.However, most existing inverse design methods cannot realise direct mapping from optical properties to photonic structures; instead, they rely on forward simulation methods to perform iterative optimization.In this study, we exploit the powerful generative abilities of artificial intelligence and propose a practical inverse design method based on latent diffusion models.Our method directly maps the optical properties to structures without requiring forward simulation and iterative optimization.In this case, the given optical properties can serve as 'prompts' and guide the constructed model to 'draw' the required photonic structures correctly.Simulations and experiments show that our direct-mapping-based inverse design method can generate fabricable subwavelength photonic structures with high fidelity while following the given optical properties, such as the transmission power, phase, and polarisation responses.This may influence the methods used for optical design and significantly accelerate the research and manufacturing of new photonic devices.