Artificial Intelligence-Generated Photonics: Map Optical Properties to Subwavelength Structures Directly via a Diffusion Model
Publish Date: 2026-05-16 02:30:00
Source Domain: www.newswise.com
-
Revolutionizing Photonic Design: A research team led by Professor Kaiyu Cui has developed the AI-generated photonic framework (AIGP) that creates subwavelength photonic structures directly from optical properties using generative diffusion models, eliminating the need for iterative optimization processes.
-
Direct Mapping of Optical Properties: The AIGP framework allows high-precision mapping from full-band transmission spectra, phase profiles, and polarization responses to corresponding photonic structures, with generation times of merely seconds for manufacturing readiness.
-
Flexible Design Capabilities: It supports flexible design constraints, including polarization insensitivity and band-specific masking, allowing for adaptable and diverse design goals without complex forward modeling.
-
Experimental Validation and Fabrication: The technology was experimentally validated on a silicon-on-sapphire platform, successfully creating structural-color meta-atoms and encoding an image onto a chip, demonstrating the capability for immediate fabrication from design.
-
Overcoming Conventional Obstacles: The AIGP framework addresses traditional challenges of non-uniqueness, robustness to unseen inputs, and the complete avoidance of iterative procedures, presenting a revolutionary approach to generative photonic design.
-
Prospects for Future Innovation: The technology promises to expedite developments in various photonic applications, from optical computing to structural colors and beam splitters, ushering in an era of large-scale AI-driven photonic innovation.
