Rapid Inverse Design of High Degree of Freedom Meta-Atoms Based on the Image-Parameter Diffusion Model

• DOI: 10.1109/jlt.2024.3391924
• Publication Date: 2024-04-22T17:39:05Z
• Authors: Lu Zhu; Wei Hua; Cong Lv; Yuanyuan Liu
• Keywords: Degree (music)

The inverse design of metamaterials is the mapping spectral response (low dimensions) to structure (high dimensions). In process design, there will be a "one-to-many" problem (similar responses are generated by multiple meta-atoms), especially for designs with high degrees freedom (DOFs). Although traditional generative models (VAE, GAN) combined optimization algorithms can convert into forward prediction solve problem, this method requires iterative calculations and has low efficiency, making it difficult achieve fast accurate design. This paper proposes novel based on image-parameter diffusion model. Through step-by-step adding noise that transforms structural parameters images Gaussian distributions, our gradually remove starting from distribution generate DOFs meta-atoms satisfy transmission coefficient additional inputs. Moreover, avoids optimizations ensures more efficient high-quality generation results. experimental results show solves high-DOFs free-form structures only one iteration, which superior conventional methods in terms model speed, accuracy, quality.