Metasurface Vision Transformer: A Generic AI Model for Metasurface Inverse Design

TL;DR

MetasurfaceViT is a new AI model that can design metasurfaces for various optical applications with high accuracy and adaptability.

Contribution

Introduces MetasurfaceViT, a universal AI model for metasurface inverse design that works across wavelengths and polarizations.

Findings

MetasurfaceViT achieves over 99% prediction accuracy for physically realistic designs.

The model successfully designs multiplexed printings, holograms, and broadband achromatic metalenses.

Physics-informed data augmentation and full-wavelength Jones matrix reconstruction enable one-shot design adaptability.

Abstract

Metasurfaces enable diverse applications by controlling light's amplitude, phase, and polarization. Although deep learning‐based inverse design has revolutionized metasurface design, current models are limited by fixed operating conditions and lack universality, often requiring retraining for new wavelengths, polarizations, or application scenarios. To address this, we introduce MetasurfaceViT (Metasurface Vision Transformer), a generic AI model for inverse design. Our solution leverages a large dataset of Jones matrices, significantly expanded via physics‐informed data augmentation. By pretraining through masking wavelengths and polarization channels, MetasurfaceViT can reconstruct full‐wavelength Jones matrices, which are then used by a fine‐tuning model for inverse design. This versatility allows one‐shot structure design for arbitrary wavelength, polarization, and application requirements. We demonstrate MetasurfaceViT's capabilities in designing multiplexed printings and holograms and broadband achromatic metalenses. Prediction accuracy exceeds 99% for physically realistic designs, showcasing a significant step toward a universal optical inverse design paradigm.

MetasurfaceViT is a generic AI model that addresses limitations in current metasurface inverse design by enabling universal control across wavelengths and polarizations. The authors detail their approach using Jones matrices and physics‐informed data augmentation for full‐wavelength reconstruction, alongside a fine‐tuning methodology. They showcase MetasurfaceViT's capabilities in designing multiplexed printings/holograms and broadband achromatic metalenses, highlighting its high accuracy and transfer learning potential.

Figures