Evolutionary Diffusion Framework Empowering High-Performance Freeform Terahertz Metasurface Sensing

演化扩散框架助力高性能自由曲面太赫兹超表面传感

阅读:1

Abstract

Metasurfaces offer an unprecedented avenue to facilitate light-matter interactions. However, traditional design methodologies rely on computationally intensive trial-and-error processes. Moreover, existing deep learning (DL) schemes are predominantly hindered by their massive data requirements and limited exploration of freeform design spaces. To overcome these challenges, a multi-model-driven generative-evolutionary strategy (GES) is proposed, for the on-demand inverse design of bespoke Terahertz (THz) metasurface sensors. By leveraging a Conditional Diffusion Generator (CDG) and an Attention-Enhanced Residual Network (ARN), this framework enables the exploration of an expansive design space encompassing 2100 possible configurations. The GES effectively overcomes the data bottleneck by selectively generating high-potential data in stages. Full-wave simulations confirm that the inversely designed metasurfaces exhibit high-contrast resonance peaks and exceptional sensitivity across low, mid, and high THz bands. This work provides a versatile paradigm for the efficient design of high-performance functional metamaterials, significantly accelerating the advancement of application-specific THz sensing.

特别声明

1、本页面内容包含部分的内容是基于公开信息的合理引用;引用内容仅为补充信息,不代表本站立场。

2、若认为本页面引用内容涉及侵权,请及时与本站联系,我们将第一时间处理。

3、其他媒体/个人如需使用本页面原创内容,需注明“来源:[生知库]”并获得授权;使用引用内容的,需自行联系原作者获得许可。

4、投稿及合作请联系:info@biocloudy.com。