Exploring AI in metasurface structures with forward and inverse design.

Area of Science:

• Metasurface design
• Artificial intelligence in electromagnetics
• Computational electromagnetics

Background:

• Metasurfaces are artificial planar devices producing unique electromagnetic responses.
• Traditional metasurface design relies on time-consuming numerical algorithms and parameter optimization.
• Existing methods often struggle to accurately match desired performance.

Purpose of the Study:

• To classify artificial intelligence (AI)-enabled metasurface design methodologies.
• To provide a unique perspective on forward and inverse AI design approaches.
• To review recent advancements, principles, advantages, and applications of AI in metasurface design.

Main Methods:

• Classification of AI-enabled design into forward and inverse approaches based on variable-performance mapping.
• Review of intelligent algorithms, particularly neural networks, for inverse design.
• Systematic examination of AI-driven metasurface design principles and techniques.

Main Results:

• AI-enabled design offers a novel approach to metasurface development.
• Forward designs are driven by intelligent algorithms, while neural networks are key for inverse design.
• The review details principles, advantages, and potential applications of AI in this field.

Conclusions:

• AI significantly enhances metasurface design efficiency and accuracy.
• This systematic review provides a holistic understanding and aids future research.
• The findings facilitate the selection of practical applications for AI-designed metasurfaces.