Design of multifunctional tunable dual-layer metalens based on deep learning

Area of Science:

• Optics and Photonics
• Materials Science
• Artificial Intelligence

Background:

• Single-layer metalenses face limitations in modulation performance and control of individual meta-atoms.
• Achieving tunable, multifunctional optical devices requires advanced material properties and design methodologies.

Purpose of the Study:

• To propose and demonstrate a dual-layer tunable multifunctional metalens using barium titanate (BTO) for enhanced optical control.
• To accelerate the design process of tunable metalenses using a deep learning algorithm.
• To achieve rapid switching between dual and quadruple functionalities in the visible light spectrum.

Main Methods:

• Utilized a dual-layer metalens architecture with tunable BTO material for holistic modulation.
• Employed a deep learning algorithm based on transformer structure with band embedding for predicting high-frequency response.
• Simulated and analyzed the switching capabilities between various optical functionalities (focal length, position, beam deflection, splitting).

Main Results:

• Demonstrated near-perfect switching between dual and quadruple functionalities in the visible light range.
• The deep learning model accurately predicted meta-atom transmittance and phase, generating 100,000 arbitrary geometric meta-atoms in under 1 second.
• Simplified geometric meta-atoms were used to enhance manufacturability with minimal performance loss.

Conclusions:

• The proposed dual-layer tunable multifunctional metalens offers rapid switching and versatile optical control.
• Deep learning significantly accelerates the design and optimization of complex meta-atom structures.
• The design presents a promising pathway for advanced, manufacturable optical devices with switchable functionalities.