Mitigating sampling-induced performance limitations in high-NA metalenses via transfer-learning-enabled nonlocal...

Area of Science:

• Optics and Photonics
• Nanophotonics
• Computational Design

Background:

• High-numerical-aperture (NA) metalenses face performance limitations due to the breakdown of the locally periodic approximation (LPA).
• This breakdown stems from aliasing from undersampling and significant inter-element coupling, particularly in peripheral regions.
• Existing design methods struggle to accurately model these complex interactions.

Purpose of the Study:

• To introduce a novel transfer-learning-enabled nonlocal inverse-design framework for metalenses.
• To capture short-range inter-element interactions effectively while maintaining computational efficiency.
• To enhance focusing efficiency and phase fidelity in high-NA metalenses.

Main Methods:

• Development of a transfer-learning-enabled nonlocal inverse-design framework.
• Design and simulation of focusing metalenses with NAs ranging from 0.85 to 0.95.
• Comparison of the proposed framework against conventional LPA-based designs.

Main Results:

• Optimized metalenses designed with the nonlocal framework exhibit significantly higher focusing efficiency and improved phase fidelity compared to LPA designs.
• At NA = 0.85, focusing efficiency increased from 33.9% to 44.5%, with a 32% reduction in phase root-mean-square error.
• The nonlocal framework effectively utilizes inter-element coupling as a designable parameter, mitigating phase errors and reducing scattering losses.

Conclusions:

• The nonlocal inverse-design framework offers a viable strategy for overcoming limitations in high-NA metalens design.
• This approach enhances metalens performance by addressing inter-element coupling and sampling issues near the Nyquist limit.
• The developed framework paves the way for more efficient and accurate metalens designs.