Small-sample efficient inverse design of anti-reflection micro-nano optical structures: a Bayesian optimization-tuned...

Area of Science:

• Optics and Photonics
• Computational Electromagnetics
• Materials Science

Background:

• Inverse design of micro-/nano-optical anti-reflection structures faces challenges due to complex parameter mapping and high computational costs.
• Existing data-driven methods struggle with small sample sizes, leading to performance degradation.

Purpose of the Study:

• To develop an efficient and accurate closed-loop inverse design framework for micro-/nano-optical anti-reflection structures.
• To overcome the limitations of existing computational and data-driven approaches in optical structure optimization.

Main Methods:

• Proposed a closed-loop inverse design framework integrating Bayesian optimization with extremely randomized trees (ERT) as a surrogate model.
• Employed differential evolution algorithm for optimization, coupled with the ERT surrogate model.

Main Results:

• The Bayesian optimization-tuned ERT model achieved an average relative error of 1.01% on independent test samples, surpassing six mainstream machine learning models.
• The framework demonstrated 100% constraint satisfaction for reflectivity ≤20% at 1064 nm across 0°-70° incident angles.
• Achieved a 99.67% reduction in finite element method calls and a ~307× computational speedup.

Conclusions:

• The proposed framework offers an engineering-feasible solution for high-performance optical structure design.
• This approach significantly enhances the efficiency and accuracy of inverse design for anti-reflection structures.
• The method addresses the critical need for faster and more reliable optical component design.