A Hybrid Optimization Strategy for Achieving Ultra-Compact and Ultra-High Q Waveguide Cavity

Despite recent advances, achieving higher quality factors while maintaining smaller mode volumes remains a major challenge in current cavity designs. Since strong spatial confinement with minimal electromagnetic leakage depends on a precisely optimized hole-radius distribution, an excessively high computational burden is often required. This paper proposes a hybrid strategy that combines a machine learning model with a genetic algorithm. A quality factor exceeding 108 is achieved, and the corresponding 15-hole pair Fabry–Perot cavity exhibits a mode volume of 0.52 (λ/n)3. The approach thus yields a high Q-factor within a cavity of such ultra-compact dimensions while significantly reducing calculation cost. A comprehensive comparison is also given, demonstrating these certain advantages over the other three methods—analytical tapering profiles, a pure genetic algorithm, and machine-learning-based estimation—applied separately in this study. The methodology considered may provide an adaptable framework for various related complex photonic optimization problems.