Flexible Dual Fabry-Perot Filters Enabling Color Balanced Narrowband Emission Designed by Physics Guided Neural Inverse Optimization.

Achieving high color purity through precise spectral control is a key optical requirement for advanced display systems. Here, we present a flexible dual Fabry-Perot interferometer (DFPI) consisting of two stacked Fabry-Perot cavities, where one generates multiple resonance peaks and the other precisely tunes their spectral alignment and intensity. The optical behavior of the DFPI was analytically modeled as the product of individual Fabry-Perot factors (FPFs) and experimentally verified on a PET substrate. Furthermore, a data-driven bi-directional neural network (Bi-NN), integrating forward and inverse models was employed to improve design efficiency, providing an accurate mapping between structure and optical variables and mitigating the non-uniqueness issue in inverse prediction. The optimized Bi-NN predictions exhibited excellent agreement with the analytical formulation results. When integrated with a blue LED and quantum dot color conversion films, the fabricated DFPI film reduced the spectral full width at half maximum (FWHM) and expanded the color gamut by over 30% relative to sRGB gamut, while maintaining excellent spectral and color stability under repeated bending conditions. This integrated optical and AI design framework offers a promising pathway for realizing next generation flexible displays with high color purity and reproducibility.