Intelligent Inverse Design of Multi-Layer Metasurface Cavities for Dual Resonance Enhancement of Nanodiamond Single Photon Emitters

Single-photon emitters (SPEs) based on nitrogen-vacancy centers in nanodiamonds (neutral NV0 (wavelength 575 nm) and negative NV- (wavelength 637 nm)) represent promising platforms for quantum nanophotonics applications, yet their emission efficiencies remain constrained by weak light-matter interactions. Multi-layer metasurfaces (MLM) offer unprecedented degrees of freedom for efficient light manipulation beyond conventional single-material metasurfaces, enabling dual-resonance cavities that can simultaneously enhance pump excitation and SPE collection. However, traditional trial-and-error and forward optimization methods face significant challenges in designing these complex structures due to the vast parameter space and computational demands. Here, we present NanoPhotoNet-Inverse, an artificial intelligence-driven inverse design framework based on a hybrid deep neural network architecture. This model efficiently performs inverse design of two dual-resonance MLM cavities to amplify pump and SPE emissions of different vacancies in nanodiamond and improve SPE collection. Our approach achieves inverse design prediction efficiency exceeding 98.7%, demonstrating three orders of magnitude amplification in SPE count rate and 50 picosecond lifetime, significantly surpassing conventional cavity designs. These remarkable enhancements in emission rates and collection efficiency position our platform as a transformative technology for advancing quantum communication networks, quantum computing architectures, quantum sensing applications, and quantum cryptography systems. Therefore, it opens new pathways for intelligent photonic device engineering in quantum technologies.