Artificial intelligence-empowered functional design of semi-transparent optoelectronic and photonic devices via deep Q-learning

Photonic-based design of semi-transparent organic solar cells (ST-OSCs) demands a careful balance between optical transparency and photovoltaic efficiency, often requiring trade-offs that complicate optimization. This study, for the first time, employs deep Q-learning, a reinforcement learning algorithm, to address this challenge, integrating transfer matrix method for precise optical calculations. The proposed framework optimizes asymmetric dielectric/metal/dielectric photonic-based transparent contact systems combined with novel PBDB-T:ITIC-based active layers, achieving superior optical and photovoltaic performance. The deep Q-learning algorithm successfully identified configurations yielding a maximum photo-current density (J_ph) while effectively maintaining average visible transmittance (AVT), balancing transparency, and photon harvesting by learning Maxwell's equations. Precise tuning of material thicknesses and optical properties further enhanced performance, ensuring color neutrality and high rendering index. These ST-OSC designs are particularly suited for building-integrated photovoltaics and photovoltaic windows, where both functionality and aesthetics are critical. This study also highlights the transformative potential of artificial intelligence in optoelectronic device design. The deep Q-learning framework accelerates optimization processes, reduces computational demands, and enables scalable solutions, surpassing traditional methods in efficiency and precision. By addressing the complex interplay of optical and photovoltaic parameters, this research advances the state-of-the-art ST-OSCs and establishes a foundation for future machine learning-driven innovations in sustainable energy technologies.