Researchers from Kyung Hee University and Hyundai Motor Group have developed an AI-driven inverse design strategy enabling full-color, high-efficiency semitransparent perovskite solar cells, advancing the prospects of solar-integrated windows and vehicle glazing. The findings were published in Opto-Electronic Advances.
The team introduced a modelling-guided framework that integrates all-dielectric multilayer coatings composed of zinc sulfide (ZnS) and magnesium fluoride (MgF₂) into perovskite photovoltaics. Unlike conventional color-tuning approaches that rely on absorptive or metallic films, the method uses transparent interference coatings optimized through a digital loop. Each multilayer configuration is encoded as a binary string and evaluated using a factorization machine-based surrogate model trained on optical simulations, with optimization formulated as a quadratic unconstrained binary optimization (QUBO) problem.
Using this approach, researchers achieved six target hues — red, green, blue, cyan, magenta, and yellow — across absorber thicknesses of 65–165 nm. A cyan device with a 110 nm MAPbI₃ absorber delivered a 20.9% increase in power conversion efficiency (PCE) over uncoated cells, alongside controlled visible transmittance. Flexible PET-based devices also demonstrated improved PCE while retaining mechanical flexibility.
Importantly, the coatings were deposited via thermal evaporation directly onto operational cells, confirming compatibility with scalable manufacturing. The framework, based solely on measured optical constants, can be extended to other thin-film PV technologies, offering a scalable pathway to aesthetically customizable, high-performance solar façades and automotive applications.
Subscribe to get the latest posts sent to your email.
