AI driven coatings enable full color solar windows without losing power

by Clarence Oxford

Los Angeles CA (SPX) Mar 30, 2026

A new study has demonstrated that artificial intelligence can design color-tunable solar windows that maintain, and even enhance, their power output compared to conventional transparent solar cells. The work focuses on semitransparent perovskite photovoltaics, a promising technology for building-integrated applications where windows and other surfaces can generate electricity while remaining see-through. By combining detailed optical modelling with an inverse design strategy guided by AI, the researchers created non-metallic, non-absorbing dielectric coatings that control the perceived color of the solar window without sacrificing efficiency.

Most existing semitransparent solar technologies face a trade-off between aesthetics and performance because color is often achieved using metal filters or absorbing layers that waste incident light. These approaches reduce the amount of light available for power generation, limit color options, and constrain how easily devices can be integrated into architecture, vehicles or electronics. In contrast, the new strategy uses transparent dielectric coatings, similar to the multilayer stacks in precision optical components, to shape how light interferes within the device. By adjusting the thickness and refractive index of these layers, the team can tailor the reflected and transmitted spectra to produce user-defined colors while keeping the photovoltaic layer highly active.

Using this AI driven inverse design, the researchers realized semitransparent perovskite solar cells that display distinct colors including red, green, cyan, magenta and gray. Instead of reducing efficiency, the optimized dielectric coatings increased power generation by up to 20 percent compared to uncoated reference devices with the same active layer. This improvement arises because the coatings can simultaneously minimize parasitic reflections at wavelengths where the perovskite absorbs strongly and redistribute light in ways that favor electricity production. The result is a device that looks like a colored window yet performs better than a transparent solar cell with a simple, unoptimized stack.

The study emphasizes that the approach does not rely on metallic mirrors or strongly absorbing pigments, both of which typically introduce substantial optical loss. Instead, it exploits high index contrast dielectric materials and interference effects to sculpt the optical response. This makes the coatings suitable for both rigid glass substrates and flexible plastic supports, extending their applicability beyond conventional flat window panes. The same design framework can be adapted to different perovskite compositions, layer thicknesses and target colors without rethinking the entire device structure, because the AI searches the design space and identifies configurations that meet the desired optical and electrical criteria.

In practical terms, this technology offers architects and product designers a new degree of freedom in integrating photovoltaics into visible surfaces. Building facades could use colored solar glazing that aligns with an overall design scheme while harvesting energy from a wide area of glass. Automotive applications might include tinted solar roofs or windows that blend with a vehicle’s styling, providing auxiliary power without altering its appearance. Consumer electronics and wearable devices could incorporate semitransparent power-generating layers that match brand colors or aesthetic themes while remaining functional and unobtrusive.

The researchers also highlight potential benefits for greenhouses and other agricultural or horticultural structures where both light transmission and spectral control matter. By tuning the color and transparency of the solar windows, it may be possible to maintain suitable illumination for plant growth while still generating electricity. More broadly, the work points to a route for designing photonic structures that balance appearance, transparency and energy capture, helping remove an important barrier to widespread adoption of building-integrated photovoltaics.

The paper notes that this AI enabled design strategy fits into a wider portfolio of research on photonic materials and devices. The group led by Prof. Sun-Kyung Kim has previously demonstrated three dimensional silicon grating nanowire photovoltaic systems that achieved record power conversion efficiencies at the single nanowire level. They have also developed strong diffraction hollow cavity growth substrates that support high efficiency InGaN/GaN light emitting diodes, outperforming state of the art commercial devices. In the field of thermal radiation control, they reported directional radiative coolers that enhance side thermal emissions to improve thermal comfort for users of personal optoelectronic devices such as smartphones.

Across these efforts, the common theme is the use of high index contrast dielectric and metal dielectric hybrid photonic structures to manage light absorption, emission and thermal radiation across a wide spectral range from ultraviolet to microwave. The current work on semitransparent perovskite solar cells extends these concepts into the domain of building-integrated photovoltaics, where visual appearance is as important as performance. By leveraging metamaterials, surface plasmon effects and advanced interference optics in conjunction with AI based inverse design, the team shows that device engineers no longer need to choose between color, transparency and efficiency.

Research Report:Modelling-guided inverse design strategy for semitransparent perovskite photovoltaics with customized colors