AI Tunes Perovskite Solar Windows to Any Hue

Mar 3, 2026 10:50 AM ET
  • AI inverse-designed ZnS/MgF2 coatings paint perovskite solar cells in user-defined colors with minimal loss, boosting PCE on glass and PET—opening sharper-colored solar windows and auto glazing.
AI Tunes Perovskite Solar Windows to Any Hue

Researchers at Kyung Hee University and Hyundai Motor Group used AI inverse design to add all‑dielectric ZnS/MgF2 multilayer coatings to semitransparent perovskite cells, delivering user‑defined colors with minimal optical loss. A factorization‑machine surrogate and QUBO optimization chose layer sequences, achieving six target hues across varying absorber thicknesses.

A cyan device with 110 nm MAPbI3 absorber reached 10.4% transmittance and 6.5% AVT on glass, raising PCE 20.9%; on PET it lifted PCE 10.4%. Coatings were thermally evaporated on cells or laminated from pre‑coated PET, and framework extends to thin films; higher‑bit encoding could sharpen hues for solar windows and vehicle glazing.

What scalability, durability, and color-uniformity hurdles remain for AI-designed perovskite coatings?

Scalability – manufacturing and cost
- Tight thickness control over multi‑layer dielectric stacks across square‑meter rolls; managing web tension, edge effects, and run‑to‑run drift in roll‑to‑roll lines.
- Throughput limits of vacuum deposition versus demand for high‑area, multi‑color product families; tool uptime, target utilization, and cycle time.
- Yield loss from pinholes/particulates in 10–100 nm films; need for inline, high‑speed ellipsometry/colorimetry and closed‑loop feedback.
- Registration of coatings to cell layouts and busbars on curved or flexible substrates; patterning without damaging perovskites.
- Supply chain and cost of high‑purity dielectrics; scalability of AI workflows from small design spaces to many SKUs and custom hues.
- Lamination or transfer steps at scale without bubbles, wrinkles, or optical adhesives adding haze or color shift; recyclability and rework.

Durability – field reliability and standards
- Thermo‑mechanical mismatch between brittle dielectrics and soft perovskite/encapsulants causing microcracks under thermal cycling and flex.
- Moisture/oxygen ingress at layer interfaces; adhesion aging and edge‑seal robustness under damp‑heat and freeze–thaw.
- UV exposure leading to polymer yellowing, dielectric densification, and perovskite/transport‑layer degradation; maintaining color and PCE.
- Ion migration and halide segregation shifting the absorber’s spectrum, breaking the color match over time.
- Abrasion and cleaning chemicals in building and automotive use; anti‑soiling topcoats that don’t perturb hue.
- Compliance with IEC/UL PV weathering, automotive glazing (impact, wiper, thermal shock), and fire/smoke standards.

Color uniformity – aesthetics and optical control
- Angle‑dependent interference causing iridescence; achieving low color shift over wide viewing cones for windows and windshields.
- Spatial thickness non‑uniformity across large areas leading to visible hue gradients; substrate roughness and TCO texture imprinting scatter.
- Batch‑to‑batch matching across absorber thickness variations and halide ratios; metamerism under different illuminants (D65 vs warm LED).
- Aging‑induced drift from perovskite or encapsulant changes; maintaining uniformity after bending, lamination, and thermal cycling.
- Edge and busbar regions darkening or hue fringes; seamless tiling between modules without visible seams.
- Quantization limits from discrete layer counts/material sets; need for higher‑bit designs without increasing parasitic absorption.