Researchers from institutions including Hong Kong Polytechnic University, the Swiss Federal Institute of Technology Lausanne (EPFL), the University of Oxford, Peking University, and Zhejiang University have developed a fully autonomous, closed-loop AI–robotics platform for perovskite solar cell fabrication. The study was published in the journal Engineering and represents one of the most comprehensive high-throughput solar cell research efforts to date.
At the heart of the system is a Recipe Language Model (RLM) trained on approximately 60,000 perovskite-related scientific publications. A seven-layer AI architecture handles everything from recipe generation and dataset construction to reasoning, evaluation, and optimization — with each experimental cycle feeding structured data back into the model to continuously improve future recommendations.
Physical fabrication is carried out by 11 robotic boxes encompassing 101 functional units, more than 1,500 components, and over 4,300 controllable parameters. The first three boxes manage chemical storage and dispensing, while boxes 4–11 perform spin-coating, laser processing, thermal annealing, vacuum thin-film deposition, and in situ characterization using integrated cameras and sensors. The system also generates a real-time digital twin of every process.
The platform progressed through four optimization stages: an initial exploratory phase yielded efficiencies between 0% and 17.4%; introducing self-assembled monolayers (SAMs) and additives pushed this to ~23%; surface passivation reached 25.6%; and the final configuration achieved 27.0% (certified at 26.5%). In total, 50,764 devices were fabricated and tested during the study.
Perovskite solar cells are widely regarded as the most promising next-generation photovoltaic technology, with the potential to surpass conventional silicon panels in both cost and performance. Automating the optimization process at this scale significantly shortens the research-to-production pipeline, making commercial high-efficiency perovskite modules a more tangible near-term goal for the global solar industry.
Source: Fabricating perovskite solar cells with robotic boxes - PV Magazine International· Based on source, with AI-assisted rewriting.
Related articles

How Much Do Solar Panels Produce in Winter? Realistic Numbers
Winter output from a rooftop solar system is a fraction of the summer peak: shorter days, a low sun, cloud and snow mean that the four darkest months together typically deliver only around a tenth to a fifth of the annual yield, depending on latitude. This guide sets realistic month-by-month expectations, explains what winter means for self-consumption and your bill, shows how a battery and smart scheduling help, and why you should never size a system for December.

Pre-Season Heat Pump Checklist: What to Check in September
A heat pump does most of its work in the coldest months, but how much electricity it burns doing so is largely decided in September. Here is what homeowners can check themselves, from the outdoor unit to the heating curve, what to leave to a qualified engineer, and how to get the most out of rooftop solar in the autumn months.

Air conditioning on solar power: what summer cooling really costs in 2026
In a heatwave, air conditioning becomes one of the biggest loads in a household — exactly when solar panels produce the most. We look at what cooling costs with and without solar, how much extra capacity you need, and what to watch for when sizing a system.

Heat Pump, Solar & H Tariff: How to Plan Together
Combining a heat pump, rooftop solar panels and a preferential electricity tariff can significantly cut household energy bills — but only if you plan all three as one integrated system. This guide walks through the decision logic, common mistakes and key considerations for homeowners.
Comments
0 commentsBe the first to comment.
