Researchers at LETI Develop Neural Network to Detect Defects in Silicon Solar Cells
Experts from St. Petersburg State Electrotechnical University “LETI” have created a neural network capable of identifying defects in the production of solar cell components. LETI shared details about this development with socialbites.ca.
Solar energy ranks among the most promising sources of renewable power today. Solar cells are typically made from silicon, a widely available and affordable semiconductor material. Even though modern silicon cell manufacturing is largely automated, errors can occur as silicon wafers move between different processing steps on the production line.
Defects on silicon wafers are detected through the electroluminescence effect. When a voltage is applied to a silicon solar cell, its surface emits light. Infrared imaging captures this glow, revealing a uniform emission in defect-free cells. Any irregular brightness patterns indicate the presence of defects, signaling areas that need attention.
A large database of infrared images was compiled from more than 100,000 silicon solar cells produced at Hevel LLC in Novocheboksarsk. Using this data, researchers trained a neural network to recognize and classify defects that can occur at various stages of solar cell production with high accuracy.
Defects in production occur at about 0.2–0.3 percent of units. While that level may seem small, it has a disproportionate impact on solar cell efficiency, especially as manufacturers push for higher performance. The analytics tool developed by LETI enables not only defect detection but also pinpointing where defects arise in the production line. With this insight, manufacturers can take corrective actions to reduce losses and improve overall efficiency in their high-efficiency silicon solar cell operations. The lead researcher, Evgeniy Terukov, a professor at the Photonics Department of LETI and Deputy Director of the Scientific and Technical Research Center for Thin Film Technologies in Energy, explained that the new method enhances production reliability for enterprises like Hevel LLC.
In addition to the solar cell initiative, LETI has previously worked on rapidly configurable microwave antennas designed for drone control, reflecting a broader portfolio of advanced materials and sensing technologies.