Process Insights into Perovskite Thin‐Film Photovoltaics from Machine Learning with In Situ Luminescence Data

Large-area processing remains a key challenge for perovskite solar cells (PSCs). Advanced understanding and improved reproducibility of scalable fabrication processes are required to unlock the technology’s economic potential. In this regard, machine learning (ML) methods have emerged as a promising tool to accelerate research and unlock the control needed to produce large-area solution-processed perovskite thin-films. However, a suitable dataset allowing the analysis of a scalable fabrication process is currently missing. In this work, a unique labeled in situ photoluminescence (PL) dataset for blade-coated PSCs is introduced and explored with unsupervised k-means clustering, demonstrating the feasibility to derive meaningful insights from such data. Correlations between the obtained clusters and the measured performance of PSC reveal that the in situ PL signal encodes information about the perovskite thin-film quality. Detrimental mechanisms during thin-film formation are detected by identifying spatial differences in PL patterns and, consequently, of device performance. In addition, k-nearest neighbors is applied to predict the performance of PSCs, motivating further investigations into ML-based in-line process monitoring of scalable PSC fabrication to detect, understand, and ultimately minimize process variations across iterations.