講演情報
[17a-PA1-20]Machine Learning Analysis of Photoluminescence Properties in Cesium Lead Bromide Perovskite Nanocrystals
〇Yoshua Albert Darmawan1, Shinichi Fujiwara1, Qing Shen2, Kenji Katayama1 (1.Chuo Univ., 2.Univ. of Electro-Commun.)
キーワード：
machine learning、perovskite、nanocrystal
Lead halide perovskite nanocrystals exhibit high photoluminescence quantum yield (PLQY) and are promising materials for optoelectronic applications; however, their photoluminescence properties often show significant sample-to-sample variation even under identical synthesis conditions. In this study, a machine learning approach combined with Shapley Additive exPlanations (SHAP) analysis was applied to elucidate the factors governing photoluminescence behavior in cesium lead bromide perovskite nanocrystals. Structural and optical descriptors extracted from X-ray diffraction, UV-vis absorption, Fourier-transform infrared spectroscopy, and photoluminescence measurements were analyzed using our recently developed ML method suitable for small number of data. The model achieved a high R2 value of 0.825 for PLQY prediction. SHAP analysis revealed that a small number of dominant descriptors, mainly associated with surface passivation conditions and structural properties, contribute most strongly to PLQY variation. These results indicate that the apparent complexity of photoluminescence variability originates from a limited set of governing factors, demonstrating the usefulness of interpretable machine learning for extracting mechanistic insight from experimental nanomaterial data.