WITHDRAWN: Machine Learning-Assisted High-Throughput Screening of Transparent Organic Light-Emitting Diodes Anode Materials

Abstract The development of data-driven scientific paradigms has revolutionized the discovery of two-dimensional (2D) functional nanomaterials. Owing to their high potential in organic light-emitting diode (OLED) displays, the discovery of 2D nanomaterials with suitable work functions is crucial for optimizing device performance. This study proposed a target-driven material design framework, strategically embedding interpretable machine learning (ML) within high-throughput virtual screening (HTVS), for accelerating the identification of novel OLED anodes materials with suitable work functions and excellent photoelectric performance. Based on CatBoost algorithm and the C2DB database, an ML regression model for predicting the work functions of 2D nanomaterials was established, achieving fast (≈ 10³ faster than density functional theory (DFT)) and accurate predictions with an accuracy of 92%. Furthermore, the SHapley Additive exPlanations (SHAP) method was used to confirm the key factors determining the work function of 2D nanomaterials from both structure-based and a series of composition-based features at the atomic scale. Our methodology, combining robust ML models, multi-condition screening, and DFT calculations, has identified a promising 2D nanomaterial—PS. The material demonstrates exceptional conductivity (σ > 10⁶ S/m), high transparency (transmittance > 85%), and favorable work functon(> 5 eV), significantly outperforming the commonly used indium tin oxide (ITO). As a result, PS emerges as a potential candidate for transparent OLED anodes, informing future experimental synthesis efforts. This study offers insights into the discovery of transparent OLED anode materials and provides a cost-effective design framework for identifying high-performance materials across various chemical fields.