摘要: The utilization of phosphorescent metal complexes as emissive dopants for organic light-emitting diodes (OLEDs) has been the subject of intense research. Cyclometalated Pt(II) complexes are particularly popular triplet emitters due to their color-tunable emissions. To make them viable for practical applications as OLED emitters, it is essential to develop Pt(II) complexes with high radiative decay rate constants (k_r) and photoluminescence quantum yields (PLQY). To this end, an efficient and accurate prediction tool is highly desirable. In this work, we propose a general yet powerful protocol achieving metal complex generation, high throughput virtual screening (HTVS), and fast predictions with high accuracy. More than 3600 potential structures are generated in a synthesis-friendly manner. Moreover, three HTVS-machine learning (ML) models are established using different algorithms with carefully designed features that are suitable for metal complexes. Specifically, 30 potential candidates are filtered out by HTVS-ML models with a three-tier screening rule and put into accurate predictions with experimental calibration Δ-learning method. The highly accurate prediction approach further reduces the stress of experiments and inspires greater confidence in identifying the most promising complexes as excellent emitters. As a result, 12 promising complexes (k_r > 10⁵ s^-1 and PLQY > 0.6) with the same superior core structures are confirmed from over 3600 Pt-complexes. Experiments revealed that two very close complexes have excellent emission properties and are consistent with the prediction results, providing strong evidence for the efficacy of the proposed protocol. We expect this protocol will become a valuable tool, expediting the rational design and rapid development of novel OLED materials with desired properties.