Exploration and Design of Carbon Dot-Based Long Afterglow Materials Using Active Machine Learning and Quantum Chemical Simulations

Long afterglow materials based on carbon dots (CDs) have attracted extensive attention in the field of optics due to their low cost and nontoxic properties. However, the targeted synthesis of specific properties of complex and unknown structures such as CDs remains a daunting challenge. In this study, the powerful nonlinear fitting ability of machine learning was used to explore the afterglow properties of CDs. The XGBoost algorithm demonstrates high prediction accuracy in determining the optimal excitation wavelength, optimal emission wavelength, and afterglow lifetime. Using Bayesian optimization, we screened and synthesized the CDs-based long afterglow materials with the longest lifetime reported so far by a one-step microwave method. By combining quantum chemical calculations with experimental data, we revealed the structure-function relationship between CDs and their precursors through electron-hole analysis. These results show that machine learning can establish nonlinear correlations between precursors and materials with unknown structures, clarify their intrinsic relationships, simplify the material design process, and thus accelerate the development of advanced materials.