Machine-Learning-Driven Synthesis of Carbon Dots with Enhanced Quantum Yields.

Abstract

Knowing the correlation of reaction parameters in the preparation process of carbon dots (CDs) is essential for optimizing the synthesis strategy, exploring exotic properties, and exploiting potential applications. However, the integrated screening experimental data on the synthesis of CDs are huge and noisy. Machine learning (ML) has recently been successfully used for the screening of high-performance materials. Here, we demonstrate how ML-based techniques can offer insight into the successful prediction, optimization, and acceleration of CDs' synthesis process. A regression ML model on hydrothermal-synthesized CDs is established capable of revealing the relationship between various synthesis parameters and experimental outcomes as well as enhancing the process-related properties such as the fluorescent quantum yield (QY). CDs exhibiting a strong green emission with QY up to 39.3% are obtained through the combined ML guidance and experimental verification. The mass of precursors and the volume of alkaline catalysts are identified as the most important features in the synthesis of high-QY CDs by the trained ML model. The CDs are applied as an ultrasensitive fluorescence probe for monitoring the Fe3+ ion because of their superior optical behaviors. The probe exhibits the linear response to the Fe3+ ion with a wide concentration range (0-150 μM), and its detection limit is 0.039 μM. Our findings demonstrate the great capability of ML to guide the synthesis of high-quality CDs, accelerating the development of intelligent material.