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

Area of Science:

• Materials Science
• Nanotechnology
• Chemical Engineering

Background:

• Optimizing carbon dot (CD) synthesis requires understanding complex reaction parameter correlations.
• Large, noisy datasets hinder traditional experimental screening for high-performance CDs.
• Machine learning (ML) offers a powerful approach for materials discovery and process optimization.

Purpose of the Study:

• To demonstrate ML's capability in predicting, optimizing, and accelerating CD synthesis.
• To establish a regression ML model for hydrothermal-synthesized CDs.
• To identify key synthesis parameters influencing CD properties, particularly fluorescent quantum yield (QY).

Main Methods:

• Development of a regression ML model using hydrothermal synthesis data for CDs.
• Utilizing the ML model to identify critical synthesis parameters affecting CD properties.
• Experimental verification guided by ML predictions to synthesize high-QY CDs.
• Application of synthesized CDs as a fluorescence probe for Fe³⁺ ion detection.

Main Results:

• An ML model was established, revealing relationships between synthesis parameters and CD outcomes.
• High-QY CDs (up to 39.3%) with strong green emission were synthesized via ML guidance.
• Precursor mass and alkaline catalyst volume were identified as key features for high-QY CD synthesis.
• The synthesized CDs demonstrated ultrasensitive linear detection of Fe³⁺ ions (0-150 μM, LOD 0.039 μM).

Conclusions:

• ML significantly enhances the prediction, optimization, and acceleration of CD synthesis.
• The developed ML model effectively guides the synthesis of high-quality CDs with superior properties.
• ML-driven CD synthesis accelerates the development of intelligent materials for applications like ion sensing.