Machine learning-guided realization of full-color high-quantum-yield carbon quantum dots

Area of Science:

• Materials Science
• Nanotechnology
• Machine Learning Applications

Background:

• Carbon quantum dots (CQDs) offer versatile luminescence applications.
• Optimizing CQD synthesis is complex due to numerous parameters and desired properties.
• Traditional methods involve extensive trial-and-error, limiting research efficiency.

Purpose of the Study:

• To develop a machine learning (ML)-guided strategy for optimizing CQD hydrothermal synthesis.
• To reduce the experimental search space and accelerate the discovery of CQDs with desired properties.
• To establish a closed-loop optimization approach for synthesizing multifunctional CQDs.

Main Methods:

• Utilized a novel multi-objective optimization strategy powered by a machine learning algorithm.
• Implemented a closed-loop system that learns from limited and sparse experimental data.
• Developed a unified objective function to simultaneously optimize photoluminescence (PL) wavelength and PL quantum yield (PLQY).

Main Results:

• Achieved the synthesis of full-color fluorescent CQDs using only 63 experiments.
• Attained high PLQY exceeding 60% across all synthesized colors.
• Successfully revealed complex relationships between synthesis parameters and CQD properties.

Conclusions:

• The ML-guided approach significantly accelerates CQD synthesis compared to traditional methods.
• This strategy enables the efficient development of CQDs with multiple, precisely controlled properties.
• Represents a significant advancement in ML-assisted materials synthesis for future applications.