Rapid Data-Efficient Optimization of Perovskite Nanocrystal Syntheses through Machine Learning Algorithm Fusion.

Area of Science:

• Materials Science
• Nanotechnology
• Chemical Engineering

Background:

• Growing demand for renewable energy and efficient devices necessitates novel nanomaterials.
• Traditional material synthesis relies on trial-and-error, which is time-consuming and inefficient.
• Machine learning (ML) offers potential for synthesis optimization but often requires extensive data.

Purpose of the Study:

• To develop a data-efficient ML approach for optimizing nanomaterial synthesis.
• To apply the method to the synthesis of cesium lead bromide (CsPbBr3) nanoplatelets.
• To demonstrate rapid optimization of material properties and synthesis processes.

Main Methods:

• Integration of three ML models with Bayesian optimization for low-data synthesis optimization.
• Utilizing precursor ratios as input parameters to predict photoluminescence emission maxima.
• Fabrication and characterization of CsPbBr3 nanoplatelets.

Main Results:

• Accurate prediction of photoluminescence emission maxima for CsPbBr3 nanoplatelet dispersions.
• Successful fabrication of previously unattainable 7- and 8-monolayer-thick nanoplatelets.
• Significant improvement in the homogeneity of 2-6-monolayer-thick nanoplatelets, indicated by narrower spectra.
• Optimization achieved with only 200 total syntheses.

Conclusions:

• The developed ML-Bayesian optimization algorithm significantly accelerates nanomaterial synthesis and optimization.
• The approach enables the production of high-quality, tailored nanomaterials with limited data.
• The versatile algorithm is applicable to a wide range of nanocrystal syntheses, facilitating materials discovery.