Machine-Learning-Optimized Perovskite Nanoplatelet Synthesis

TL;DR

This paper demonstrates how combining machine learning models with Bayesian Optimization can efficiently optimize the synthesis of perovskite nanoplatelets, achieving high-quality results with limited experimental data.

Contribution

The study introduces a novel approach merging three machine-learning models with Bayesian Optimization to improve nanoplatelet synthesis with minimal data.

Findings

Achieved high-quality CsPbBr3 nanoplatelets with only ~200 syntheses.

Predicted photoluminescence emission maxima based on precursor ratios.

Enabled synthesis of previously unattainable 7 and 8 monolayer nanoplatelets.

Abstract

With the demand for renewable energy and efficient devices rapidly increasing, a need arises to find and optimize novel (nano)materials. This can be an extremely tedious process, often relying significantly on trial and error. Machine learning has emerged recently as a powerful alternative; however, most approaches require a substantial amount of data points, i.e., syntheses. Here, we merge three machine-learning models with Bayesian Optimization and are able to dramatically improve the quality of CsPbBr3 nanoplatelets (NPLs) using only approximately 200 total syntheses. The algorithm can predict the resulting PL emission maxima of the NPL dispersions based on the precursor ratios, which lead to previously unobtainable 7 and 8 ML NPLs. Aided by heuristic knowledge, the algorithm should be easily applicable to other nanocrystal syntheses and significantly help to identify interesting compositions and rapidly improve their quality.