Synthesizer: Chemistry-Aware Machine Learning for Precision Control of Nanocrystal Growth.

Area of Science:

• Materials Science
• Nanotechnology
• Artificial Intelligence

Background:

• Precise control over nanocrystal synthesis is crucial for tuning optical properties in halide perovskites.
• Current methods face challenges in achieving reproducible and tailored results.
• Developing adaptable frameworks for nanocrystal optimization is essential.

Purpose of the Study:

• To introduce a broadly adoptable, machine learning-guided framework for nanocrystal synthesis optimization.
• To demonstrate nm-level precision in tuning optical properties of halide perovskites.
• To establish a practical platform for accelerating materials discovery.

Main Methods:

• Utilized Gaussian Process regression and Bayesian optimization within the Synthesizer framework.
• Employed chemistry-aware molecular encodings and systematic feature engineering.
• Translated interpretable machine learning tools into a benchtop platform for ambient condition optimization.

Main Results:

• Achieved nm-level precision in photoluminescence peak tuning (430 nm to 520 nm) for CsPbBr3 nanocrystals.
• Obtained benchmark narrow linewidths down to 70 meV via lateral confinement control.
• Demonstrated robust photoluminescence quantum yield optimization and identified key mechanistic parameters.

Conclusions:

• The Synthesizer provides an adoption-ready platform for data-efficient, uncertainty-aware synthesis design.
• The framework enables reproducible pathways for accelerating materials discovery in halide perovskites and beyond.
• Generalizability confirmed across different chemical spaces and material systems (CsPbI3).