Machine Learning-Aided Design of Gold Core-Shell Nanocatalysts toward Enhanced and Selective Photooxygenation

Area of Science:

• Materials Science
• Catalysis
• Computational Chemistry

Background:

• Designing plasmonic nanoparticles for enhanced chemical reactions is crucial.
• Predicting electric field enhancement around nanoparticles is complex.
• Optimizing singlet oxygen (¹O₂) generation is key for organic synthesis.

Purpose of the Study:

• To utilize machine learning (ML) for predicting electric fields around core-shell nanoparticles.
• To guide the design of gold-silica nanoparticles for improved ¹O₂ sensitization.
• To enhance selectivity and reaction rates in organic synthesis.

Main Methods:

• Deep neural network (DNN) algorithms for feature importance analysis.
• Development of a linear descriptor (θ ∝ aD⁰.²⁵t⁻¹) for electric field prediction.
• Synthesis and characterization of gold-silica nanoparticles using STEM-EELS mapping.
• Time-dependent density functional theory (TD-DFT) calculations.

Main Results:

• Identified a linear descriptor for electric field prediction around core-shell plasmonic nanoparticles.
• Synthesized nanoparticles with optimized plasmonic intensity (θ = 0.40).
• Achieved ~3-fold increase in reaction rate for anthracene photooxygenation.
• Improved selectivity by 4% for dihydroartemisinic acid (DHAA) photooxygenation.
• Demonstrated synergetic effects of plasmonic enhancement and fluorescence quenching for ¹O₂ generation.

Conclusions:

• ML tools offer rapid and accurate prediction of electric fields for catalyst design.
• The developed descriptor effectively guides the synthesis of high-performance photocatalysts.
• This data-driven strategy significantly reduces experimental costs and accelerates catalyst development for organic synthesis.