Utilizing a SHAP-Assisted Machine Learning Algorithm for Optimization of Plasmonic Properties of Gold Nanostructure

Area of Science:

• Nanotechnology
• Materials Science
• Computational Chemistry

Background:

• Gold nanostructures (GNSs) offer tunable optical and catalytic properties crucial for molecular sensing and nanomanufacturing.
• Conventional trial-and-error methods hinder the efficient design of GNSs with desired optical characteristics.

Purpose of the Study:

• To develop a predictive model for the synthesis of GNSs with specific plasmonic features.
• To elucidate the relationship between synthetic conditions and GNS plasmonic properties using machine learning.

Main Methods:

• Utilized a SHapley Additive exPlanations (SHAP)-assisted extreme gradient boosting (XG-Boost) model.
• Investigated the synthesis of GNSs using silver nitrate and ascorbic acid in various surfactants at room temperature.
• Employed SHAP to interpret model predictions and identify key influencing factors, such as surfactant contributions.

Main Results:

• The XG-Boost model demonstrated high predictive accuracy (R² > 0.98, RMSE < 0.03).
• SHAP analysis revealed the significant, previously underestimated, role of surfactants in determining GNS plasmonic properties.
• Successfully designed and synthesized GNSs with tunable localized surface plasmon resonance peaks and enhanced stability.

Conclusions:

• The developed machine learning framework significantly improves the design and synthesis of GNSs.
• The SHAP-XG-Boost strategy enables precise tailoring of GNSs for diverse applications, including molecular detection and luminescent chip fabrication.
• This approach highlights the potential of AI in materials science for accelerating the development of functional nanomaterials.