Data-Driven Inverse Design of Silver Nanoparticle Size for Controlled Synthesis Across Multiple Systems Using...

Area of Science:

• Nanotechnology
• Materials Science
• Computational Chemistry

Background:

• Controlling nanoparticle size is crucial for tailored applications.
• Traditional methods for silver nanoparticle (AgNP) synthesis often lack precision.
• Data-driven approaches offer potential for optimizing synthesis parameters.

Purpose of the Study:

• To develop and evaluate a data-driven inverse-design framework for AgNP synthesis.
• To identify synthesis conditions yielding specific AgNP sizes.
• To compare the performance of various conditional generative models.

Main Methods:

• Developed a forward surrogate model to map synthesis parameters to particle size.
• Coupled the forward model with conditional generative adversarial networks (cGAN, cWGAN, cWGAN-GP) and other models.
• Evaluated models using three AgNP datasets from microfluidic and chemical synthesis routes.

Main Results:

• Conditional adversarial models outperformed non-adversarial baselines.
• cWGAN demonstrated consistent performance across datasets; cGAN was effective in complex cases.
• The framework captured the one-to-many nature of inverse design, generating multiple synthesis conditions for a target size.
• Prediction errors increased near the feasible size interval boundaries.

Conclusions:

• The proposed framework enables target-oriented AgNP synthesis design.
• Forward-constrained conditional generative modeling is suitable for limited-data settings.
• Inverse design challenges are more pronounced at the boundaries of attainable particle sizes.