A neural network parametrized coagulation rate model for <3 nm titanium dioxide nanoclusters

Area of Science:

• Materials Science
• Chemical Engineering
• Computational Chemistry

Background:

• Coagulation significantly impacts nanocluster size distribution during high-temperature metal oxide nanomaterial synthesis.
• Existing population balance models rely heavily on coagulation rate coefficients, which are complex for nanoscale particles.
• Accurate modeling of nanocluster coagulation is crucial for predicting nanoparticle growth and properties.

Purpose of the Study:

• To develop an improved model for titanium dioxide (TiO2) nanocluster coagulation rate coefficients.
• To utilize molecular dynamics (MD) simulations to train a neural network (NN) for predicting coagulation outcomes.
• To enhance population balance models for TiO2 nanoparticle formation.

Main Methods:

• Conducted MD trajectory calculations for TiO2 nanoclusters (0.6-3.0 nm diameters) with varying velocities and impact parameters.
• Incorporated dipole-dipole, dispersion, and repulsive interactions in MD simulations.
• Trained a NN model to predict coagulation based on nanocluster size, velocity, and impact parameter, achieving >95% accuracy.

Main Results:

• The NN model accurately predicts TiO2 nanocluster coagulation probabilities.
• NN predictions for coagulation rate coefficients at 300 K and 1000 K closely match MD results (within 0.65-1.54 range).
• The NN model successfully captures the local minimums in coagulation rate coefficients observed in MD simulations.

Conclusions:

• A validated NN model can accurately predict TiO2 nanocluster coagulation rate coefficients.
• This NN model offers a computationally efficient tool for improving population balance models in TiO2 synthesis.
• The findings contribute to a better understanding and control of nanomaterial formation processes.