Deep-learning-enhanced modeling of electrosprayed particle assembly on non-spherical droplet surfaces

Area of Science:

• Colloid and Surface Science
• Computational Materials Science
• Additive Manufacturing

Background:

• Monolayer assembly of charged colloidal particles at liquid interfaces is crucial for developing advanced thin film materials and devices.
• Understanding particle dynamics at curved interfaces is essential for controlling assembly structures.

Purpose of the Study:

• To investigate the dynamics of electrosprayed colloidal particles at curved droplet interfaces.
• To develop a computational framework combining physics-based simulations and machine learning for predicting particle assembly.
• To identify optimal parameters for achieving high-fidelity particle assembly.

Main Methods:

• Utilized a mesh-constrained Brownian dynamics (BD) algorithm coupled with Ansys® electric field simulations.
• Employed deep neural network surrogate modeling to predict radial distribution functions (RDF) of particle assembly.
• Integrated surrogate modeling with Bayesian optimization to determine optimal particle and substrate charge densities.

Main Results:

• Electrostatic repulsion, electrophoretic forces, and Brownian motion were identified as key factors governing assembly structure.
• The deep learning model accurately predicted RDF profiles, achieving 96.4% similarity with experimental data.
• Optimal charge densities were identified, leading to simulated assembly differing by less than 5% in average bond order parameter from experiments.

Conclusions:

• A deep learning-based approach significantly reduces computational cost while maintaining high accuracy in predicting colloidal assembly.
• The inferred charge densities offer critical insights into surface charge accumulation during electrospray processes.
• This methodology advances additive manufacturing of thin films with tailored properties through precise control of particle assembly.