Modeling Electrosprayed Colloidal Particle Assembly at Curved Droplet Interfaces Integrating Physics and Machine Learning

The assembly of charged colloidal particles into monolayers at liquid interfaces offers promising advancements in fabricating thin-film materials and devices with customizable properties. This study explored the behavior of electrosprayed colloidal particles at curved droplet interfaces using a combination of physics-based simulations and machine-learning techniques. We implemented a mesh-constrained Brownian dynamics (BD) algorithm integrated with ANSYS electric field solver to model the movement and assembly of charged particles on non-spherical droplet surfaces. Our findings highlight electrostatic repulsion, electrophoretic forces from substrate surface charges, and Brownian motion as important factors shaping the compactness and structural order of the particle assemblies. Additionally, we trained a surrogate model with the artificial neural network using data from BD simulations to predict the radial distribution functions of the particle assembly. This surrogate model alongside Bayesian optimization, was utilized to identify optimal particle and substrate charge densities that best match the simulation assembly with experimental data. Insights from the inferred charge densities can shed light on surface charge accumulation in the electrospray process.