Computational modeling and design of self-stratifying colloidal materials

Abstract

Self-stratification is a self-assembly process in which two or more types of colloidal particles dispersed in a solvent spontaneously separate into layers as the solvent is removed by drying. Self-stratification can enable fabrication of not only multilayered materials but also a variety of other continuous gradient materials with tailored properties and functions, all in a single processing step. However, self-stratification is sensitive to numerous physicochemical properties of the colloidal particles and the processing conditions. Experimental data on self-stratification are sparse and time-consuming to collect, and as a result, self-stratification is poorly understood.

I used computational modeling to understand the dynamics in these drying processes and then designed self-stratifying colloidal materials. Existing models for self-stratification are computationally expensive or not sufficiently accurate. I developed a continuum model for self-stratification based on dynamic density functional theory (DDFT), which is a mesoscale model that predicts how particle concentrations evolve from the particle-level interactions and dynamics. DDFT has two key inputs: a free-energy functional and a model for the hydrodynamic interactions (HI). For the first input, I compared the different approximations of the free-energy functional for inhomogeneous systems and evaluated their accuracy in simulating drying colloidal films. I found that fundamental measure theory (FMT) provides a good balance between performance and accuracy compared to other approximations when simulating drying suspensions. For the second input, I compared the different approximations of HI and incorporated the appropriate approximation in the DDFT framework to simulate drying droplets. I found pairwise HI are important when simulating these structures and also highlight the challenges in incorporating it in the DDFT framework. I then developed an inverse design strategy to couple simulations with optimization strategies to design self-assembled coatings with targeted particle distributions. I demonstrate that surrogate models trained on particle-based simulations can be used to predict parameters to achieve different target self-assembled particle distributions of dried colloidal dispersions.