Utilizing machine learning to model interdependency of bulk molecular weight, solution concentration, and thickness of spin coated polystyrene thin films

Spin coating is a quick and inexpensive method to create nanometer-thick thin films of various polymers on solid substrates. Since the film thickness determines the mechanical, optical, and degradation properties of the coating, it is essential to develop a simple method to predict thickness based on other manipulatable factors. In this study, a three-dimensional manifold simultaneously relating initial solution concentration, film thickness, and monodisperse bulk molecular weight is developed utilizing curve-fit machine learning on a dataset of spin coated polystyrene samples. Given values for any two of the three factors, the manifold presents an accurate corresponding value for the unknown.