Self-consistent field theory (SCFT), the mean-field theory of polymer thermodynamics, is a powerful tool for understanding ordered state selection in block copolymer melts and blends. However, the nonlinear governing equations pose a significant challenge when SCFT is used for phase discovery because converging an SCFT solution typically requires an initial guess close to the self-consistent solution. This Viewpoint provides a concise overview of recent efforts where machine learning methods (particle swarm optimization, Bayesian optimization, and generative adversarial networks) have been used to make the first strides toward converting SCFT from a primarily explanatory tool into one that can be readily deployed for phase discovery.
Read full abstract