CBE Doctoral Dissertation Defense: “Polymer Phase Behavior Simulation Study via Deep Learning and Binding-Induced Polarization” (Dongqi Zhao)

Abstract:
Polymer materials show rich phase behavior that is difficult to predict because of their structural and interaction complexity. This work develops computational tools that improve both the efficiency and physical realism of polymer modeling. A deep-learning-accelerated SCFT method is introduced that predicts density fields directly from mean-field potentials, effectively replacing the expensive propagator calculation and enabling large speedups while maintaining accuracy. Particle-based models are developed to describe reversible Lewis-type bonding and the polarization that arises when these bonds form. These models clarify how tunable bond strength and induced dipoles influence phase separation in homopolymer mixtures. Extending the approach to block copolymers shows that reversible bonding can reorganize microdomain structure, shift phase boundaries, and alter interfacial composition. Taken together, these methods provide faster simulation strategies and new insight into how dynamic interactions shape polymer phase behavior.