Software and Methods for Computational Inverse Design of Soft Materials with Physical Relevance

Abstract

Soft materials are used in a wide range of applications, but they remain difficult to design because their properties emerge from coupled molecular interactions and struc- tures. This dissertation develops and applies computational methods for inverse design of soft materials, with an emphasis on physically meaningful models, reproducible work- flows, and practical simulation tools. First, open-source Python software is developed to make molecular-simulation workflows for optimization more transparent, reusable, and extensible. Second, surrogate models based on Chebyshev interpolation and Smolyak sparse grids are used to reduce the cost of relative-entropy-based inverse design for low- dimensional interaction potentials. Third, relative-entropy minimization is used to coarse- grain star poly(ethylene glycol) into simpler bead-spring models. Finally, molecular sim- ulations are used to determine how polymer architecture affects axial dispersion in mi- crochannels. Together, these studies provide practical approaches for modeling, design- ing, and understanding soft materials using molecular simulation.