University of Akron,
Keywords: informatics, machine learning
Summary:The last decade has seen the emergence of a new frontier in polymer chemistry: the facile synthesis of copolymers with specific monomeric sequences. This advance may be the key to realizing synthetic polymers exceeding the performance of biopolymers such as proteins and DNA. Achieving this goal will require solutions to two challenges. First, the design space afforded by even a sequence specific polymer can be astronomical, or even Avogadroan, in scale. How are we to reproduce, on a human timescale, the billions of years nature has dedicated to designing these complex molecules? Here I describe a new approach to this problem, combining molecular dynamics simulations, machine learning, and evolutionary algorithms to design sequence-specific polymers with extremal properties and then reverse-engineer their underlying physics. We apply this strategy to design model sequence-specific copolymer compatibilizers that reduce the energy of a polymer/polymer interface more efficiently than sequence-nonspecific surfactants. Study of these designed sequences ultimately yields new physical insight into the role of chain sequence in polymer surfactants. These results suggest new opportunities for control of surface interactions via designed sequence-specific surfactants and ultimately point to the prospect of a new century of polymer science built around designed sequence specificity. Finally, we demonstrate that this strategy for materials design and understanding is extensible to the design and understanding of other classes of soft materials. We employ a hybrid computational/informatic genetic algorithm to design model molecules with targeted glass formation behavior – a problem of relevance to material challenges ranging from next-generation battery materials to lightweight structural materials. The authors acknowledge the W. M. Keck Foundation for generous financial support of this research.