Rational Design of Polymer Dielectrics Using First Principles Computations and Machine Learning

While intuition-driven experiments and serendipity have guided traditional materials discovery, computational strategies have become increasingly prevalent and a powerful complement to experiments in modern day materials research. A novel approach for efficient materials design is “rational co-design”, where high-throughput computational screening is used synergistically with experimental synthesis and testing. In this Thesis, the utility and promise of such an approach was demonstrated for the design of advanced polymer dielectrics for electrostatic energy storage applications. Density functional theory computations were applied to study the structural, electronic and dielectric properties of polymers, based on which targeted synthesis and property measurements were carried out for promising candidates. These co-design efforts led to the identification of potential replacements for present day “standard” dielectrics (such as biaxially oriented polypropylene) not only by new organic polymer candidates within known generic polymer subclasses (e.g., polyurea, polythiourea, polyimide), but also by organometallic polymers, a hitherto untapped but promising chemical subspace. Further, the prospects of significantly accelerating the materials design process using state-of-the-art machine learning techniques were explored. Vast computational data generated as part of this work was mined for the development of accurate ‘instant prediction’ and ‘design’ models for the relevant properties of polymers. These models were converted into user-friendly polymer design tools, and along with the computational and experimental data, compiled in the form of a web-based application (http://khazana.uconn.edu/polymer_genome/) to facilitate the rapid design and discovery of polymer dielectrics