Date of Completion
polymer dielectrics, machine learning, first principles
Ramamurthy (Rampi) Ramprasad
George Rossetti, Jr.
Avinash M. Dongare
Field of Study
Materials Science and Engineering
Doctor of Philosophy
Polymers offer a nearly infinite variety of material systems with diverse properties.
Until recently, the formulation of polymers for specific applications was based
on trial and error, guided by intuition. In this work, first principles computations
and machine learning approach are employed to guide the design of polymers, in
the present case for dielectric applications. Specically, we adopt two strategies, (1)
functionalization of a well understood polymer dielectrics, such as PE and PP, to
enhance its dielectric response, and (2) discovery of entirely new classes of polymer
dielectrics, both organic and organometallic. Different polymer classes are explored,
from C-based organic polymers to novel Si-, Ge-, and Sn-based polymers, and the
search is based on two properties, band gap and dielectric constant. Newly developed
high throughput DFT methods were used first to accurately determine the
dielectric constant and band gap of dierent polymer systems for a set of limited
compositions and congurations. Machine learning methods were then used to predict
the properties of systems spanning a much larger part of the congurational
and compositional space. Based on this strategy, we are able to provide a "map" of
the achievable combination of properties within the chemical space explored.
Wang, Chenchen, "Polymer Dielectrics Design Using First Principles Computations and Machine Learning" (2015). Doctoral Dissertations. 654.