About this Abstract |
Meeting |
2020 TMS Annual Meeting & Exhibition
|
Symposium
|
Computational Discovery and Design of Emerging Materials
|
Presentation Title |
Frequency-dependent Dielectric Constant Prediction of Polymeric Dielectrics with Machine Learning |
Author(s) |
Lihua Chen, Rohit Batra, Chiho Kim, Tran Huan, Rampi Ramprasad |
On-Site Speaker (Planned) |
Lihua Chen |
Abstract Scope |
The dielectric constant is a vital property for the design of novel polymeric dielectrics for a wide range of applications, e.g., energy storage, photovoltaic devices and high-voltage insulations. It is usually computed using the density functional perturbation theory and classical molecular dynamics simulations. However, these theoretical methods are limited in terms of their accuracy, suffer from high computational cost and can only provide such information in the high-frequency regime. To fill this gap, we develop a machine-learning (ML) based model to predict the frequency-dependent dielectric constant of polymers, using a dataset including several-hundred experimentally measured dielectric constant values at different frequencies. This is achieved using the advanced fingerprinting scheme to numerically represent the polymers and the Gaussian process regression algorithm to train the model. The resulting ML model can accurately and rapidly predict the frequency-dependent dielectric constant of new candidates, allowing the design of new polymers for specific applications. |
Proceedings Inclusion? |
Planned: Supplemental Proceedings volume |