About this Abstract |
Meeting |
MS&T24: Materials Science & Technology
|
Symposium
|
Materials Processing and Fundamental Understanding Based on Machine Learning and Data Informatics
|
Presentation Title |
Image Processing of Charge Density from DFT to Predict Properties in Complex Materials |
Author(s) |
Hossein Mirzaee, Ramin Soltanmohammadi, Nathan Linton, Jacob Fischer, Serveh Kamrava, Pejman Tahmasebi, Dilpuneet S. Aidhy |
On-Site Speaker (Planned) |
Dilpuneet S. Aidhy |
Abstract Scope |
With its origins in digital media, Image Processing is rapidly becoming an important tool in microstructure analysis. We demonstrate the integration of convolutional neural networks (CNNs) into density functional theory (DFT) calculations to predict materials properties. Specifically, we use DFT-derived charge density distribution to build a CNN machine learning model to predict properties in complex alloys. We show that charge density is the only descriptor that is needed to develop models for accurate predictions of elastic constants and stacking fault energies in high entropy alloys. We further show that the model only needs to be trained on simpler/binary alloys to predict properties in complex alloys, thereby opening a pathway to explore large compositional space in the field of high entropy materials. We propose that our framework can potentially be agnostic, and can be applied to experimental studies of materials processing. |