Developing Workflows and ML Structure/Property Relationships for Catalysis on Metal Oxide Systems

Abstract

This project will investigate and develop machine learning models to accelerate the study of surface chemistry on metal oxides. High-throughput computational chemistry workflows will be developed to build training datasets for how reaction species such as methanol adsorb to oxide surfaces. Graph convolution models will be developed and tested to accelerate fundamental challenges in oxide materials: predicting local magnetic moments, predicting sensitivity of calculations to correction schemes such as DFT+U, and accounting for surface termination changes under reaction conditions. The dataset and machine learning models will be used to identify trends across composition and structure for intermediate adsorption energies and surface stability. This project has both basic science and practical implications for contaminant reduction and other surface chemistries.

Document Details

Document Type
DoD Grant Award
Publication Date
Jul 09, 2020
Source ID
W911NF2010188

Entities

People

  • Zachary Ulissi

Organizations

  • Army Contracting Command
  • Massachusetts Institute of Technology
  • United States Army

Tags

Readers

  • Computational Modeling and Simulation
  • Neural Network Machine Learning.
  • Quantum Chemistry

Technology Areas

  • AI & ML
  • AI & ML - Neural Networks