Accelerating Materials Discovery with AI

Goals

Finding new materials to serve as the next generation catalysts, batteries, solar cells, superconductors or electronic devices can have a potentially transformative impact on our lives and society. Here, we seek to leverage state-of-the-art machine learning methods to accelerate the process of materials discovery and design far beyond what is possible using conventional simulation and screening algorithms.  

Issues Involved or Addressed

Designing novel materials with targeted functionalities, or inverse design, remains a grand challenge in the materials sciences. To address this, we will develop new data-driven algorithms for inverse design at the atomic level, with a goal towards achieving chemical accuracy. Here we will investigate novel machine learning architectures focused around graph neural networks for efficiently capturing spatial and chemical information, and incorporating physical constraints and domain knowledge. Optimization in chemical latent spaces and generative modelling will also be studied. Developing effective testing platforms and benchmarks will be crucial for achieving this goal in a timely manner. Ultimately the team will deploy these models for real-world problems and datasets for materials discovery.

Partners/Sponsors

N/A