The Alliance for AI-Accelerated Materials Discovery (A3MD) is a new initiative that seeks to leverage the power of artificial intelligence to design the next generation of high-performance materials. Uniting recent advances in machine learning and high-throughput experimentation, the team seeks to accelerate the discovery and commercialization of new, efficient catalysts and consumer electronic materials. A3MD brings together world-leading researchers from the University of Toronto, McMaster University and the National Research Council of Canada, as well as industrial partners LG and Total.
Material discovery can be concisely described as a three-step process:
- Computational screening of the available chemical space using first-principle calculations
- Down-selection of the most promising candidates and synthesis of these materials
- Characterization and analysis of the successfully synthesized candidates
A3MD accelerates each stage of the materials discovery process.
Opportunities with A3MD
For all A3MD inquiries please contact:
Ted Sargent, Principal Investigator: firstname.lastname@example.org
Brandon R. Sutherland, Executive Director: email@example.com
First-principle calculations are a necessary first step to explore the available chemical space of a new class of materials. These calculations – particularly density functional theory (DFT) – are computationally expensive, limiting the number of different material combinations that can practically be evaluated. However, if a machine learning model is trained on just a small number of DFT-calculated materials, it can then be used to rapidly explore a much larger set of materials. This acceleration enables the screening of several orders-of-magnitude more compounds than if DFT is used alone, and at a similar accuracy.
As laboratory robotics improve and become more accessible, researchers can perform experiments at an increased rate. Traditional research methods – in which a researcher sequentially optimizes a synthesis procedure to make a new material, one at a time – limit the number of experiments to <50 /month. New advances in parallel experimentation have increased this number to 10s-100s of experiments a day. A3MD is partnered with the National Research Council of Canada (NRC) to acquire and host the infrastructure required to perform autonomous, parallelized experimentation to accelerate materials discovery.
Parallel Analysis and Characterization of New Materials
The final – and perhaps most crucial – step in the materials discovery process is to evaluate the performance of the newly synthesized material. This analysis is often the bottleneck in the discovery process, and can eliminate much of the benefit gained from accelerating the first two steps. A3MD will place an equal emphasis on acquiring infrastructure to rapidly test the performance of synthesized materials. Where accelerated evaluation of a certain metric is impossible, A3MD will again leverage machine learning to learn proxies for the desired metric.
Tying it all Together
In addition to accelerating the materials discovery workflow, A3MD will maximize the learning that happens from each experimental cycle. With techniques such as Bayesian Inference, machine learning can guide researchers to help decide which set of experiments should be performed next. In a similar fashion to how research is currently conducted, each previous iteration will inform the next set of experiments; Bayesian optimization has the added benefit of learning trends in a highly multi-dimensional space that are not obvious to human researchers.