ABSTRACT:

The acquisition of materials properties is the key bottleneck in the design and discovery of new materials. While the combination of ab initio methods and Moore’s law in computing power has substantially decreased the cost of materials property acquisition in the past few decades, property prediction remains prohibitively expensive, especially for large, complex systems. In this talk, I will demonstrate how machine learning (ML) is powering the next revolutionary leap in our ability to predict materials properties at low cost while maintaining high transferability accuracy. We now possess the ability to develop accurate interatomic potentials for complex systems, which can be used to probe complex “high-entropy” alloys as well as lithium superionic conductors. Graph deep learning models have also matured to the extent that they can now be used to reliably access vast compositional spaces for materials design. Finally, I will also discuss the consequences of the exponentially declining cost of property prediction on the field of materials science.  

Join Here: 
https://umbc.webex.com/umbc/j.php?MTID=m3000a2cfcc900d12cec0d02b930edf96


Meeting number: 2621 955 5740
Password: uPJX6JMQg69

Join by video system
Dial 26219555740@umbc.webex.com
You can also dial 173.243.2.68 and enter your meeting number.

Join by phone
+1-202-860-2110 United States Toll (Washington D.C.)

Access code: 262 195 55740