2019 Van Horn Distinguished Lectures: 3: machine learning of materials structure & synthesis

2019 Van Horn Distinguished Lectures: Part 3 - machine learning of materials structure and synthesis.

The Kent R. van Horn Lectureship is an endowed Lectureship at the Case Western Reserve University and dates from 1974. It honours Kent R. van Horn, an alum, who had a distinguished career as a metallurgist, director of research, and ultimately corporate vice-president of Alcoa. Three lectures on varied topics are to be delivered over three successive days. The 2019 lectures were delivered by Gerbrand Ceder, who is the Chancellor's Professor of Materials Science and Engineering at University of California, Berkeley. He is notable for his pioneering research in high-throughput computational materials design, and in the development of novel lithium-ion battery technologies.

Abstract: Machine learning is particularly well suited for problems for which no deductive theory exists. I will show two examples of such efforts: While structure prediction is usually treated as a mathematical minimization problem of the (free) energy, I will show that it can be learned much more effectively by training an algorithm on the tens of thousands of known crystal structure assignments known in structure databases.

A more difficult problem is that of materials synthesis. With no deductive theory available to predict how novel compounds can be synthesized, it may be possible to learn from the hundreds of thousands published recipes with which inorganic compounds have been synthesized. I will show how we use advanced natural language processing techniques to extract codified synthesis recipes from over 3 Million journal papers. The large collection of synthesis recipes can be mined and used to predict synthesis conditions for new materials.

These videos are reproduced courtesy of the Materials Science and Engineering Department of Case Western Reserve University.

http://engineering.case.edu/materials...
http://www.phase-trans.msm.cam.ac.uk