Russell Gleason

Leveraging Machine Learning to Predict Microstructural and Macroscopic Properties of Alumina


Russell Gleason has Bachelor’s degrees in Physics and Mathematics from California State University, Fullerton, and a Master’s degree in Physics from California State University, Long Beach.  His Master’s thesis emphasis was in experimental condensed matter physics, specifically using nanosphere lithography to examine the properties of magnetic thin films.  Before pursuing a doctoral degree in Materials Science, Russell was highly regarded as an educator and taught Physics, Astronomy, and Science Education at universities in California and Colorado.


Alumina is one of the most widely used advanced ceramic materials, yet abnormal grain growth during production remains problematic. The difficulty in controlling the microstructural development of alumina leads to the commensurate difficulty in controlling its macroscopic properties. Ever since Coble’s discovery in the 1950’s that doping alumina with small amounts of magnesium could inhibit grain growth, ceramists have sought to understand the mechanisms behind grain growth in alumina and how macroscopic properties alumina depend on its microstructure. Though many different approaches have been used to frame the problem, from doping, to surface energy and bulk defect chemistry, and more recently complexions, but a complete understanding is still elusive. This work aims to take a new approach to the subject, namely Machine Learning. This approach leverages the power of data informatics to search for previously unrecognized phenomenological relationships among powder properties, chemical composition, manufacturing processes, and the resulting microstructural and macroscopic properties of alumina. Through the use of neural networks and regression analysis, the properties of alumina samples produced by widely varying methods can be compared to each other in order to extract relationships and elucidate methods of better controlling microstructural development and macroscopic properties.