Paper on Machine Learning of In Situ RHEED Published in Physical Review Materials

RHEED K-means clustering
K-means clustering of SrTiO3 film grown on TbScO3

Our first lead author paper has been published in Physical Review Materials! Dr. Sydney Provence led this work on k-means clustering and principal component analysis (PCA) of films grown in our lab and in Prof. Bharat Jalan’s lab at the University of Minnesota. Using these big-data analysis approaches, we examined how the film surface evolves during the early stages of growth. We showed that k-means clustering can accurately determine when strain relaxation occurs in BaSnO3 and shuttering-steps in MBE growth of LaNiO3. PCA and k-means also can be used to infer order/disorder transitions in the growth process of SrTiO3 and determine whether a homoepitaxial film is stoichiometric. With the clustering we are able to determine what features in the RHEED pattern are changing the most during layer-by-layer growth of SrTiO3 on TbScO3. The key takeaway is that there is a great deal more information that can be gleaned from RHEED than simply monitoring oscillations of the specular peak.

Sydney has shared the code through her GitHub account. We encourage other groups to record videos of all their growths and use the code for further analysis of their data. There is a rich playground to extract information from RHEED during growth and we are only beginning to scratch the surface. If you’re interested in learning more, please contact Sydney or Ryan for more information. This work was supported in part by our NSF project, DMR-1809847.