Machine Learning Analysis of Reflected High Energy Electron Diffraction Images of Epitaxial Oxide Thin Films

Reflected high energy electron diffraction (RHEED) is a highly common form of real time analysis used in growth systems such as molecular beam epitaxy (MBE) and pulsed laser deposition (PLD). Traditional RHEED analysis focuses only on the intensity and shape of the diffraction pattern for a few still images taken during growth and is mostly qualitative. While this information has proven insightful, there is far more information that can be gleaned from RHEED. In order to obtain greater insight from RHEED recordings, the novel machine learning techniques principle component analysis (PCA) k-means clustering, and convolutional neural networks were applied to the recordings of the RHEED taken during the molecular beam epitaxy growth of epitaxial thin film perovskite oxides. Neural network analysis is extended to predict X-ray photoelectron spectroscopy data from RHEED images as a proxy for film composition and material classification. These methods yield more quantitative results from the RHEED with minimal time requirements and open the door for future development of real-time computer control of film growth for optimal growth conditions.