Luminance histogram analysis for RHEED images using peak fitting approach with machine learning technique

Reflection high-energy electron diffraction (RHEED) is widely used for in situ observation during physical vapor deposition (PVD) process such as molecular beam epitaxy (MBE) to precisely control of the two-dimensional materials fabrication, which is promising for next generation electronic devices. Although automatic analysis using machine learning technique has been attracting for taking in the abundant information contained in RHEED, there are still problems such as the robustness of image misalignment or possibility of in situ analysis. In this study, we propose an automatic analysis method based on a luminance-histogram representation of RHEED images, in which the histogram is treated as a spectrum and analyzed using a high-throughput peak fitting technique. The method is robust to image misalignment and has in situ analysis possibility. Applied to RHEED images of the indium deposition process on to Si(111) surface, we achieved to estimate the optimal deposition time for fabricating each surface superstructure and the external parameter changing time.