A combined approach of atom probe tomography and unsupervised machine learning to understand phase transformation in (AlxGa1−x)2O3
Jith Sarker, Scott Broderick, A F M Anhar Uddin Bhuiyan, Zixuan Feng, Hongping Zhao, Baishakhi Mazumder
Abstract
In this paper, we investigated the evolution of microstructural chemistry of metal organic chemical vapor deposition grown (010) (AlxGa1−x)2O3 films with varying Al contents, x = 0.10–1.0, using atom probe tomography (APT). At a low Al content (x ≤ 0.25), the films are homogeneous, where layer inhomogeneity appears at a high Al content (x > 0.25). Further increasing the Al content up to x ≥ 0.60 results in a homogeneous (AlxGa1−x)2O3 layer. This change in microstructural features was linked to the phase transformation of (AlxGa1−x)2O3 using a manifold learning approach to capture the governing features hidden in the data dimensionality. Combining APT to unsupervised machine learning enables APT to be an independent material characterization tool to investigate the microstructure, chemical composition, and phase related information.