Achievement of High‐Quality Gallium Oxide Epitaxial Growth via Machine Learning

Abstract The development of gallium oxide (Ga 2 O 3 )‐based electronics has been hampered by the persistent challenge of obtaining high‐quality epilayers with device‐grade performance metrics. While metal‐organic chemical vapor deposition has proven effective in producing device‐grade films, the realization of ideal step‐flow epitaxial growth with high deposition rates in Ga 2 O 3 remains a significant scientific challenge. Here, a machine learning (ML)‐guided approach is presented to overcome conventional epitaxial limitations, which are historically constrained by narrow process windows and empirical growth paradigms. Specifically, our ML‐guided approach successfully achieves perfect step‐flow epitaxy at 1.2 µm h −1 on universal β‐Ga 2 O 3 substrates, eliminating traditional requirements for large mis‐cut angle substrates and low growth rates. This breakthrough resolves the long‐standing trade‐off between crystal quality and growth efficiency. The resultant epilayers exhibit atomically smooth surfaces with record‐low roughness (0.121 nm) featuring 6 Å‐high single‐atomic‐layer steps, coupled with exceptional electrical properties. The technological viability is further validated through Schottky barrier photodiodes with excellent solar‐blind detection performance: ultra‐fast decay time (3.28 µs), ultra‐high photo‐to‐dark current ratio (PDCR > 10 5 ), and ultralow dark current density (6.2 × 10 −9 A cm −2 ). This work not only establishes ML as a revolutionary accelerator for β‐Ga 2 O 3 development but also provides a transformative methodology for next‐generation semiconductor manufacturing.