Machine-learning-decoupled absorption spectroscopy reveals bandgap shrinkage competition in Si-doped β-Ga 2 O 3

β-Ga 2 O 3 is promising for power and solar-blind UV photodetection due to its wide bandgap, high critical field, and ease of n-type doping with Si. However, a doping limit arises at high Si concentrations, attributed to band-gap narrowing and renormalization caused by dopants and defects. Dopant ionization (via the Franz–Keldysh effect) and defects (via the Urbach effect) both induce band-gap narrowing and alter the absorption coefficient, yet their individual contributions are difficult to decouple from absorption spectra near the bandgap (4.5–5 eV). This study extends analysis to 2–5 eV using machine learning, enabling the separation of each effect’s contribution. The resulting metric directly reflects changes induced by doping. This unreported finding clarifies the intrinsic band-gap narrowing mechanism in Si-doped β-Ga 2 O 3 and offers valuable guidance for other n-type doping studies.