Physics-informed machine learning to analyze oxide defect-induced RTN in gate leakage current

In this work, we introduce a physics-informed machine learning framework that models stochastically-behaving defects in the gate oxide of an individual device. We demonstrate the extraction of the total number of active defects and their corresponding behavior from oxide defect-induced random telegraph noise signals existent in time-resolved constant stress-induced leakage current. Physics-informed machine learning framework conducts an automated data-driven analysis of this stress-induced leakage current data with decision-making guided by the incorporated defect physics. Finally, this work includes the generated statistics from a successful implementation of the physics-informed machine learning framework on stress-induced leakage current data of multiple deeply-scaled high-κ metal gate devices.