Gallium nitride high-electron-mobility transistors (GaN HEMTs) are increasingly deployed in power conversion applications due to their superior efficiency and switching performance compared to silicon devices. However, long-term reliability under electrical stress remains a critical concern for grid integration and power electronics manufacturers. A detailed study of depletion-mode GaN-on-SiC HEMTs reveals that charge trapping mechanisms create history-dependent degradation patterns not captured by conventional accelerated-stress models.
Researchers subjected three nominally identical devices to sequential electrical stress sequences, monitoring gate leakage current continuously while periodically measuring threshold voltage and on-resistance. The experimental protocol included negative gate biases ranging from −5 to −9 V, positive bias recovery phases, repeated negative stress cycles, and drain-current stress up to 0.5 A under controlled temperature.
Key findings indicate that lower negative bias levels (−5 to −7 V) produced stronger time-dependent parameter drift, whereas higher stress voltages (−8 to −9 V) showed reduced incremental degradation. This counterintuitive behavior reflects cumulative stress effects rather than simple voltage acceleration. Positive gate bias induced partial recovery of both threshold voltage and on-resistance, demonstrating reversibility in some trapping mechanisms. However, reapplication of stress after recovery exhibited altered degradation kinetics, confirming strong dependence on electrical history.
Drain-current stress alone produced limited long-term drift, but combined positive gate bias with drain current increased instability significantly. These results challenge conventional reliability testing approaches that typically apply single, isolated stress conditions. For power system applications—including renewable energy inverters, HVDC converters, and EV charging infrastructure—the findings suggest that realistic duty cycles exposing devices to varying stress states may produce different degradation trajectories than standard qualification tests.
While the study does not uniquely identify specific trap energies or spatial locations, it establishes a reproducible experimental framework for evaluating GaN device stability under realistic operating conditions, enabling more accurate lifetime predictions for critical power electronics applications.
