Device-to-device variability in positive bias temperature instability (PBTI) remains a critical reliability concern for 4H-SiC MOSFETs because conventional electrical characterization provides limited capability for predicting long-term degradation behavior before stress. This study proposes a physics-informed statistical screening framework that predicts post-turnaround PBTI variability using pre-stress electrical measurements together with an early-stage transient descriptor. The proposed framework combines Triple-sense electrical features, principal component analysis (PCA), a transient descriptor, and statistical regression. Six bias-history-dependent threshold-voltage features extracted from Triple-sense measurements were reduced to two PCA-derived electrical descriptors, preserving 94.9% of the total variance. The turnaround time, tbottom , was introduced as a transient descriptor characterizing the temporal transition from the initial negative threshold-voltage shift to the subsequent post-turnaround degradation. Regression analysis demonstrated that the combination of the two PCA-derived electrical descriptors and the transient descriptor improved prediction of post-turnaround threshold-voltage variability compared with stress-time-based models. Device-wise leave-one-device-out cross-validation yielded an R2 of 0.930, indicating reproducible predictive performance for previously unseen devices evaluated under the same experimental protocol. The proposed framework is not intended as a universal physical degradation model or a microscopic trap-characterization technique. Instead, it demonstrates that pre-stress bias-history-dependent electrical responses contain statistically useful predictive information regarding subsequent post-turnaround PBTI variability. The proposed framework provides a practical approach for predicting device-to-device reliability variability by combining pre-stress electrical measurements with the experimentally observed turnaround time under a defined stress condition.
This paper focuses on studying p-GaN gate high-electron-mobility transistors (HEMTs) characteristics under drain voltage stress (VD,stress). A novel double-pulse circuit was developed to evaluate turn-on transient characteristics. It significantly reduces the measurement delay from several microseconds in traditional circuits to less than 20 ns. Experimental results show that increasing the stress duration induces a positive shift in threshold voltage (VTH), which prolongs the turn-on time and reduces the peak drain current. At a VD,stress of 300 V, VTH increases from 1.89 to 2.09 V as the stress duration increases, accompanied by an increase in the drain-voltage switching time from 7.84 to 12.67 ns and a decrease in the peak drain current from 12.5 to 10.5 A. The competition between the prolonged turn-on time and reduced peak current results in a non-monotonic variation in turn-on energy loss (EON).
Power conversion system plays a critical role in modern electric vehicles and renewable energy grids, etc. As one of core components, the health monitoring of Insulated Gate Bipolar Transistors (IGBTs) is essential for system reliability. This study explores a method for extracting operational parameters from the spatially radiated magnetic field of IGBT modules, and we propose a physics-informed Transformer framework that decodes the IGBT operational status through non-contact measurements of spatial magnetic field spectrum data. The spectral data are converted into grayscale images, and a multi-head attention mechanism integrated with gated recurrent units (GRUs) is employed to learn the mapping between spectral features and electrical parameters. Trained on experimental data, the framework successfully extracts the operational parameters from the spatially radiated magnetic field of IGBT modules, while the attention mechanism enhances the model interpretability by revealing the frequency-dependent parameter correlations. It is believed that this research will be useful for supporting condition awareness of power electronic components and providing a new perspective on the future degradation-aware monitoring the IGBT modules.
In the long-term application of silicon carbide MOSFETs in space, not only will they be affected by radiation and experience changes in electrical performance, but also they will age due to the inherent positive bias temperature instability (PBTI) effect of the gate dielectric over time. Currently, there are relatively few studies on the influence of irradiation on the BTI aging of devices. In this work, Si-ion irradiation experiments were performed on a commercial 1200 V, 68 A n-channel 4H-SiC power MOSFET at two energies and five fluences. The irradiated devices were then subjected to a 48 h PBTI stress at 450 K with VGS = 20 V. The experimental results showed that the PBTI effect of the devices became more pronounced after irradiation, and both the heavy ion energy and fluence had different effects on the variation pattern of the PBTI effect. Low-frequency-noise measurements, an ionization-dominated Co-60 γ-ray control experiment and ERETCAD simulations point to the same picture. Under the sequential irradiation and PBTI conditions investigated in this work, displacement-damage-related interface and near-interface defects appear to be the dominant additional contributors to the aggravated PBTI response. Ionization-induced oxide-trapped charge mainly modifies the initial threshold-voltage state, and it may also contribute to the subsequent threshold-voltage evolution through charge recovery and compensation. A device that shows no functional failure immediately after irradiation can therefore still drift during subsequent long-term operation. Therefore, before SiC MOSFETs are widely used in space, it is very important to pay attention to the synergistic relationship between radiation damage and the aging of the devices themselves.
This work investigates the temperature-dependent degradation and failure behavior of a 160 V commercial N-channel split-gate trench MOSFET (SGTMOSFET) under high gate overstress at different environmental temperatures. The results reveal a two-stage evolution: pre-breakdown gate-oxide degradation characterized by a non-monotonic threshold-voltage shift in the early stage, followed by time-dependent hard breakdown at longer stress times. At high gate overstress voltage, the recovery effect appears earlier and becomes more pronounced as temperature increases. The subthreshold midgap technique (SMGT) separates the measured threshold-voltage shift and shows that the oxide charges reverse during prolonged stress. Optical beam induced resistance change (OBIRCH) and emission microscopy (EMMI) localize the failure position, and the focused ion beam (FIB) slicing analysis at this position further confirms aluminum melting. Technology computer-aided design (TCAD) simulation associates the threshold-voltage recovery with the temperature-dependent evolution of oxide charge and trap-related processes. The simulated gate current agrees with the measured one, and the aluminum melting point is reached during the current runaway, which indicates that the melting is a consequence of the gate-current runaway rather than its trigger. Results show that the temperature-dependent current distribution around the gate oxide is associated with the electrothermal response under gate overstress. This temperature-dependent reliability mechanism further reveals different behaviors of SGTMOSFETs in extreme environments.
