Gate Leakage Current Characterization and Remaining Useful Lifetime Prediction in Silicon Carbide MOSFETs

This article presents a comprehensive study on gate leakage current characterization of silicon carbide (SiC) MOSFETs and the prediction of the device’s remaining useful lifetime (RUL) using machine learning (ML) methods. Recent studies in SiC MOSFETs reliability recognize gate leakage currents as an aging indicator. However, most measurements are confined to periodic measurements of one device type. This limitation raises concerns in SiC MOSFETs, particularly for safety-critical applications like aerospace and electric vehicles (EVs). To address this concern, this study proposes a functional circuit capable of measuring gate leakage current at 100 samples per second rate, facilitating the aging and failure characterization accurately. In addition, it showcases the effective application of practical ML methods for predicting device longevity. Experimental measurements across devices from different vendors consistently reveal declining gate leakage current and increasing threshold voltage with aging. Thanks to the high-resolution measurement capability, for the first time it is also shown that the gate leakage current surges from nanoampere levels to milliampere levels within 1–2 s at the end of the lifetime, signifying impending device failure. This time window provides a safe-stop opportunity that enables an additional protective action way before any fault-related current or voltage overshoots appear. Finally, through the application of a combination of gate leakage current measurements and practical ML methods, the accurate prediction of the RUL of SiC MOSFETs is achieved.