Third quadrant overvoltage ruggedness of p-gate GaN HEMTs

Abstract

Overvoltage ruggedness is an essential reliability requirement for non-avalanche AlGaN/GaN power devices against inductive transient spikes in the OFF-state breakdown events. In this work, we first report the 3rd-quadrant (i.e. source-to-drain reverse operation) overvoltage ruggedness of the p-gate GaN high-electron-mobility transistors (HEMTs) by performing static current-voltage (I-V) sweeps and dynamic unclamped-inductive-switching experiments. Both experimental results reveal an inferior breakdown performance of the device under 3rd-quadrant blocking conditions, including a low static voltage of 98 V and a dynamic breakdown voltage (BV) of 155 V. In particular, this transient voltage is evidently degraded from 155 V to 114 V (i.e. a 26% reduction) by increasing the load current, which is significantly different from the stable BV (1.4 kV) achieved in the 1st-quadrant dynamic breakdown. Such unusual degradation behavior can be explained by the trapping-related carrier transport mechanism, which is supported by temperature-dependent breakdown characteristics. Further numerical simulations, de-capsulation failure analyses and repeated UIS experiments are carried out to study the failure mechanism and gain insight into the overvoltage capability of the device. These results suggest that the rational design of the p-gate GaN HEMTs with the desired 3rd-quadrant overvoltage ruggedness requires careful consideration, especially in reverse conduction events

Insights into dynamic switching behavior and electric field distribution of depletion-mode GaN HEMT with bonding pad over active layout

“…To reveal the effect of the different BPOA structures (D-, Sand G-BPOAs) on the dynamic switching and off-state blocking characteristics of the device, a physics-based technology computer aided design (TCAD) simulation in Silvaco Atlas is performed [13][14][15], where the key parameters are depicted in table 1 for simulation calibration. The ionization model with a shallow donor (∼1 × 10 15 cm −3 ) and a deep donor (∼2 × 10 17 cm −3 ) is used to calibrate the ionized doping concentration in the buffer layer, while taking into account other numerical models such as Shockley-Read-Hall, Auger model, polarization model, Fermi-Dirac statistics and impact ionization [16,17]. Meanwhile, Fe or C doped acceptor-like traps with E V + 0.5-1 eV are implemented for donor compensation [14,18].…”