Published online by Cambridge University Press: 13 June 2012
In this paper, we compare degradation modes and failure mechanisms of different AlGaN/GaN HEMT technologies. We present data concerning reverse-bias degradation of GaN-based HEMTs, which results in a dramatic increase of gate leakage current, and present a timedependent model for gate degradation. Some of the tested technologies demonstrated to be immune from this failure mechanism up to drain-gate voltages in excess of 100 V. When this was the case, the main failure mode consisted of drain current degradation during on-state tests, resulting from charge trapping in the gate-drain access region attributed to hot-electron effects. Finally, the use of diagnostic techniques such as electroluminescence microscopy and Deep Level Transient Spectroscopy for the identification of failure modes and mechanisms of GaNbased HEMTs is reviewed. Concerning reverse-bias degradation of GaN-based HEMTs, we demonstrate that, (i) when submitted to reverse-gate stress, HEMTs can show both recoverable and permanent degradation. (ii) recoverable degradation consists of a decrease in gate current and threshold voltage, which are ascribed to the simultaneous trapping of negative charge in the AlGaN layer, and of positive charge close to the AlGaN/GaN interface. (iii) permanent degradation is manifested by the generation of parasitic leakage paths. Time-dependent analysis suggests that permanent degradation can be ascribed to a defect generation and percolation process. Results supports the existence of a time to breakdown for HEMT degradation, which significantly depends on the stress voltage level. On the contrary, AlGaN/GaN technologies which were found to be resistant to gate degradation (off-state critical voltage larger than 100 V for a 0.25 um gate device) were subjected to on-state tests at different gate and drain voltage levels. All tests showed a non-recoverable degradation of electrical parameters (drain saturation current, threshold voltage and on-state resistance) and electroluminescence signal EL, with a strong dependence on the EL value of the bias point, and a negligible dependence of temperature. Once verified that EL intensity represents a reliable estimate of channel hot electron effects, we attributed the degradation to hot electron trapping in the gate-drain access region. Using EL intensity as a measure of the stress acceleration factor, we derived an acceleration law for GaN HEMT hot electron degradation similar to the one already demonstrated for GaAs devices.