Physics of electrical degradation in GaN high electron mobility transistors

Abstract

The deployment of GaN high electron mobility transistors (HEMT) in RF power applications is currently bottlenecked by their limited reliability. Obtaining the required reliability is a difficult issue due to the high voltage of operation. In order to improve reliability, it is essential to develop detailed physical understanding of the fundamental degradation mechanisms. In this thesis, we investigate the physical mechanisms behind the electrical degradation of GaN HEMTs by performing systematic stress experiments on devices provided by our industrial collaborators. These devices are electrically stressed under various bias conditions while regularly characterized by a benign characterization suite. We observe that electrical stress beyond a critical voltage results in an increase in drain resistance, a decrease in maximum drain current, and a sharp increase in reverse gate current. We show that this mode of degradation is driven by electric field and that current is less relevant. Behind this degradation is trap formation that occurs at the critical voltage. To understand this, we have developed a new trap-analysis methodology. It is found that under stress, the density of traps increases in the AlGaN barrier layer in the proximity to the gate edge on the drain side of the device. We show that this degradation is enhanced under mechanical uniaxial tensile strain that is externally applied to the device. From our experiments, we propose a degradation mechanism of defect formation through the inverse piezoelectric. In this mechanism, high vertical electric field at the gate edge under high voltage increases tensile stress in the AlGaN layer due to piezoelectricity of the material.

(cont.) When the elastic energy in the crystal exceeds a critical value, crystallographic defects are formed. These defects trap electrons and reduce drain current as well as provide leakage paths and increase gate current. We theoretically validate the plausibility of this hypothesis and provide a model for the critical voltage that agrees with experimental observations. Unlike conventional wisdom, hot electrons do not appear to be the direct cause of electrical degradation in the devices that we study. Our studies suggest several possibilities to improving the electrical reliability of GaN HEMTs.