Description
GaN-based high electron mobility transistors (HEMTs) offer excellent high-power and high-frequency performance, allowing them to amplify high power signals at microwave frequencies very efficiently. The current in GaN HEMTs flows in a two-dimensional electron gas at the interface between GaN and AlGaN layers, which offers excellent electrical properties (high carrier density and mobility). Understanding the physics of failure in these devices remains an important issue, with both abrupt failures and gradual parametric degradation having been observed. These devices may show sudden and permanent damage when subjected to very high reverse bias-stress. This failure mechanism has been addressed by improvements in processing technology. However, the parametric degradation that occurs in the �semi-ON� state at moderate drain biases remains an issue. In this condition, the device is biased close to pinch-off, but the relatively small numbers of electrons that are flowing are accelerated by a high electric field. The resulting energetic carriers can activate, by dehydrogenation, or reconfigure defects near the interface. The defect generation is greatest at the end of the gate on the gate-drain side, where the lateral electric field is at its maximum. This may lead to significant reductions in drain current and transconductance, as well as shifts in threshold voltage, resulting in poor DC, RF and large-signal performance. GaN/AlGaN HEMTs grown under various conditions (i.e., gallium-rich, nitrogen-rich, and ammonia-rich) have been analyzed and the defects responsible for degradation in each device type have been identified. The atomic-scale nature of the traps that produce changes in threshold voltage, leakage current, and drain current have been related to changes in transconductance and turn-off voltage using a combination of electrical measurements, quantum mechanical calculations, Monte-Carlo device simulations, and accelerated degradation tests. A relatively simple formulation has been developed under the assumption that the hot-electron scattering cross-section is independent of the electron energy. In this case one can relate the change in defect concentration to the operational characteristics of a device, such as the spatial and energy distribution of electrons (electron temperature), electric field distribution and electron energy loss to the lattice. The number of electrons with energy (obtained from device-level Monte Carlo simulations) in excess of that required to activate a defect (obtained from density functional theory) is used to predict the degradation rate. The results of quantum mechanical calculations of candidate defect formation energies as functions of growth conditions and Fermi level position were used to identify the primary defects responsible for the degradation as hydrogenated Ga vacancies or hydrogenated N antisite defects. In each case, degradation occurs by the hot electrons providing the energy needed to release a hydrogen atom. The calculations also yield the activation energy for hydrogen release.
Speaker(s):