Analytical modeling of drain-current characteristics of AIGaN/GaN HFETs with incorporation of the impacts of virtual-gate and transferred-electron effect

GaN-based heterostructure field effect transistors (HFETs) have gained considerable attention in high-power microwave applications. So far, unsurpassed current levels and high output power at microwave frequencies have been achieved. However, the dominant factors limiting the reliability of these devices under high-power operation are still unsettled. Drain current collapse is one of the major encumbrances in the development of reliable high-power devices in this technology. In this thesis, an accurate and versatile analytical model based on the concept of virtual gate formation due to the existence of acceptor type surface states is developed to model the current-collapse phenomenon. The implementation of this simple and at the same time precise analytical model demonstrates superb agreement with the experimental observations of permanent/semi-permanent current collapse in AlGaN/GaN HFETs. An analytical model, with incorporation of transferred-electron effect, for drain-current characteristics of AlGaN/GaN HFETs is also presented. Oftentimes, the transferred electron effect is neglected in modeling the drain-current characteristics of III-V HFETs. The broader steady-state electron drift-velocity overshoot of GaN in comparison to other direct semiconductors such as GaAs and InP, in addition to the larger difference between the peak and saturation drift-velocity, and the wider bandgap of this semiconductor predict the importance of the incorporation of transferred-electron effect (Le. steady-state drift-velocity overshoot) in modeling the drain-current of these devices. Simulation results are compared with the results of the adoption of Ridley's mobility model which does not take into account the transferred-electron effect. Solving the Poisson's equation through a simple iterative method and considering the diffusion component of current are at the core of this model. The iterative nature of this approach has considerably relieved the outcome of the implementation from the choice of fitting parameters