Alexander Grill Dipl.-Ing. Publications

Compared to other technologies, gallium-nitride (GaN) metal-insulator-semiconductor high-electron mobility transistors (MIS-HEMTs) offer superior electronic properties in terms of breakdown voltage, on-state resistance, and switching behaviour. The large band gap allows the design of power devices with very high breakdown voltages. Large spontaneous and piezoelectric polarization values lead to very high sheet carrier densities at the heterointerfaces with high mobility values. This enables the design of HEMTs with high power, low on-state resistance, and fast switching dynamics. The gate insulating dielectric minimizes gate leakage and enables the design of normally-off HEMTs.
One major reliability issue, in both normally-on and normally-off MIS-HEMTs, is the threshold voltage (Vth) drift at forward gate bias stress. Studies of the degradation behavior have revealed broad distributions of capture and emission times as well as second-order effects attributed to the interaction of trapped charges with the channel through the barrier and a coulomb feedback effect on the surface potential.
Our investigations showed that the experimentally observed Vth drift in GaN MIS-HEMTs can be well described using a non-radiative multi-phonon model for charge capture and emission. The large trap density together with the fact that in GaN HEMTs the electron channel is separated from the interface through a barrier layer results in a strong sensitivity of the surface potential on the trap occupancy. This effect leads to severe challenges in identifying the physical trap properties, because the active energy area and the surface potential are a function of stress and recovery history. Furthermore, the inhomogeneous potential in the oxide leads to changes in the kinetics for each individual trap, which also has to be considered for an accurate model.
The identification and modeling of the physical properties of the involved traps, their dynamic behavior, and the transport of electrons from the channel to the traps are the central tasks of our current research activities.

Fig. 1: The surface potential for increasing stress times during recovery at Vg = 10 V. The equilibrium value is given by the dotted red line. The permanent shift in 's causes a shift of the effective trap energies. This shift depends on the previous stress and recovery cycles.

Fig. 2: Simulated and measured Vth recovery curves for measurement-stress-measurement sequences with stress voltages of 4 V (red), 5 V (green), and 10 V (blue). The measured forward drift behaviour and its stress time and bias dependency is described by two sets of NMP trap parameters.