Simulation results have been obtained for AlGaN/GaN HEMTs. Fig. 7.50 shows the comparison of simulation and measurements of a = 0.2 m AlGaN/GaN HEMT for = 6 V. The gate-width of the measured device was 260 m. Three factors are most important for the simulation: First, a realistic description of the ohmic contact situation needs to be applied.
For the AlGaN/GaN HEMT simulated, non-alloyed ohmic contacts are introduced, so the contact situation II from Fig. 3.25 is assumed. Electrons have to pass the AlGaN/GaN band gap discontinuity by RST at the drain side. It was found, that performing an analysis, such as in , does not match the transport physics of this HEMT correctly, since the transfer characteristics would not saturate, as indicated with the simulation of contact Case I. We see in Fig. 7.50, that assuming contact situation Case II, does deliver agreement for lower , however for higher fields the simulation shows a rapid change which has a different position in the measurement. The transconductance is strongly affected by the barrier conduction in the AlGaN, and Fig. 7.50 underestimates the characteristics of the measured device for higher . Further understanding of the high field transport in AlGaN and GaN is required to perform the investigations for these HEMTs for higher currents and temperatures. The understanding of the resistances and is especially desirable to use these device for frequencies of operation above 10 GHz.
Second, the available carrier concentration requires careful evaluation due to the parasitic effects. A good agreement can be found for a background concentration of 110 cm in all materials. The device was not passivated by SiN for the measurements performed. A negative charge concentration of -0.310 cm is introduced at the air/AlGaN interface for the simulation. At the channel/barrier interface, a positive surface charge of 1.510 cm is used in the simulation. At the substrate a negative charge of -110 cm is applied. These surface charges introduce the effects of the parasitic properties of the semiconductor materials resulting in the high current densities at the AlGaN/GaN interfaces. The inclusion is necessary, since otherwise, as was shown by Sacconi et al. , the effective band energy edges are not modeled correctly. Third, the maximum DC-power in this simulation is about 4.2 W/mm, which is about twice a typical GaAs HEMT power compliance. This requires to account for self-heating in the simulation. As the band structure at higher temperatures is still under discussion, this topic remains for further analysis.