Conclusions

In this article, we have presented calculations of the DC and RF characteristics in comparable AlGaN/GaN and AlGaAs/GaAs HEMTs. The sheet charge density calculations for different gate biases are made using a self-consistent solution of the Schroedinger and Poisson equations. The carrier mobility in the GaN HEMT is extracted from previous Monte Carlo simulations. A quasi-2D model is then used to calculate the drain-current voltage characteristics. The effects of the spontaneous and piezoelectrically induced polarization fields in the AlGaN/GaN HEMT are incorporated into the model by the boundary condition at the hetero-interface. That is the electric displacement must be continuous at the hetero-interface. The model used to characterize the AlGaN/GaN HEMT device is based on a nonlinear polarization model since the macroscopic polarization of nitride alloys is indeed nonlinear as a function of composition. Since a fair comparison of the GaAs and GaN HEMTs requires comparable device structures, we have chosen a device structure for which experimental data are available for GaAs. Unfortunately, no experimental data for this particular structure in GaN exists. Therefore, the model is calibrated for the GaAs calculation to experimental data. Previous calculations were made comparing the GaN calculations to experiment but for different structures. Nevertheless, excellent agreement was obtained between the model and experiment for GaN HEMTs indicating that the model can be successfully employed to the study of GaN HEMTs. Using this model, we have examined six different performance measures for GaN and GaAs HEMTs. The six performance measures are the sheet carrier concentration, capacitance-voltage characteristics, maximum drain current at similar gate and drain bias, pinch-off voltage, transconductance and cutoff frequency. It is found that the maximum drain current and the magnitude of the pinch-off voltage in the GaN device are larger than those in the GaAs device due to a significantly larger sheet charge induced by the polarization field at the hetero-interface. On the other hand, the maximum transconductance in the GaAs device is higher than that in the GaN device. This is because the low-field mobility in the GaAs HEMT is significantly larger than that in the GaN HEMT. Also increments of the sheet charge in the 2D channel with respect to a variation of the gate voltage (DQ2DEG/DVg) in the GaAs device are higher than that in the GaN device. Both of these two factors cause a higher peak transconductance in the GaAs device. The maximum cutoff frequency is very close in these two devices since the GaAs device has a larger Cgs corresponding to its peak gm. However, the transconductance in the GaN device decreases slowly as the gate voltage switches to a forward bias. This indicates that the GaN device can be employed in a larger voltage swing. Consequently, the GaN HEMT exhibits the potential for high power and high frequency applications.

Introduction

Due to the superior material properties such as high breakdown voltage, high saturation velocity, and high thermal conductivity, III-nitride devices are expected to offer better high frequency, high power, and high temperature performance compared to conventional Si and GaAs devices. The AlGaN/GaN high electron mobility transistor (HEMT) is one of the most promising devices because of its modulation-doped structure and a significantly larger polarization field induced at the hetero-interface. Recent progress in growth and process technology has already led to very impressive results for AlGaN/GaN HEMTs. A transconductance as high as 270 mS/mm has been achieved in devices with 0.7 μm gate length [1]. AlGaN/GaN HEMTs grown on a SiC substrate with a power density as high as 9.8 W/mm at 8 GHz have also been demonstrated, which is about ten times higher than GaAs-based FETs [2], and current gain cutoff frequencies of 101 GHz have been reported [3]. Recent studies show that there is about a 20% deviation for the sheet charge density in the AlGaN system between theoretical predictions and experimental measurements [4]. Nitride device modeling [5-10] has assumed that the polarization in nitride alloys is linearly interpolated from the values of the parent binary compounds [11-12]. However, recent research indicates that the macroscopic polarization in III-nitride alloys is a nonlinear function of the material composition [4]. Moreover, recent experimental measurements in AlGaN/GaN heterostructures illustrate that reasonable agreement is obtained using the nonlinear polarization formulation [4,13-14]. Also our research has shown that excellent agreement between the simulation and experimental data is obtained when the nonlinear polarization formulation is employed in the AlGaN/GaN HEMT model [15].

Since both the spontaneous and piezoelectric polarization fields can induce a significantly larger sheet charge and alter the band bending at the AlGaN/GaN hetero-interface, the device performance in the AlGaN/GaN HEMT is quite different from that in the AlGaAs/GaAs HEMT. Therefore, it is important to make a comparison between AlGaN/GaN and AlGaAs/GaAs HEMTs to illustrate how this natural advantage translates into improved device performance. 

Several theoretical AlGaN/GaN HEMT simulation models have been presented and reported in the literature [5-10]. Rashmi’s group presented analytical models including a one-dimensional (1D) and two-dimensional (2D) model to examine DC and RF characteristics of HEMTs [5-7]. A full band Monte Carlo simulation also was presented by Ando et al [8]. Sacconi et al. investigated DC performance by using a quasi-2D model [9]. Within the framework of the gradual channel approximation, Albrecht et al. included the thermal effects in the I-V calculation of a GaN/AlGaN HEMT [10]. Though much progress has been made by these researchers, to the author’s knowledge this is the first theoretical study of the comparison between AlGaN/GaN and AlGaAs/GaAs HEMTs with the same basic geometric structure. The AlGaN/GaN HEMT model includes a nonlinear formulation of the polarization effects [15].

In order to affect a reasonable comparison of the AlGaAs/GaAs and AlGaN/GaN HEMTs comparable device structures must be examined. Unfortunately, experimental data are only available for the specific structure examined for the AlGaAs/GaAs HEMT. Thus the  AlGaAs/GaAs HEMT device calculation is calibrated by using measured data. In an earlier work, we have compared our calculations of AlGaN/GaN HEMTs to other structures for which experimental data were available [15]. Excellent agreement was obtained thus ensuring that the present calculations for the AlGaN/GaN device have also been calibrated albeit to different device structures.

 The sheet charge density, a function of gate voltage, is calculated from the Schroedinger and Poisson equations self-consistently. The carrier mobility in the AlGaN/GaN HEMT is extracted from Monte Carlo simulation [16]. Based on the self-consistent charge control model and the field-dependent mobility, a quasi-2D model is used to calculate the drain-current characteristics in the HEMT devices. The most important characteristics including the sheet charge density vs. gate voltage, gate to source capacitance, drain current-voltage, transconductance, and cut-off frequency are used to examine the performance of GaN and GaAs HEMTs.