Comparison between AlGaN/GaN and AlGaAs/GaAs HEMTs

To make a comparison between AlGaN/GaN and AlGaAs/GaAs HEMTs, we consider a representative AlGaAs/GaAs HEMT structure that consists of a 3nm undoped Al_0.27Ga_0.73As spacer and a 27nm thick Al_0.27Ga_0.73As doped layer with a doping level of 2×10¹⁸ cm^-3 [28]. The same geometry structure and doping level in the doped layer are used in the Al_0.15Ga_0.85N/GaN HEMT. The material compositions are chosen so as to yield a structure with similar conduction band edge discontinuities and comparable well depths. The conduction band discontinuities are 0.286V and 0.25V are for the Al_0.27Ga_0.73As/GaAs and Al_0.15Ga_0.85N/GaN HEMT, respectively. The material parameters used in the calculation are compiled in Ref. 15 Table I.

Figures 6 and 7 show the calculated sheet carrier concentration as a function of gate voltage in the Al_0.27Ga_0.73As/GaAs and Al_0.15Ga_0.85N/GaN heterostructures, respectively. The free electrons and neutralized donors in the doped layer are calculated by integrating n_3D(z) and  in equations (1)-(2) over the barrier thickness. As the gate voltage increases, the free electrons in the 2D channel tend to go into the doped layer as 3D free electrons or to neutralize the donors. Therefore, the sheet carrier concentration in the 2D channel saturates, which limits the maximum voltage swing in the HEMTs. The maximum sheet charge density of 5.12×10¹² cm^-2 in the Al_0.15Ga_0.85N/GaN heterostructure is significantly larger than that of 1.14×10¹² cm^-2 in the Al_0.27Ga_0.73As/GaAs heterostructure due to a larger polarization field induced at the AlGaN/GaN hetero-interface. Also, to deplete this larger sheet charge at the hetero-interface, the threshold voltage in the GaN device becomes more negative, corresponding to a larger voltage swing.

The gate capacitance is calculated by differentiating the sheet charge density under the gate with respect to the applied gate voltage. The calculated capacitance per unit area is shown in Figures 8 and 9 as a function of gate bias for the Al_0.27Ga_0.73As/GaAs and Al_0.15Ga_0.85N/GaN heterostructures respectively. The curve denoted by C_2DEG is the capacitance due to electrons in the 2D channel, C_3D is due to electrons in the AlGaAs or AlGaN layer, and C_Neu is due to neutralized donors in the barrier. As can be seen from Figures 8-9, near the threshold voltage, donors throughout the doped layer are fully ionized with no free electrons in the barrier, and 2D electrons are distributed deep into the GaAs or GaN channel, resulting in a small capacitance. As the gate voltage increases, the capacitance increases quickly because of the band bending and an increase in the confined electrons induced by the gate voltage. Once the 2DEG is formed, the effective position of the 2DEG from the hetero-interface is no longer noticeably influenced. Hence, the capacitance remains almost constant in a small voltage range around the maximum capacitance. When the gate voltage is further forward biased, donor neutralization and parallel conduction become significant. As a consequence, the capacitance due to the 2DEG in the channel decreases while the total capacitance increases. It is noted that the peak capacitance of 0.276 mF/cm² due to 2DEG in the GaAs device is larger than that of 0.247 mF/cm^-2 in the GaN device. The peak capacitance in the GaAs device is higher than that in the GaN device since in the GaAs device the gate voltage alone mainly controls the sheet charge. In contrast within the GaN device it is the combined effect of the polarization field and the gate voltage that controls the sheet charge. In addition, as the gate voltage is further forward biased, the total gate capacitance in the GaAs device increases rapidly. Thus the cutoff frequency drops quickly as the gate voltage increases.

A comparison of the I-V characteristics for the Al_0.27Ga_0.73As/GaAs and Al_0.15Ga_0.85N/GaN HEMTs is shown by Figures 10-11. As can be seen from Figure 10, the calculated I-V characteristic of the Al_0.27Ga_0.73As/GaAs HEMT is plotted along with experimental measurements [28]. Relatively good agreement between the experimental data and simulation results is obtained. To obtain the best agreement with the experimental data, the critical field e_c, saturation velocity v_sat, and the low-field mobility m₀ in Eq.(5) used in the model are 3.5×10⁵ V/m, 1.6×10⁷ cm/s, and 9,143 cm²/V-s, respectively. The maximum sheet carrier concentration has been set to1.08×10¹² cm^-2 for gate voltages greater than 0.5V. To consider a finite output conductance, the drain current is multiplied by (1+lV_DS) in the saturated current regime [29]. Here l is the fitting parameter and is 0.07 in this calculation. Furthermore, the source and drain resistances are also taken into account and have been adjusted to 6W.

The maximum drain current of 66.3 mA at a gate bias of 0.9V and a drain bias of 5V in the Al_0.15Ga_0.85N/GaN HEMT is larger than that of 36.3 mA for the Al_0.27Ga_0.73As/GaAs HEMT with the same biases. The conducting channel within the Al_0.15Ga_0.85N/GaN device closes-off completely at V_gs = -3.98V with a drain current as low as 1.0 mA/mm at V_ds=5V. The corresponding voltage in the Al_0.27Ga_0.73As/GaAs device is V_gs = -0.44V.  Moreover, in the Al_0.15Ga_0.85N/GaN HEMT, the on-off state current ratio of 663 is larger than that of 363 in the Al_0.27Ga_0.73As/GaAs HEMT.

The transconductance and cutoff frequency for both devices are shown in Figure 12 at a drain bias of 5V. The maximum transconductance of 292 mS/mm in the Al_0.27Ga_0.73As/GaAs device with a V_gs= 0.175V and I_ds=124 mA/mm is larger than that of 187 mS/mm in the Al_0.15Ga_0.85N/GaN HEMT with a V_gs= -1V and I_ds=347 mA/mm. The maximum g_m is influenced by two factors: capacitance due to 2DEG and low-field mobility. The 2DEG capacitance indicates how much charge can be supplied to the 2D channel with respect to a variation of the gate voltage. The excess supply charge directly contributes to an increase of the drain current. Therefore, the maximum g_m strongly depends on the peak of the 2DEG capacitance. The maximum 2DEG capacitance of the Al_0.27Ga_0.73As/GaAs device is larger than that of Al_0.15Ga_0.85N/GaN structure as shown in Figures 8-9. In addition to the modulation of charge in the 2D channel, the electrons in the channel are not all traveling at the saturated or peak electron velocity except those at the drain edges. Hence, g_m is also proportional to the low-field mobility. The low-field mobility of 9143 cm²/V-s in the Al_0.27Ga_0.73As/GaAs device is significantly larger than that of 500 cm²/V-s in the Al_0.15Ga_0.85N/GaN device in this plot. Therefore, both these two factors cause a higher peak transconductance in the Al_0.27Ga_0.73As/GaAs device. However, the Al_0.15Ga_0.85N/GaN device still exhibits a transconductance of 146 mS/mm at a drain current of 663 mA/mm and a gate bias of 0.9 V. For the GaAs device, the transconductance drops more precipitously at 0.9 V gate bias to a value of 66.3 mS/mm.

Consequently, the GaN HEMT has a larger useful voltage swing implying better high power performance.

Figure 12 also shows the calculated cutoff frequency as a function of gate voltage for the GaAs and GaN HEMT structures. When the gate is forward biased, the parallel conduction and donor neutralization cause a decrease of g_m and also lead to an increase of C_gs. Therefore, the cutoff frequency falls off at forward biases. As can be seen from Figure 12, the maximum cutoff frequency (f_T)_max of 13.8 GHz in the GaN device is very closed to that of 13.6 GHz in the GaAs device. This difference is attributable to the fact that the value of C_gs corresponding to (f_T)_max in the GaN device is less than that in the GaAs device.