Effect of Polarization Coulomb Field Scattering on Electrical Properties of the 70-nm Gate-Length AlGaN/GaN HEMTs

This research presents the first experimental observation of the enhancement of the polarization Coulomb field (PCF) scattering by aggressive lateral scaling of GaN HEMTs. By decreasing the source-drain distance to 300 nm through n+-GaN ohmic regrowth, 70-nm gate AlGaN/GaN HEMTs achieved an extremely low electron mobility. Different from the electron mobility of the traditional device, which was determined by polar optical phonon scattering, the electron mobility of the 70-nm gate AlGaN/GaN HEMTs was dominated by PCF scattering due to the enhanced nonuniform strain distribution of the AlGaN barrier layer. Furthermore, compared with the parasitic access resistance at gate-source voltage VGS = 0 V, the parasitic access resistance at VGS = −2.5 V showed an increase of approximately 700%, which was also responsible for the enhanced PCF scattering.

voltage V DS of 100 mV at different gate biases were used. The main scattering mechanisms in the AlGaN/GaN HEMTs include polar optical phonon (POP), polarization Coulomb field (PCF), interface roughness (IFR), acoustic phonon (AP), and dislocation (DIS) scatterings 25,28,31,[36][37][38][39][40][41] . Based on the two-dimensional scattering theory, the 2DEG electron mobility was obtained by applying the self-consistent iteration calculation 25,26,42 , as shown in Fig. 3. The detailed parameters and calculation process are the same as in ref. 42 . As the gate-source voltage was decreased from 0 V to −2.5 V, the 2DEG electron mobility of the three samples decreased. The detailed electron mobility determined by POP, IFR, AP, and DIS scatterings (labeled as μ POP , μ IFR , μ AP , and μ DIS , respectively) was the same for the three samples 25,26 . However, the electron mobility determined by PCF scattering, labeled as μ PCF , differed among the three samples. The PCF scattering in the 70-nm gate AlGaN/GaN HEMTs was the strongest scattering mechanism and obviously dominated the total electron mobility variation, causing the electron mobility to decrease with the gate bias. The electron mobility of Sample 2 was lower than that of Sample 1 but higher than that of Sample 3. This means that an increase in the source-drain distance or gate width can decrease the electron mobility.
Before the device processing, the strain distribution in the AlGaN barrier layer was consistent, and the polarization charges at the AlGaN/GaN interface were uniform [21][22][23][24] . Because of the converse piezoelectric effect, the gate bias can change the strain of the AlGaN barrier layer under the gate region, causing the strain distribution of the AlGaN barrier layer to become nonuniform [25][26][27] . The device scaling makes the nonuniformly distributed strain more obvious. This means that the non-uniform distribution of the polarization charges is enhanced. The PCF scattering, which originates from the nonuniform distribution of the polarization charges at the AlGaN/ GaN interface, is also enhanced with the device scaling 25,26,28,[31][32][33] . The difference between the non-uniformly distributed polarization charges and the uniformly distributed ones is defined as the additional polarization charges. The negative gate bias decreases the tensile strain of the AlGaN barrier layer and reduces the polarization charges under the gate region [25][26][27] . When the effect of the PCF scattering on the electron under the gate region is considered, as shown in Fig. 4(a), the positive additional polarization charges are located at the gate-source and gate-drain regions, which are increased with the decrease of the gate bias [25][26][27] . The PCF scattering is enhanced by the increased additional polarization charges [25][26][27][31][32][33] , causing the electron mobility to decrease with the decreased gate bias. Samples 1 and 2 have the same gate length and width. However, Sample 2 has bigger gate-source and gate-drain regions, as well as a larger number of additional polarization charges, compared with Sample 1. This means that the PCF scattering of Sample 2 is stronger than that of Sample 1. Therefore, Sample 2 has lower electron mobility. Samples 2 and 3 have the same gate length, gate-source spacing, and gate-drain spacing. Because Sample 3 has a bigger gate width, the increased additional polarization charges originating from this increased gate width enhances the PCF scattering 34,35 ; thus, Sample 3 has lower electron mobility than Sample 2. Figure 5(a and c) shows the calculated parasitic access resistance versus the gate-source voltage. The gate-source parasitic access resistance (R S ) and the gate-drain parasitic access resistance (R D ) are obviously increased with the decreased gate-source bias. The difference between the parasitic access resistance under the negative gate bias (R S and R D ) and that under the zero gate bias (R S0 and R D0 ) was also obtained, as shown in   5(b and d). Compared with the parasitic access resistance at gate-source voltage V GS = 0 V, the parasitic access resistance at V GS = −2.5 V showed an increase of approximately 700%. Furthermore, Samples 1 and 3 both showed a bigger increase than Sample 2. This means that the device with a smaller gate-source/gate-drain distance or a larger gate width has a more obvious increase in parasitic access resistance with the decreased gate-source bias.
Under the negative gate-source voltage, the polarization charges under the gate region are lower than those under the gate-source/gate-drain region. Therefore, considering the electrons in the gate-source/gate-drain channel, as shown in Fig. 4(b), the negative additional polarization charges were under the gate region, and increased with the decreased gate-source bias 24,25 . The increased additional polarization charges enhanced the PCF scattering and increased the parasitic access resistance. The additional polarization charges under the gate region are the same for Samples 1 and 2. However, Sample 1 has smaller gate-source/gate-drain distances than Sample 2. Thus, the influence of the additional polarization charges on the smaller gate-source/gate-drain distances is stronger 24,25 , causing the PCF scattering to be enhanced and making the increase in the parasitic access resistances more obvious. Sample 3 has a bigger gate width, and the additional polarization charges under the gate region are larger for Sample 3 than for Sample 2. Furthermore, the PCF scattering is stronger and the parasitic access resistances showed a larger increase for Sample 3 34,35 .

Conclusions
In summary, the electron mobility and parasitic access resistances versus the gate-source voltage for the 70-nm gate AlGaN/GaN HEMTs were obtained. The PCF scattering, originating from the strain variation of the AlGaN barrier layer, was shown to have a more significant influence on the device characterization with device scaling. This could present a possible approach toward improving the performance of 70-nm gate AlGaN/GaN HEMTs by decreasing the PCF scattering.

Methods
Sample fabrication. The AlGaN/GaN heterostructures were grown on a sapphire substrate by molecular beam epitaxy (MBE) (see Fig. 1(a) for a more detailed material structure description). The sheet electron concentration and electron mobility obtained from Hall measurements were 9.27 × 10 12 cm −2 and 2020 cm 2 /V•s, respectively. The AlGaN barrier layer in the ohmic contact regions was etched into the GaN channel layer by inductively coupled plasma reactive ion etching (ICP-RIE), followed by MOCVD re-growth of highly Si doped n + -GaN (3 × 10 19 cm −3 ). The source and drain electrodes were formed by using non-alloyed Ti/Pt. The transmission-line matrix measurements showed that the ohmic contact resistance R C was 0.58 Ω•mm. Ni/Au T-shaped gate with a 70-nm gate length (L G ) was fabricated and located in the middle of the source and drain contacts. Finally, the devices were passivated by using 50-nm-thick SiN deposited by PECVD. As shown in Fig. 1(b), the device with L SD = 300 nm and W G = 20 μm was labeled as Sample 1, the device with L SD = 600 nm and W G = 20 μm as Sample 2, the device with L SD = 600 nm and W G = 40 μm as Sample 3.
Measurements. The current-voltage (I-V) measurements were carried out at room temperature by using an Agilent B1500A semiconductor parameter analyzer.