Investigation of the Effect of Different SiN_x Thicknesses on the Characteristics of AlGaN/GaN High-Electron-Mobility Transistors in Ka-Band

Abstract

The effect of different SiN_x thicknesses on the performance of AlGaN/GaN high-electron-mobility transistors (HEMTs) was investigated in this paper. The current, transconductance (G_m), cut-off frequency (f_T), maximum oscillation frequency (f_max), power performance, and output third-order intercept point (OIP3) of devices with three different SiN_x thicknesses (150 nm, 200 nm, and 250 nm) were measured and analyzed. The DC measurements revealed an increase in both the drain-source current (I_DS) and G_m values of the device with increasing SiN_x thickness. The S-parameter measurement results show that devices with a higher SiN_x thickness exhibit improved f_T and f_max. Regarding power performance, thicker SiN_x devices also improve the output power density (P_out) and power-added efficiency (PAE) in the Ka-band. In addition, the two-tone measurement results at 28 GHz show that the OIP3 increased from 35.60 dBm to 40.87 dBm as the SiN_x thickness increased from 150 nm to 250 nm. The device’s characteristics improved by appropriately increasing the SiN_x thickness.

1. Introduction

Recently, extensive research has been conducted on the application of AlGaN/GaN high-electron-mobility transistors (HEMTs) in high-frequency microwave power devices, ranging from essential everyday needs to defense applications such as 5G small-cell base stations, satellite communications, defense and military applications, and industrial radar [1,2,3,4,5,6]. Compared to other III-V compound semiconductors, GaN materials possess excellent characteristics, such as a wide bandgap, high electron mobility, high saturation velocity, and a high breakdown field [7,8,9]. Additionally, the combination of GaN and AlGaN layers forms a two-dimensional electron gas (2DEG) at the interface, which is characterized by a high electron concentration. These excellent characteristics allow GaN devices to maintain high output power at higher frequencies [10,11,12].

In the face of the arrival and commercialization of the fifth generation of mobile communication (5G), characterized by faster transmission speeds, low latency, large bandwidth, and high density, it is advantageous for the development of services such as big data, the Internet of Things (IoT), and Artificial Intelligence (AI) [13,14]. These advancements can drive high-quality audiovisual entertainment, self-driving cars, drones, smart healthcare, intelligent factories, smart retail, smart cities, and other value-added innovative applications. This has become a focal point of development in countries around the world. Due to the growing demand for high-speed transmission in 5G communication, there is an increasing need for high-frequency devices operating in the Ka-band to support wider bandwidth requirements [15,16,17,18].

In wireless communication, when two or more signals pass through nonlinear amplifier devices, intermodulation distortion (IMD) occurs. Because IMD generates unwanted spurious emissions, it leads to an increase in the occupied bandwidth and interferes with adjacent channels, ultimately reducing audio clarity or increasing spectral occupancy. Among the nonlinear distortions generated by devices within the system, the third-order intermodulation distortion (IM3) is the most influential factor [19,20,21]. It is closest to the main signal, difficult to filter out, and easily affects the signal quality. Therefore, improving device linearity by enhancing the performance metrics, including the output third-order intercept point (OIP3) and reducing IM3, is an important concern. The literature has put forward various methods to enhance HEMT linearity, including δ-Doped [22], gate dielectric [23], N-Polar GaN MIS-HEMT [24], selective-area charge implantation [25], fin-like configuration [26], and etched-fin gate structure [27]. For traditional AlGaN/GaN HEMTs, the transconductance (G_m) of the device rapidly decreases after reaching its peak as the gate-source voltage (V_GS) increases, resulting in severe nonlinearity [28]. To improve the DC characteristics of a device, many studies have proposed methods such as modifying the device structure or reducing the gate leakage current to enhance the overall DC and RF performance of the device [29,30,31,32]. However, there is limited literature on improving device performance by increasing the SiN_x thickness [33].

In this study, we investigate the effect of different SiN_x thicknesses on the characteristics of AlGaN/GaN HEMT devices in the Ka-band. The DC characteristics, S-parameter measurement, power performance, and OIP3 of devices with three different SiN_x thicknesses (150 nm, 200 nm, and 250 nm) were measured. The measurement results demonstrate an enhancement in the device’s performance as the SiN_x thickness increases to 250 nm.

2. Materials and Methods

The AlGaN/GaN HEMT heterostructure used in this study was grown on a 4-inch SiC substrate using metal–organic chemical vapor deposition (MOCVD). A cross-sectional schematic diagram of the AlGaN/GaN HEMT is displayed in Figure 1. The dimensions of the device were as follows: L_g is 150 nm, H_head is 350 nm, and the thicknesses of the SiN_x layer (t_SiNx) were 150 nm, 200 nm, or 250 nm. The device fabrication process flow is shown in Figure 2. Device fabrication began with the Ohmic contact formation of Ti/Al/Ni/Au. The Ohmic metal layers were deposited using an E-gun evaporator, followed by rapid thermal annealing (RTA) at 830 °C for 30 s in a nitrogen atmosphere to form Ohmic contacts. Next, device isolation was achieved using boron ion implantation. A photoresist was used as a mask to protect the active region, and boron ions (B¹¹⁺) were implanted using an ion implantation machine to complete the device isolation process. To define the gate region, the first SiN_x passivation layer was deposited using plasma-enhanced chemical vapor deposition (PECVD). The thicknesses of the first SiN_x layers (t_SiNx) were 150 nm, 200 nm, and 250 nm. Afterward, stepper lithography was performed twice to complete the gate fabrication. The first step of stepper lithography defined the areas for etching the first SiN_x film, followed by etching using inductively coupled plasma (ICP). The second step of stepper lithography defined the areas for depositing the gate metal. After two steps of stepper lithography, the Ni/Au gate metal stack was deposited using an E-gun evaporator to complete the gate fabrication. Before depositing the gate metal, the wafer was immersed in a diluted HCl solution for one minute to remove any natural oxides from the AlGaN barrier layer. After gate fabrication was completed, the first SiN_x layer was removed using ICP, followed by the deposition of the second SiN_x layer using PECVD. Finally, for metallization, holes were opened in the passivation film via ICP, and then a 1.5 µm thickness of Au metallization layer was deposited.

3. Results and Discussion

The DC characteristics of the 4 × 50 μm AlGaN/GaN HEMTs were measured using Keysight B1505A. Figure 3 shows the typical I_DS-V_GS and G_m-V_GS characteristics at V_DS = 20 V for 4 × 50 μm AlGaN/GaN HEMT devices with different SiN_x thicknesses (150 nm, 200 nm, and 250 nm). While the SiN_x thickness increases from 150 nm to 200 nm, the device exhibits an increase in the saturation drain current (I_DSS) from 766.5 mA/mm to 861 mA/mm, and the maximum transconductance (G_m,max) increases from 290.1 mS/mm to 312.9 mS/mm. Increasing the SiN_x thickness to 250 nm no longer significantly increases the I_DSS and G_m,max of the device, as shown in

Table 1. DC characteristics of the devices with various SiN_x thicknesses at V_DS = 20 V.

Parameters	tSiNx
	150 nm	200 nm	250 nm
IDSS (mA/mm)	766.5	861	861.5
Gm,max (mS/mm)	290.1	312.9	320.9
Vth (V)	−3.1	−3.7	−3.7
GVS (V)	1.0	1.1	1.2
$a_1$	−0.52938	−0.47639	−0.51950
$a_3$	−0.25548	−0.22635	−0.24283
$a_3$/$a_1$	0.48260	0.47514	0.46743
$a_5$	−0.00589	−0.00491	−0.00528
$a_5$/$a_1$	0.01113	0.01031	0.01016

. Compared to other thicknesses, the device with a SiN_x thickness of 250 nm exhibits the highest I_DSS and G_m,max. The increase in the device’s current and G_m is due to the increase in the thickness of the SiN_x passivation layer, which leads to an increase in the biaxial tensile stress applied to the AlGaN layer, further increasing the surface charge density at the heterointerface [34,35]. Furthermore, the threshold voltage (V_th) for a 150 nm SiN_x thickness is −3.1 V, whereas the V_th for SiN_x thicknesses of both 200 nm and 250 nm is −3.7 V. The V_th is defined as the V_GS when I_DS = 1 mA/mm.

To further analyze the linearity, a polynomial curve fitting technique is applied to the transfer characteristic function of these devices, as shown in Equation (1) [36].

$$I_{DS}\left(V_{GS}\right)=a_0+a_1{V}_{GS}+a_2{V}_{GS}^2+a_3{V}_{GS}^3+a_4{V}_{GS}^4+a_5{V}_{GS}^5$$

(1)

In addition, Equation (2) can be used to estimate the IM3 levels generated by the device [37]. Equation (3) illustrates the relationship between OIP3, transconductance (G_m), drain conductance (G_ds), and G_m″ [38].

$$\mathrm{IM}3=\frac{3}{8}{a}_3{A}^3+\frac{50}{32}{a}_5{A}^5$$

(2)

$$\mathrm{OIP}3\propto \frac{{\left(G_m\right)}^3}{G_m^{″}{G}_{ds}^2{R}_L}$$

(3)

The linearity of the device is primarily determined by the flatness of the G_m profile. In other words, the coefficient $a_1$ should be larger, whereas the ratios $a_3$/$a_1$ and $a_5$/$a_1$ should be minimized [39]. The polynomial curve fitting coefficients and the gate voltage swing (GVS), defined as a 10% drop in G_m,max, are presented in Table 1. Compared to devices with SiN_x thicknesses of 150 nm and 200 nm, the device with a SiN_x thickness of 250 nm exhibits lower $a_3$/$a_1$ and $a_5$/$a_1$ values, indicating that the device with a SiN_x thickness of 250 nm has better linearity. Furthermore, the two-tone measurement results at the end of the article also indicate an improvement in the linearity of the devices with thicker SiN_x.

Next, the S-parameters of the 4 × 50 μm devices with different SiN_x thicknesses (150, 200, and 250 nm) were measured using a Keysight E8361A PNA Network Analyzer in the frequency range from 100 MHz to 67 GHz, as shown in Figure 4. The values of f_T can be obtained by converting the measured S-parameters to H-parameters, whereas the values of f_max can be obtained by extrapolating the gain curve with a slope of −20 dB/decade. When increasing the SiN_x thickness from 150 nm to 250 nm, the f_T value increases from 33.7 GHz to 53.1 GHz, as indicated in

Table 2. The f_T and f_max values of the devices with SiN_x thicknesses of (a) 150 nm, (b) 200 nm, and (c) 250 nm.

Parameters	tSiNx
	150 nm	200 nm	250 nm
fT	33.7	51.9	53.1
fmax	76.9	132.9	138.9

. Furthermore, there is an increase in f_max, which improves from 76.9 GHz to 138.9 GHz. The gate stem height changes due to the varying SiN_x thicknesses. According to the definitions of f_T and f_max in Equations (4) and (5), increasing the SiNx thickness to enhance the device’s G_m will result in higher f_T and f_max values [40]. Increasing the gate stem height also reduces C_gs and C_gd, further enhancing the f_T and f_max of the device [40]. Moreover, C_gs will also be affected when the gate is facing towards the drain side [41]. However, increasing the gate stem height to 300 nm does not lead to a further significant reduction in the capacitance [42]. Additionally, higher gate stem heights contribute to an increase in the gate resistance (R_g) of the device.

$$f_T=\frac{g_m/2\pi }{\left(C_{gs}+C_{gd}\right)\times [1+(R_s+R_d)g_0]+{g_{m}C}_{gd}(R_s+R_d)}\cong \frac{g_m}{2\pi \left(C_{gs}+C_{gd}\right)}$$

(4)

$$f_{max}=\frac{f_T}{2\sqrt{g_0(R_g+R_s)+2f_T{C}_{gd}{R}_g}}$$

(5)

Figure 5 shows the load-pull measurement results at 28 GHz for 4 × 50 μm devices with three different SiN_x thicknesses (150 nm, 200 nm, and 250 nm). All devices were measured at a bias of class AB (1/4 I_DSS, I_DSS = I_DS at V_GS = 0) and V_DS = 20 V. When increasing the SiN_x thickness of the device from 150 nm to 200 nm, the maximum output power density (P_out,max) increases from 2.21 W/mm to 2.61 W/mm. Furthermore, when increasing the SiN_x thickness from 200 nm to 250 nm, P_out,max increases from 2.61 W/mm to 2.72 W/mm. From the load-pull measurement results, it can be observed that the P_out, PAE, and linear gain of the device show improvement when the SiN_x thickness is increased from 150 nm to 250 nm. The device with a 250 nm SiN_x thickness exhibits the best power performance, with a P_out,max of 2.72 W/mm, PAE_max of 31.78%, and a linear gain of 10.68 dB. Therefore, increasing the thickness of SiN_x improves the power performance of the device.

To evaluate the linearity of the devices, two-tone measurements were performed at 28 GHz with a 5 MHz tone spacing using the Keysight E8267D signal generator, N9030B spectrum analyzer, U84888A power sensor, N6700C modular power system, and an AMP4072 power amplifier. Figure 6 shows the two-tone measurement results at 28 GHz for 4 × 50 μm devices with different SiN_x thicknesses. During the measurements, each device was measured at a bias of class AB (1/4 I_DSS, I_DSS = I_DS at V_GS = 0) and V_DS = 20 V. The measurement results show that the OIP3 value improves from 35.60 dBm to 40.87 dBm when the SiN_x thickness of the device is increased from 150 nm to 250 nm. Based on the previous polynomial curve fitting results, it can be observed that as the SiN_x thickness increases, the ratios of $a_3$/$a_1$ and $a_5$/$a_1$ decrease, resulting in an improvement in the device’s linearity. Moreover, the increase in OIP3 could be attributed to the increase in SiN_x thickness, which leads to a higher current and changes in the G_m distribution, thereby enhancing the linearity of the device.

4. Conclusions

In this paper, the effects of increasing the SiN_x thickness on the characteristics of AlGaN/GaN HEMTs were investigated. The DC measurement results indicate that as the SiN_x thickness increases, the current and G_m for the device also increase. The device’s f_T and f_max also improve with an increase in the SiN_x thickness. For the 4 × 50 μm device with a SiN_x thickness of 250 nm, the load-pull measurement results at 28 GHz show a P_out,max of 2.72 W/mm, a PAE_max of 31.78%, and a linear gain of 10.68 dB. Furthermore, the two-tone measurement results show that the improvement in OIP3 is due to the increase in the current for the device, resulting in a change in the G_m profile. The final results show that when the SiN_x thickness of the device increases from 150 nm to 200 nm, and even up to 250 nm, its characteristics such as I_DSS, G_m,max, P_out,max, f_T, f_max, and OIP3 improve. Thus, we have demonstrated that device characteristics can be improved by increasing the SiN_x thickness to a sufficient value, such as 250 nm.