Operational Characteristics of AlGaN/GaN High-Electron-Mobility Transistors with Various Dielectric Passivation Structures for High-Power and High-Frequency Operations: A Simulation Study

Abstract

This study investigates the operational characteristics of AlGaN/GaN high-electron-mobility transistors (HEMTs) by employing various passivation materials with different dielectric constants and passivation structures. To ensure the simulation reliability, the parameters were calibrated based on the measured data from the fabricated basic Si₃N₄ passivation structure of the HEMT. The Si₃N₄ passivation material was replaced with high-k materials, such as Al₂O₃ and HfO₂, to improve the breakdown voltage. The Al₂O₃ and HfO₂ passivation structures achieved breakdown voltage improvements of 6.62% and 17.45%, respectively, compared to the basic Si₃N₄ passivation structure. However, the increased parasitic capacitances reduced the cut-off frequency. To mitigate this reduction, the operational characteristics of hybrid and partial passivation structures were analyzed. Compared with the HfO₂ passivation structure, the HfO₂ partial passivation structure exhibited a 7.6% reduction in breakdown voltage but a substantial 82.76% increase in cut-off frequency. In addition, the HfO₂ partial passivation structure exhibited the highest Johnson’s figure of merit. Consequently, considering the trade-off relationship between breakdown voltage and frequency characteristics, the HfO₂ partial passivation structure emerged as a promising candidate for high-power and high-frequency AlGaN/GaN HEMT applications.

1. Introduction

AlGaN/GaN high-electron-mobility transistors (HEMTs) are increasingly being adopted for high-power applications due to their superior material properties, such as a wide energy bandgap (3.4 eV) and a high critical electric field (3.39 MV/cm). GaN exhibits a higher electron saturation velocity and current density than conventional silicon and gallium arsenide [1,2,3]. AlGaN/GaN heterostructure HEMTs exhibit exceptional performance due to the formation of a two-dimensional electron gas (2-DEG) at the surface, which arises from spontaneous and piezoelectric polarization effects [4,5]. Consequently, these HEMTs are extensively employed in power electronics and devices that operate under high-power and high-frequency conditions. To optimize these outstanding characteristics, we developed various dielectric passivation structures that enhance the breakdown voltage (V_BD) and cut-off frequency (f_T) simultaneously. After conducting the simulation for each structure, Johnson’s figure of merit (JFOM), which can be expressed as V_BD × f_T, was used to evaluate the operational characteristics [6,7].

High-k materials are commonly used in the passivation layer because of their advantages. Due to its higher dielectric constant than conventional materials such as SiO₂ and Si₃N₄, a high-k material functions as a thicker dielectric layer without physically increasing its thickness. It effectively reduces leakage current under both off-state and on-state conditions [8,9,10]. In addition, high-k materials contribute to the electric field redistribution, which improves the breakdown voltage of HEMT devices [11,12]. However, the large dielectric constant of these materials also increases parasitic capacitances, such as the gate-to-source capacitance (C_GS) and gate-to-drain capacitance (C_GD), which can degrade the frequency characteristics [13]. Therefore, it is crucial to balance the trade-off between V_BD and frequency characteristics when selecting high-k materials for the passivation layer of AlGaN/GaN HEMTs [14,15].

In this study, we simulated and analyzed three distinct passivation structures: entire passivation (EP), hybrid passivation (HP), and partial passivation (PP) structures, using different dielectric materials such as Si₃N₄, Al₂O₃, and HfO₂ to improve the V_BD with minimal degradation in frequency characteristics. The JFOM was calculated and analyzed for seven different structures in terms of the trade-off relationship between V_BD and f_T. First, the Al₂O₃ and HfO₂ EP structures were modeled by substituting the passivation material of the basic Si₃N₄ passivation structure. We confirmed that the EP structure with HfO₂ passivation, which has the highest dielectric constant among the dielectric materials, exhibited the highest V_BD because it effectively redistributed the electric field when a high drain voltage (V_DS) was applied. Conversely, when a high-k material was applied as the passivation layer, the parasitic capacitances also increased, leading to the degradation of f_T [16,17,18]. To minimize the degradation of f_T caused by the use of a high-k material in the passivation layer, we suggest the use of HP and PP structures to improve the frequency characteristics. The AlGaN/GaN HEMT with a properly designed dielectric passivation structure with high V_BD and f_T is expected to be a good candidate for high-power and high-frequency applications, such as GaN monolithic microwave integrated circuit power amplifiers for military radars and GaN radio frequency (RF) electronic devices for 5^th generation mobile telecommunication and autonomous driving.

2. Materials and Methods

A 0.16 μm T-gate AlGaN/GaN HEMT was fabricated, and a cross-sectional view of the unit device is shown in Figure 1a. Figure 1b shows a magnified image of the gate electrode, featuring a 0.16 μm gate foot opening in the 1st passivation layer, which is covered on top by a 2nd passivation layer.

Figure 2 illustrates a cross-sectional view of a 0.16 μm T-gate AlGaN/GaN HEMT, which was used for technology computer-aided design (TCAD) modeling. In this figure, S is the source electrode, S-FP is the source-connected field plate, G is the gate electrode, and D is the drain electrode. The specific geometric parameters of the modeled device are listed in

Table 1. Specific geometric parameters of the 0.16 μm gate foot length of the basic Si₃N₄ EP HEMT.

Parameters	Value (μm)
① LSource-Drain	4.585
② LGate-Head-Top	0.26
③ LGate-Head-Bottom	0.71
④ LGate-Foot	0.16
⑤ LGate-Drain	3.175
⑥ LGate-Source	0.7
SiC-4H substrate	5
Nucleation layer	0.02
GaN buffer	1.04
AlN insertion layer	0.001
AlGaN barrier	0.018
1st passivation	0.02
2nd passivation	0.25

.

The AlGaN/GaN HEMT was grown on top of a 4-inch SiC-4H substrate using metal–organic chemical vapor deposition. The epitaxial layers were sequentially stacked and grown as follows: a 20 nm thick nucleation layer, a 1.04 μm thick Fe-doped GaN buffer layer, a 1 nm thick AlN insertion layer, and an 18 nm thick AlGaN barrier layer with 28% Al composition. The Ti/Au/Ni/Au alloyed ohmic contacts for the source and drain electrodes were formed by rapid thermal annealing at 775 °C for 30 s. Device isolation was achieved via P⁺ ion implantation. Subsequently, a 20 nm thick Si₃N₄ layer was deposited on the AlGaN barrier layer using plasma-enhanced chemical vapor deposition (PECVD). The first metal interconnection with the source and drain electrodes was established by Ti/Au evaporation after the etching of the 1st Si₃N₄ passivation layer. A planar gate was then created using single-layer electron beam lithography. A gate foot opening of 0.16 μm was achieved by exposing a polymethyl methacrylate resist to an electron beam, followed by the removal of the 1st Si₃N₄ passivation layer beneath the gate foot opening pattern through dry etching using inductively coupled plasma. The planar gate was defined using a Ni/Au metal stack deposited via electron–beam evaporation and subsequent lift-off processes. After defining the gate shape, a 250 nm thick 2nd Si₃N₄ passivation layer was deposited for device passivation using PECVD. The source-connected field plate (S-FP) was formed using a Ti/Au metal lift-off process. Finally, wafer thinning and backside via-hole processes were conducted [19].

To accurately predict the operational characteristics of a device, it is crucial to apply appropriate simulation parameters, such as electrical and thermal parameters, for each epitaxial layer. This meticulous approach ensures reliable and consistent simulation data. Consequently, the simulation parameters were meticulously calibrated to closely align with the actual device operating characteristics. For example, to mitigate the electron punch-through effect and reduce the substrate leakage current, iron (Fe) acceptor trap doping was leveraged in the GaN buffer layer to enhance the V_BD [20]. In this simulation, a Gaussian acceptor doping profile was employed, with an acceptor doping concentration of 8.813 × 10¹⁴/cm³ at the AlGaN/GaN interface region and a peak trap concentration of 10¹⁸/cm³ [21]. In addition, a Selberherr impact ionization model was applied to simulate the V_BD. Other simulation parameters such as electron mobility and heat models were accurately controlled to obtain reliable simulation results. The specific simulation parameters applied to the GaN and AlGaN layers are summarized in

Table 2. Material parameters for the simulation at a room temperature.

Parameters	Units	GaN	AlGaN
Bandgap energy	eV	3.39	3.88
Electron affinity	eV	4.2	2.3
Dielectric constant	-	9.5	9.38
Low-field electron mobility	-	Farahmand modified Caughey–Thomas Model
High-field electron mobility	-	GANSAT Model
Electron saturation velocity	cm/s	1.9 × 107	1.12 × 107
Hole saturation velocity	cm/s	1.9 × 107	1.00 × 106
Electron Shockley–Read–Hall lifetime	s	1.0 × 10−8	1.0 × 10−8
Hole Shockley–Read–Hall lifetime	s	1.0 × 10−8	1.0 × 10−8

[22].

After determining the appropriate simulation parameters, simulations were conducted to analyze the direct current (DC) and RF characteristics. The transconductance equation can be expressed as follows:

$${\mathrm{g}}_{\mathrm{m}}={\displaystyle \frac{{\partial \mathrm{I}}_{\mathrm{D}\mathrm{S}}}{\partial {\mathrm{V}}_{\mathrm{G}\mathrm{S}}}},$$

(1)

where ${\mathrm{g}}_{\mathrm{m}}$, ${\mathrm{I}}_{\mathrm{D}\mathrm{S}}$, and ${\mathrm{V}}_{\mathrm{G}\mathrm{S}}$ denote the transconductance, drain current, and gate voltage, respectively. The electric displacement was explained by Equation (2), as follows:

$$\mathrm{D}=\mathsf{\epsilon }\mathrm{E},$$

(2)

where $\mathrm{D},\mathsf{\epsilon }$, and $\mathrm{E}$ denote the electrical displacement, dielectric constant of the material, and electric field, respectively. Before evaluating the frequency characteristics of each structure, the relationship between the parasitic capacitances, such as C_GS and C_GD, and the frequency characteristics was investigated as follows:

$$\mathrm{C}={\displaystyle \frac{\mathsf{\epsilon }\mathrm{A}}{\mathrm{d}}},$$

(3)

where $\mathrm{A}$ and $\mathrm{d}$ denote the overlapped area between two electrodes and the distance between the electrodes, respectively.

Next, f_T can be determined using Equation (4), as follows:

$${\mathrm{f}}_{\mathrm{T}}={\displaystyle \frac{{\mathrm{g}}_{\mathrm{m}}}{2\mathsf{\pi }({\mathrm{C}}_{\mathrm{G}\mathrm{S}}+{\mathrm{C}}_{\mathrm{G}\mathrm{D}})}}\approx {\displaystyle \frac{{\mathrm{g}}_{\mathrm{m}}}{2\mathsf{\pi }{\mathrm{C}}_{\mathrm{G}\mathrm{S}}}},$$

(4)

where ${\mathrm{C}}_{\mathrm{G}\mathrm{S}}$ and ${\mathrm{C}}_{\mathrm{G}\mathrm{D}}$ denote the gate-to-source capacitance and gate-to-drain capacitance, respectively. As described in Equation (4), ${\mathrm{C}}_{\mathrm{G}\mathrm{S}}$ and ${\mathrm{C}}_{\mathrm{G}\mathrm{D}}$ have an inverse relationship with f_T, which makes it crucial to minimize parasitic capacitances to maximize the frequency characteristics [23]. Therefore, we propose various dielectric passivation structures using materials with different dielectric constants, such as Si₃N₄, Al₂O₃, and HfO₂, to analyze the RF characteristics related to capacitances. The specific material characteristics of these materials are summarized in

Table 3. Material characteristics of Si₃N₄, Al₂O₃, and HfO₂.

Parameters	Units	Si3N4	Al2O3	HfO2
Dielectric constant	-	~7.5	~9	~25
Bandgap energy	eV	5.3	7.1	5.8

below [24,25].

3. Results

3.1. Matching Simulated and Measured Data for the Basic Si₃N₄ Entire Passivation Structure

To validate the simulation accuracy, a comparative analysis was conducted between the simulated and measured drain current–gate voltage (I_DS-V_GS) transfer characteristics of the fabricated Si₃N₄ EP structure device. The measured and simulated data exhibited close agreement in terms of I_DS at V_GS = 0 V (I_dss), maximum transconductance (G_m), and threshold voltage (V_th). Figure 3a compares the measured and simulated I_DS-V_GS transfer characteristics. The measured and simulated I_dss values were 817.10 and 811.99 mA/mm, respectively. Similarly, the measured and simulated maximum transconductance values were 400.39 and 397.65 mS/mm, respectively. Furthermore, the measured V_th was −3.1 V, and the simulated value was −3 V. These results confirm a close match between the measured and simulated data for I_dss, G_m, and V_th with error rates of 0.6%, 0.7%, and 3.2%, respectively.

The measured and simulated f_T values of the basic Si₃N₄ EP structure are shown in Figure 3b. The RF characteristics were evaluated under V_DS = 10 V and V_GS = −2 V conditions for both measurement and simulation. More specifically, f_T was defined as the intersection of the extension line at the current gain point (H₂₁) with the x-axis with a slope of −20 dB/decade [26]. The measured and simulated f_T values were 29.26 and 29.51 GHz, respectively, demonstrating excellent agreement with the minimal error rate of 0.9%.

3.2. Comparative Analysis of Entire Passivation Structures Based on Dielectric Materials

To accommodate high-power applications, the passivation layer of the Si₃N₄ EP structure was replaced with a high-k material. Two distinct dielectric materials were modeled (Al₂O₃ and HfO₂) for the EP structure, as shown in Figure 4. All structural parameters except for the passivation material remained unchanged during the simulation.

3.2.1. Simulation Results of the DC Characteristics

The DC characteristics of the Al₂O₃ and HfO₂ EP structures were compared with those of the basic Si₃N₄ EP structure. Figure 5 shows the I_DS-V_GS transfer characteristics of the three structures at V_DS = 10 V. No significant variations in the I_DS were observed, and the V_th remained constant at −3.0 V.

Figure 6a shows the electric field distributions within the channel layer at V_DS = 500 V for the three passivation structures. Compared with the Si₃N₄ EP structure, the Al₂O₃ and HfO₂ EP structures exhibited more efficient electric field dispersion, resulting in a lower maximum electric field in the channel layer due to their high dielectric constant. As the maximum electric field increased, the impact ionization that caused the generation of electron–hole pairs was enhanced; therefore, electric field redistribution effectively improved V_BD [27]. The dielectric constant of Al₂O₃ is lower than that of HfO₂, resulting in a relatively lower V_BD [28]. The V_BD characteristics were simulated under a V_GS = −7 V pinch-off condition to ensure a completely off device state. We defined V_BD as the V_DS when the I_DS exceeded 1 mA/mm after completely turning off the device by applying a voltage of −7 V across the gate. As shown in Figure 6b, the Si₃N₄, Al₂O₃, and HfO₂ EP structures exhibited V_BD values of 519.97, 554.39, and 610.70 V, respectively. The V_BD of the Al₂O₃ and HfO₂ EP structures were improved by 6.62% and 17.45%, respectively, compared with that of the Si₃N₄ EP structure.

3.2.2. Simulation Results of the RF Characteristics

Figure 7 shows the parasitic capacitance characteristics of the Si₃N₄, Al₂O₃, and HfO₂ EP structures. As shown in Figure 7a,b, the C_GS and C_GD were determined at V_DS = 10 V and V_GS = −2 V. The HfO₂ EP structure exhibited the highest C_GS and C_GD values, which can be attributed to the dielectric constant of HfO₂, as described by Equation (3). Conversely, the Al₂O₃ EP structure exhibited lower parasitic capacitance values than the HfO₂ EP structure, due to its lower dielectric constant.

Figure 8 shows the simulated f_T and V_BD values for the three EP structures. f_T simulations were conducted at V_DS = 10 V and V_GS = −2 V. According to Equation (4), the f_T values of the three EP structures were affected by transconductance (g_m) and C_GS. The Si₃N₄, Al₂O₃, and HfO₂ EP structures exhibited f_T values of 29.51, 28.16, and 16.07 GHz, respectively. Compared with the Si₃N₄ EP structure, the Al₂O₃ and HfO₂ EP structures exhibited reductions of 4.57% and 45.54%, respectively.

3.2.3. Simulation Results of the Hybrid Passivation Structure

To address the trade-off between enhanced V_BD values and degraded f_T associated with the application of Al₂O₃ and HfO₂ to the EP structures, HP structures were proposed by employing Al₂O₃ and HfO₂ into the 1st passivation and Si₃N₄ into the 2nd passivation. Figure 9 shows the schematics of the HP structures with Al₂O₃ and HfO₂.

Figure 10a shows the electric field distributions for three different structures. The maximum electric fields of the Al₂O₃ and HfO₂ HP structures were lower than those of the basic Si₃N₄ EP structure. The dielectric constant of Al₂O₃ is lower than that of HfO₂, resulting in a relatively lower V_BD. Specifically, the V_BD values of the Al₂O₃ and HfO₂ HP structures were 546.39 and 572.87 V, respectively, as shown in Figure 10b. However, compared with the EP structure, the HP structure exhibited a reduced V_BD because of the use of a high-k material only for the 1st passivation.

Figure 11 shows the parasitic capacitance characteristics of the different 1st passivation materials. Given that L_Gate-Source was shorter than L_Gate-Drain, C_GS exhibited a larger value than C_GD, indicating that the capacitance was affected by the distance between the electrodes [29]. Figure 11 shows that the HfO₂ HP structure exhibited the highest C_GS and C_GD values. Conversely, the Al₂O₃ HP structure exhibited lower parasitic capacitance values than the HfO₂ HP structure, which was due to the relatively low dielectric constant of Al₂O₃.

Figure 12 compares the simulated f_T values for the three dielectric passivation structures. Simulations conducted at V_DS = 10 V and V_GS = −2 V revealed f_T values of 28.63 and 26.46 GHz for the Al₂O₃ and HfO₂ HP structures, respectively. Compared with the Si₃N₄ EP structure, these values represent f_T reductions of 2.98% and 10.34% for the HfO₂ and Al₂O₃ HP structures, respectively. According to Equation (4), the decrease in f_T can be attributed to the increase in C_GS. Compared to the high-k EP structure, the HP structure compensated for the decrease in RF characteristics by applying a high-k material only at the 1st passivation layer.

3.3. Comparative Analysis of Partial Passivation Structures Based on Al₂O₃ and HfO₂

To mitigate the degradation of the RF characteristics observed in the HP structure while preserving the benefits of high-k materials, a PP structure was introduced. By implementing the HP structure, the RF characteristics were improved compared to the high-k EP structure. However, the f_T of the HP structure was lower than that of the basic Si₃N₄ EP structure. To minimize the degradation of the RF characteristics, we applied the PP structure with a high-k material only for the 1st passivation layer at the drain-gate region. Figure 13 shows schematic diagrams of the PP structure with Al₂O₃ and HfO₂.

3.3.1. Simulation Results of the DC Characteristics

The I_DS-V_GS transfer characteristics, electric field distribution, and V_BD characteristics of the Al₂O₃ and HfO₂ PP structures were simulated. As shown in Figure 14, the I_DS, g_m, and V_th values remained unaffected by variations in the material of the 1st passivation layer at the drain–gate region.

Figure 15a shows that the maximum electric field for the HfO₂ PP structure, which exhibited the highest dielectric constant, decreased and was dispersed in the drain–gate region. Conversely, the lower dielectric constant of Al₂O₃ in the Al₂O₃ PP structure resulted in less pronounced electric field dispersion. Figure 15b shows that the V_BD of the HfO₂ PP structure exhibited the highest V_BD value of 564.27 V, while the Si₃N₄ EP and Al₂O₃ PP structures exhibited comparable values of 519.97 and 532.08 V, respectively. Notably, the use of a high-k material as the 1st passivation layer at the drain–gate region, where the electric field peak occurs, resulted in a slight decrease in V_BD for the PP structure compared to the HP structure.

3.3.2. Simulation Results of the RF Characteristics

Figure 16 shows the parasitic capacitance characteristics of the Si₃N₄ EP, HfO₂, and Al₂O₃ PP structures. Given that all three structures employed Si₃N₄ as a passivation layer at the source–gate region, the C_GS remained consistent, as shown in Figure 16a. However, Figure 16b shows that the HfO₂ PP structure exhibits the highest C_GD value.

Figure 17 shows the simulated f_T values for different dielectric passivation structures at V_DS = 10 V and V_GS = −2 V. Notably, the f_T values of the Si₃N₄ EP, Al₂O₃ HP, and HfO₂ HP structures exhibited minimal variations (29.51, 29.44, and 29.37 GHz, respectively). Equation (4) indicates that f_T is primarily influenced by C_GS, and a negligible change in C_GS results in the observed f_T consistency. These results highlight the effectiveness of the PP structure in mitigating the degradation of the RF characteristics.

4. Discussion

This study simulates and analyzes the DC and RF characteristics of various dielectric passivation structures.

Table 4. A summary of DC and RF characteristics of various dielectric passivation structures of HEMT.

Parameters	Units	Si3N4	Al2O3			HfO2
Structure type	-	EP	EP	HP	PP	EP	HP	PP
Peak electric field	MV/cm	5.16	4.98	5.09	5.13	3.83	4.22	4.69
Breakdown voltage (VBD)	V	519.97	554.39	546.63	532.08	610.70	572.87	564.27
Cut-off frequency (fT)	GHz	29.51	28.16	28.63	29.44	16.07	26.46	29.37
Johnson’s figure-of-merit (JFOM)	THz-V	15.34	15.63	15.65	15.66	9.81	15.16	16.75

summarizes the DC and RF characteristics, including JFOM, for seven different dielectric passivation structures of the AlGaN/GaN HEMT. Among the Si₃N₄, Al₂O₃, and HfO₂ EP structures, the HfO₂ EP structure exhibited the highest V_BD. However, the high-k passivation layer inevitably entailed a decrease in f_T due to parasitic capacitance. To minimize the degradation of f_T, HP and PP structures were applied. The JFOM was calculated to analyze the trade-off relationship between V_BD and f_T. The basic Si₃N₄ EP structure has a JFOM of 15.34 THz-V. The JFOMs with three different Al₂O₃ passivation structures were not significantly different from the Si₃N₄ EP structure. However, the proposed HfO₂ PP structure exhibited the highest JFOM of 16.75 THz-V with enhanced V_BD while maintaining f_T.

5. Conclusions

This study investigates the DC and RF characteristics of AlGaN/GaN HEMTs using various passivation material configurations via TCAD simulation. The simulation parameters were obtained by matching the simulation data with the measured data of a fabricated basic Si₃N₄ EP structure of HEMT to ensure the reliability of the simulation results. The JFOM was calculated to assess the operational characteristics of each proposed dielectric passivation structure considering the trade-off between the breakdown voltage and cut-off frequency. Consequently, based on the highest calculated JFOM among the investigated structures, the HfO₂ PP structure was proposed as the optimal dielectric passivation structure for achieving superior breakdown voltage and frequency characteristics. This structure shows promise for high-power and high-frequency AlGaN/GaN HEMT applications.