Optimized Device Geometry of Normally-On Field-Plate AlGaN/GaN High Electron Mobility Transistors for High Breakdown Performance Using TCAD Simulation

Abstract

This study presents the optimization of the lateral device geometry and thickness of the channel and barrier layers of AlGaN/GaN high electron mobility transistors (HEMTs) for the enhancement of breakdown voltage (V_BR) characteristics using a TCAD simulation. The effect of device geometry on the device performance was explored by varying the device design parameters, such as the field plate length (L_FP), gate-to-drain length (L_GD), gate-to-source length (L_GS), gate length (L_G), thickness of the Si₃N₄ passivation layer (T_ox), thickness of the GaN channel (T_ch), and AlGaN barrier (T_barrier). The V_BR was estimated from the off-state drain current versus the drain voltage (I_DS–V_DS) curve, and it exhibited a strong dependence on the length and thickness of the parameters. The optimum values of V_BR for all the device’s geometrical parameters were evaluated, based on which, an optimized device geometry of the field-plated AlGaN/GaN HEMT structure was proposed. The optimized AlGaN/GaN HEMT structure exhibited V_BR = 970 V at I_GS = 0.14 A/mm, which was considerably higher than the results obtained in previous studies. The results obtained in this study could provide vital information for the selection of the device geometry for the implementation of HEMT structures.

1. Introduction

AlGaN/GaN high-electron mobility transistors (HEMTs) have attracted extensive attention for high-frequency and high-voltage applications owing to their excellent properties, such as the high electron mobility of their two-dimensional electron gas (2-DEG) channels, their wide energy band-gap, and their high breakdown field [1,2]. It is well known that the breakdown voltage (V_BR) of power devices is one of the most important parameters for providing reliable performance for high-power applications; however, the reported values are considerably below the theoretical limit [3,4,5]. Hence, it is necessary to improve the V_BR, especially without increasing the device’s size. Trew et al. reported a model that described the dependence of the V_BR on the avalanche process associated with the maximum electric field generated at the gate edge toward the drain side, resulting in a drain-to-gate breakdown [6,7].

The high breakdown field of GaN allows high voltages to be sustained between the drain and the gate. However, because of the much higher critical electric field of AlGaN, an early breakdown occurs near the heterointerface [8]. Various edge termination techniques, such as floating gates, field modulating plates, source-extended field plates, and multiple field plates, have been proposed to achieve a high V_BR and improve the device performance in AlGaN/GaN HEMTs [9,10,11]. To enhance the device’s performance for high-voltage power devices and microwave applications, GaN-based HEMTs were developed using field-plate technology, through which tremendous improvements in the V_BR and power densities were demonstrated [1,12,13,14,15,16]. The field plate is an extension of the gate deposited onto the passivation layer toward the drain side to minimize the electric field at the AlGaN surface. This leads to a reduction in the DC-to-RF dispersion, resulting in an increase in the V_BR [17]. Berzoy et al. [18] demonstrated an improvement in the V_BR in an AlGaN/GaN HEMT structure through the inclusion of various field plates on their structures. They reported an optimum V_BR of 880 V at a gate current of 40 A/mm for the best field plate case.

In addition to the field plate dimension, the device’s geometrical parameters, such as the gate length (L_G), gate width (W), source-drain distance (L_SD), source-gate separation (L_GS), gate-drain separation (L_GD), thicknesses of the Si₃N₄ passivation layer (T_ox), GaN channel (T_ch), and AlGaN barrier (T_barrier) can significantly affect the device’s characteristics. These parameters are correlated with and affected by each other. In other words, all the geometrical parameters should be tuned to achieve high performance GaN-based HEMTs with field plates. Nevertheless, reports on the effect of the device’s geometry on device performance are limited. For instance, as summarized in

Table 1. V_BR comparison of the various HEMT dimensions of the device structure (S: Simulation, E: Experiment).

Ref. Author, Year	S/E	LGD (µm)	LG (µm)	LFP (µm)	VBR (V)
[16] Karmalkar. S. (2001)	S	4.7	0.4	2	630
[21] Saito. W. (2003)	E	5 10	1.5 1.5	1.6 5	350 600
[22] Lin Zhu (2016)	S/E	3 5 7	3 3 3	- - -	120 220 320
[18] A. Berzoy (2017)	S	6.9	0.7	1.4	880
[23] D. Nirmal (2018)	S	2.7 4 6	0.25 0.25 0.25	1 1 1	291 370 420
[24] L. Wang (2018)	S	3 5 10	- - -	- - -	620 700 800
[25] P. Bhavana (2019)	E	4 6	0.7 0.7	- -	72 118
[26] A.S.A. Fletcher (2019)	E	2.7	0.25	0.9	300
[5] B. Liao (2019)	S	22	3	4	450
Present Study	S	5	1.5	1.5	970

, previous reports mainly investigated the effects of a few device geometrical parameters, such as L_GD, L_G, and field plate length (L_FP), on the operational characteristics of HEMTs, mainly in terms of V_BR. Unlike previous simulation studies, which focused primarily on the structural optimization of HEMTs by changing the limited number of geometrical parameters, the present work performed an extensive and systematic investigation of the impact of all the possible device design parameters, such as L_G, W, L_SD, L_GS, L_GD, T_ox, T_ch, and T_barrier on the DC output and V_BR characteristics of the field-plated AlGaN/GaN HEMTs, using technology computer aided design (TCAD) simulations. In the present study, the simulation was performed by varying one of the geometrical parameters of the field-plated AlGaN/GaN HEMT structure, while keeping the other geometrical parameters constant. Based on the simulation results, we propose an AlGaN/GaN HEMT structure with optimized device geometry to achieve the best possible device performance in terms of V_BR. In addition, a simulation of the output characteristics, transfer curve, and transconductance was performed. The optimized AlGaN/GaN HEMT structure exhibited a V_BR of 970 V at I_GS = 0.14 A/mm. The recorded values of the on-resistance and the power device figure of merit (FOM), defined by V_BR²/R_ON [19], were 3.12 Ω·cm and 0.3 MV²/Ω·cm, respectively. It should be noted that the operational behavior of the devices simulated by varying the key device geometrical parameters, i.e., L_GD, L_G, and L_FP, showed a similar tendency to previous works. In other words, a thorough comparison of the simulated results demonstrated in this work exhibited a close match with the previously reported experimental and simulated results. This implies that the accuracy, along with the validation, of our simulation approach is sufficient to provide verification of the results, although the present study does not include any experimental verification [20]. Furthermore, the AlGaN/GaN HEMT structure proposed can be achieved from a technological standpoint, of which optimization with realistic means is intricate, expensive, and time consuming. The optimized device design parameters obtained from the simulation in this study provide a potential guideline for the development of high-performance AlGaN/GaN HEMTs.

2. Materials and Methods

The schematic cross-section of the field-plated AlGaN/GaN HEMT structure used for the simulation and the corresponding energy band profile is shown in Figure 1. It was obtained using two-dimensional (2D) TCAD device simulator software. The transistor structure consisted of a 2 µm undoped GaN buffer layer on a sapphire substrate, a 200 nm thick GaN channel, and a 15 nm thick Al_0.28Ga_0.72N barrier layer [27]. The doping concentration of the GaN channel layer and AlGaN barrier layer were assumed to be 10¹⁵ cm⁻³ and 5 × 10¹⁶ cm⁻³, respectively, which is similar to those reported in Refs. [16,21,23,27]. The Al_0.25Ga_0.75N barrier layer provides an appropriate confinement of electrons towards the channel from the top. Without intentional doping, a huge number of electrons appears at the GaN channel due to piezoelectric and spontaneous polarization effects, resulting in two-dimensional electron gas (2DEG), as shown in Figure 1b. Si₃N₄ was used as the dielectric because it exhibits a high dielectric constant and good breakdown strength, and helps to avoid the cross-talk and noise interference from the atmosphere [23]. The gate field plate was deposited on the Si₃N₄ passivation layer, and the surface state at the interface between the passivation film and AlGaN was neglected. The source and drain ohmic contacts were formed on the 50 nm thick n-type GaN layer with a doping concentration of 5 × 10¹⁹ cm⁻³ grown on the GaN channel. The simulations were performed using a 2D device simulator and Atlas TCAD device software. The material parameters used in the simulation are listed in

Table 2. Physical parameter of AlGaN/GaN used in Atlas TCAD simulation [16,21,23,27].

TCAD Parameters	GaN	AlGaN	Unit
Electron mobility (µn)	900	600	cm2/V-s
Hole mobility (µp)	10	10	cm2/V-s
Energy band gap (Eg)	3.40	3.96	eV
Conduction band density of state (Nc)	1.07	2.07	1018/cm3
Valance band density of state (Nv)	1.16	1.16	1019/cm3
Saturation velocity (Vsat)	2.00	1.10	107 cm/s
Relative permittivity (ɛ)	9.50	9.50	-
Trap lifetime (e, p)	1	1	107 s

[16,21,23,27]. The gate electrode formed for the proposed structure was assumed to be metal, with a work function of 5.23 eV. The Newton method was used to solve the built-in equations in the Atlas TCAD, such as the transport equation, and Poisson’s equation. The simulations were performed by varying the various device geometrical parameters, such as the L_FP, L_GD, L_GS, L_G, T_ox, T_ch, and T_barrier. The V_BR was calculated from the off-state drain current versus drain voltage (I_DS–V_DS) curve with a gate bias of V_G = −6 V. The V_BR was determined as the drain voltage for a drain current of 1 mA/mm. Following the simulation results, the values of the geometrical parameters were optimized. These optimized parameter values are included in the caption of Figure 1.

3. Results and Discussion

In this study, a TCAD simulation was performed for an AlGaN/GaN HEMT structure, shown in Figure 1, by varying the dimensions of the lateral device geometry. It is well known that the length of the various geometries of the HEMT structure affects the device’s performance and V_BR. The field plate connected to the gate was introduced to modify the electric field at the gate edge on the drain side, which significantly suppressed the effect of the surface traps and increased the V_BR of the HEMT devices [1]. The V_BR is directly related to the L_FP because the extension of the gate edge point redistributes the peak electric field in the channel layer and reduces the electron punch-through effect. A simulation of the effect of the L_FP on the V_BR was carried out for the proposed HEMT structure, as shown in Figure 1. The results are depicted in Figure 2 as a function of L_FP varying in the range of 0.5–2.5 µm, with the other geometrical parameters constant. It was observed that the V_BR increased along with the increase in L_FP until it reached a value of 1.5 µm, at a maximum V_BR of 956 V. This increase in V_BR with an increase in L_FP was associated with the formation of a depletion layer under the field plate electrode, which reduces the electric field strength at the gate on the drain side [10]. However, a further increase in the L_FP beyond 1.5 µm led to a reduction in the value of the V_BR, which was attributed to the increase in the electric field-driven impact ionization process at the drain side of the field plate edge [21,28]. Furthermore, a longer field plate extension increased the main electric field peak under the field plate edge, resulting in an increase in the gate capacitance and a decrease in the V_BR [16].

It is well known that the L_GD affects the V_BR characteristics of HEMTs, and an increase in the L_GD leads to an increase in the V_BR of the device, which is associated with the higher spacing [29]. Figure 3a shows the breakdown characteristics of the AlGaN/GaN HEMTs at various values of L_GD varying in the range of 3–15 µm. Figure 3b shows the variation in the V_BR with respect to the L_GD. It is evident that the V_BR increased along with the increase in the L_GD. This may have been associated with the reduction in the peak electric field at the gate edge of the drain side. However, V_BR strongly depends on the electric field at the gate field-plate edge on the drain side. Further, it can be observed from Figure 3a that the breakdown occurred at I_DS = 0.5 A, regardless of the L_GD. Figure 3 shows the V_BR characteristics for the field-plated AlGaN/GaN HEMT, with the device’s structure shown in Figure 1a, as a function of varying gate-drain lengths in the range of 3–15 µm at V_GS = −6 V. It is noteworthy that the drain current remained constant until a voltage of 400 V, regardless of the L_G. However, for L_GD = 3 µm, the current increased drastically for voltages greater than 400 V. For the AlGaN/GaN HEMT, the current levels were sustained until higher voltages, which increased with an increase in the L_GD. From Figure 3b, it can be observed that the V_BR increased to 600 V, 900 V, 1150 V, and 1175 V, with an increase in the L_GD of 3 µm, 5 µm, 10 µm, and 15 µm. The V_BR increased proportionally with the increase in the L_GD when the breakdown was primarily due to the high electric field at the drain side of the gate edge. As the L_GD increased, the electric field at the gate edge reduced, with an increase in the distance between the gate and the drain. The breakdown voltage increased along with the increase in the L_GD associated with the increased parasitic channel resistance [30].

Figure 4 displays the V_BR curves of the AlGaN/GaN HEMT at various values of L_GS varying from 1 µm to 5 µm. Figure 4b depicts the values of V_BR with the variation in L_GS at V_G = −6 V. It is notable that the impact of the L_GS on the I_DS was negligible. The V_BR increased along with the increase in the L_GS at V_G = −6 V, resulting in the extraction of higher output power. Unlike an increase in L_GD, an increase in L_GS does not increase the width of the leakage path. This is because at high leakage, the current flows primarily through the source terminal, and not the source-side channel. Therefore, an increase in the L_GS will lead to an increase in the resistance of the bugger layer, resulting in an increase in the V_BR. This could be associated with the increase in the space between the source and gate, enabling the distribution of the electric field over a wider area. As shown in Figure 4, the L_GS increased as a function of V_BR. As shown in the plot of the simulation results, the V_BR increased along with the increases in the L_GS of 1 µm, 2 µm, 3 µm, and 5 µm at V_G = −6 V, which then increased to 885 V, 920 V, 930 V, and 980 V, respectively. The enhancement in the electric field with the spacer between the source and the gate resulted in the improvement of the V_BR performance. The optimization of L_GS has a significant impact on the V_BR of the device. This is because of the source-injection buffer leakage, in contrast to the decrease in V_BR caused by the increase in L_GD, with the decrease in the width of the current flow.

Figure 5a shows the V_BR characteristics of the AlGaN/GaN HEMTs as a function of L_G varying from 0.2 µm to 2.5 µm. As the L_G increased, the area of the gate electrode increased, while the areas of the source and drain electrodes remained the same. The leakage current decreased along with an increase in L_G. As can observed in Figure 5b, the V_BR increased along with the increase in L_G, varying in the range of 110–960 V, with the L_G varying in the range of 0.2–2.5 µm. Furthermore, the current levels were lowered significantly with the increase in L_G. This could be due to the fact that an increase in L_G suppresses the punch-through and improves V_BR [31]. In addition, a higher value of L_G lowers the impact ionization at the gate edge on the drain side, which significantly lowers the leakage current. Additionally, a higher value of L_G effectively suppresses the drain-induced barrier lowering [32].

The V_BR behavior of the AlGaN/GaN heterostructure field-plated HEMT as a function of the thickness of the Si₃N₄ insulator passivation layer was investigated through simulation. Figure 6a shows the variation in the V_BR of the AlGaN/GaN HEMTs as a function of the Si₃N₄ thickness in the range of 0.1–0.7 µm. It is noteworthy that the leakage current increased along with the increase in Si₃N₄ thickness. By contrast, it can be observed from Figure 6b that the V_BR of the HEMT increased initially, exhibiting a higher value of 970 V for the Si₃N₄ layer with a thickness of 0.2 µm, followed by a drastic decrease in the V_BR along with increasing thickness of the Si₃N₄ insulator passivation layer. Generally, in the case of an insulator layer with thickness at the lower end of a certain thickness range, the electric field is concentrated at the field plate edge. Conversely, for a layer with maximum thickness, the field concentration is at the gate edge [17]. For a layer with an optimum thickness, the field is distributed between the gate edge and the field plate edge. The observed decrease in the V_BR with an increase in the thickness above 0.2 µm (the optimum thickness) implies that the nitride layer used was extremely thick and the field plate was far away from the semiconductor, so that the field plate did not have any effect on the electric field distribution. This demands the optimization of the thickness of the insulator passivation layer to produce a high-performance AlGaN/GaN HEMT.

Figure 7a shows the V_BR characteristics of the AlGaN/GaN HEMT as a function of the GaN channel layer thickness (T_ch) varying between 100 nm, 200 nm, 500 nm, 1000 nm, and 2000 nm. From Figure 7a, it can be seen that the leakage current increased along with the increase in the channel layer thickness. This resulted in a decrease in the V_BR, as shown in Figure 7b. This increase in leakage current with an increase in the channel layer thickness was associated with the wider current path and higher 2DEG concentration resulting from the thicker channel region. This lowered the channel resistance, which tends to lower the V_BR. Furthermore, the decrease in V_BR might have been associated with the increase in the drain-induced barrier lowering with an increase in the channel thickness. In addition, the deterioration of the gate control causes higher punch-through leakage [33].

Figure 8a shows the V_BR characteristic curves of the AlGaN/GaN HEMTs as a function of the AlGaN barrier layer thickness varying in the range of 10–35 nm. It was observed that the drain current increased slightly along with the increase in the thickness of the AlGaN barrier layer. A total of V_BR = 860 V was obtained for the AlGaN/GaN HEMT (Figure 8b) with an AlGaN barrier layer thickness of 10 nm. The V_BR increased to 935 V along with the increase in the thickness of the AlGaN barrier layer to 15 nm, and subsequently decreased along with the increase in the thickness of the AlGaN barrier layer. This implies that the critical value of the thickness of the AlGaN barrier layer is 15 nm for effective device operation. The decrease in the V_BR below a certain critical value of the thickness of the AlGaN barrier layer could be explained as follows. For the AlGaN/GaN HEMT with low AlGaN barrier thickness, the surface donor-like trap level was below the Fermi level. At the critical barrier thickness, the surface traps reached the Fermi level and the electrons from these traps were driven into the channel by the strong polarization-induced electric field in AlGaN, creating the 2DEG and leaving behind the positive charge [34]. Until all the surface traps were empty, the Fermi level remained essentially at the donor energy and an increase in the number of electrons transferred with an increase in the barrier thickness. As the negative voltage was increased, the electrons from the gate might have leaked into the trap states in the ungated surfaces and created a “virtual gate” modulating the depletion region. This abrupt change in the gate voltage led to the RF drain current collapse phenomenon.

The simulated output and transfer characteristics of the proposed field-plated AlGaN/GaN HEMT structure shown in Figure 1, with an optimized geometry of L_GS = 2 µm, L_G = 1.5 µm, L_GD = 5 µm, T_ox = 0.2 µm, and L_FP = 1.5 µm, are presented in Figure 9a,b, respectively. The I_DS–V_DS characteristics in Figure 9a are shown as a function of V_GS varying from –4 V to +1 V in steps of 1 V. It can be noted that the current increased distinctly along with the increase in V_GS. The drain current I_DS was 186 mA/mm at 14 V for V_GS = 0 V. Figure 9b displays the corresponding transfer characteristic and transconductance curve at V_DS = 10 V. The device exhibited a V_th of approximately −4 V, I_max of 310 mA/mm at V_GS = 2 V, and a peak transconductance (G_m) of 70 mS/mm at V_GS = 1 V. The V_BR of the optimized device structure with the gate bias was maintained at −6 V, and the device displayed a value of 970 V at I_GS = 0.14 A/mm, as shown in Figure 9c. The various values obtained from the optimized AlGaN/GaN HEMT structure are mentioned in Figure 1 and the results are presented in Table 2. An on-resistance (R_ON) of 3.12 Ω cm, determined by the slope of the I_DS–V_DS curve at V_GS = 0 V, and the power device figure of merit (FOM) of 0.3 MV²/Ω cm, are comparable with the results obtained from the simulated and experimentally fabricated devices possessing various HEMT structures reported earlier.

4. Conclusions

The impact of device geometry and the thickness of the channel and barrier layer on the V_BR characteristics of field-plated AlGaN/GaN high-electron mobility transistors (HEMTs) was explored using Atlas TCAD numerical simulation. The effect of device geometry was investigated by varying the parameters, such as the L_FP, L_GD, L_GS, L_G, T_ox, T_ch, and T_barrier. The V_BR strongly depends on the length and thickness of the aforementioned parameters. Based on the optimum V_BR values obtained for all the geometrical parameters, an optimized device geometry of the field-plated AlGaN/GaN HEMT structure was proposed following a simulation of the output, transfer characteristics, and transconductance. A V_BR of 970 V was obtained at I_GS = 0.14 A/mm for the optimized AlGaN/GaN HEMT structure. This optimized device geometry of the proposed AlGaN/GaN HEMT structure could be useful for the implementation of HEMT-structured AlGaN/GaN heterostructures.