Effect of gate length on breakdown voltage in AlGaN/GaN high-electron-mobility transistor
Luo Jun1, Zhao Sheng-Lei1, Mi Min-Han1, Chen Wei-Wei2, Hou Bin2, Zhang Jin-Cheng1, Ma Xiao-Hua1, 2, Hao Yue1, †,
Key Laboratory of Wide Band-Gap Semiconductor Materials and Devices, School of Microelectronics, Xidian University, Xi’an 710071, China
School of Advanced Materials and Nanotechnology, Xidian University, Xi’an 710071, China

 

† Corresponding author. E-mail: yhao@xidian.edu.cn

Project supported by the National Natural Science Foundation of China (Grant Nos. 61334002, 61106106, and 61204085).

Abstract
Abstract

The effects of gate length LG on breakdown voltage VBR are investigated in AlGaN/GaN high-electron-mobility transistors (HEMTs) with LG = 1 μm∼ 20 μm. With the increase of LG, VBR is first increased, and then saturated at LG = 3 μm. For the HEMT with LG = 1 μm, breakdown voltage VBR is 117 V, and it can be enhanced to 148 V for the HEMT with LG = 3 μm. The gate length of 3 μm can alleviate the buffer-leakage-induced impact ionization compared with the gate length of 1 μm, and the suppression of the impact ionization is the reason for improving the breakdown voltage. A similar suppression of the impact ionization exists in the HEMTs with LG > 3 μm. As a result, there is no obvious difference in breakdown voltage among the HEMTs with LG = 3 μm∼20 μm, and their breakdown voltages are in a range of 140 V–156 V.

1. Introduction

Owing to their high two-dimensional electron gas (2DEG) density, electron mobility, and high breakdown electric field, AlGaN/GaN high-electron-mobility transistors (HEMTs) have become excellent candidates in power electronics and high frequency applications.[16] The value of gate length has influences on transconductance, cutoff frequency, and maximum frequency of oscillation.[7,8] Numerous studies focused on reducing the gate length to improve high-frequency characteristics.[9,10] Breakdown voltage is also a very important parameter for these applications. However, the effect of gate length LG on breakdown voltage VBR has rarely been investigated.

In this paper, the effect of gate length on breakdown voltage is investigated through experiments and simulations. With the increase of LG, VBR is first increased, and then saturated at LG = 3 μm. In order to explain the relation between VBR and LG, leakage currents and software simulation are utilized to discuss the buffer leakage current and buffer-leakage-induced impact ionization.

2. Device structure and fabrication

Figure 1 shows the schematic cross section of AlGaN/GaN HEMT fabricated in this paper. The AlGaN/GaN heterostructure was grown on sapphire substrate in a metal–organic chemical vapor deposition (MOCVD) system. The epitaxial layers included a 100-nm AlN nucleation layer, 1.4-μm GaN buffer layer, 1-nm AlN interlayer, 21-nm AlGaN barrier layer, and 1.5-nm GaN cap layer. The sheet carrier density and 2DEG mobility were tested to be 1.01×1013 cm−2 and 1915 cm2/V·s respectively by Hall effect measurements.

Fig. 1. Schematic cross section of AlGaN/GaN HEMT.

Source and drain electrodes were formed by evaporating Ti/Al/Ni/Au followed by annealing in N2 ambient at 830 °C for 30 s. Then, mesa isolation was carried out by Cl-based reactive ion etching (RIE). Next, a Schottky gate was formed by evaporating Ni/Au/Ni. The devices were passivated with SiN layers to reduce surface states. The gate-source spacing and gate-drain spacing are all 3 μm. Six kinds of devices were fabricated with different gate length LG values, and the gate length values are 1, 3, 5, 8, 13, and 20 μm.

3. Results and discussion

The source–drain spacing increases with the increase of gate length from 1 μm to 20 μm, leading to the increase of on-state resistance. As a result, the maximum drain current IDmax decreases with the increase of gate length. As shown in Fig. 2(a), IDmax of the HEMT with LG = 1 μm is 743 mA/mm. In contrast, IDmax is reduced to 342 mA/mm for the HEMT with LG = 20 μm. In addition, the maximum transconductance Gm,max decreases with the increase of gate length as shown in Fig. 2(b). Gm,max decreases from 161 mS/mm for the HEMT with LG = 1 μm to 82 mS/mm for the HEMT with LG = 20 μm.

Fig. 2. IV transfer characteristics of the AlGaN/GaN HEMTs with different gate length values: (a) IDS versus VGS and (b) Gm versus VGS.

Figure 3 shows the breakdown characteristics of the AlGaN/GaN HEMTs with different gate length values. For the HEMT with LG = 1 μm, breakdown voltage VBR is 117 V. In contrast, the values of VBR are in a range of 140 V–156 V for the HEMTs with LG = 3 μm–20 μm. VBR is first increased, and then saturated at LG = 3 μm. In order to explain the relation between VBR and LG, breakdown characteristics of the HEMTs with LG = 1 μm and LG = 3 μm are first analyzed.

Fig. 3. (a) Breakdown characteristics and (b) VBRLG dependence of the AlGaN/GaN HEMTs with different gate length values.

Figure 4 shows three leakage currents, ID, IG, and IS, in the HEMTs with LG = 1 μm and LG = 3 μm during the breakdown measurements. Source current IS first decreases, and then increases, resulting in an inflection point.[11] As shown in Fig. 5, IS consists of source–gate leakage current Isg and buffer leakage current Ids, which flow in different directions. Before the inflection point, the source current is mainly composed of source–gate leakage current, which is part of gate leakage current. The buffer leakage current increases with the increase of the drain voltage. When the buffer leakage current exceeds the source–gate leakage current, the inflection point occurs. After the inflection point, the increase of IS results from the increase of the buffer leakage current.[11]

Fig. 4. Variations of ID, IG, and IS with VDS (drain–source voltage) during the breakdown measurements in HEMTs with LG = 1 μm and LG = 3 μm.
Fig. 5. Off-state leakage currents in AlGaN/GaN HEMTs during the breakdown measurements. Gate leakage current IG consists of source–gate leakage current Isg and drain–gate leakage current Idg.

As shown in Fig. 4, VBR for the HEMT with LG = 3 μm is 148 V, which is 31 V higher than that for the HEMT with LG = 1 μm. For the HEMT with LG = 1 μm, IS increases quickly at VDS = 82 V, and achieves 1 mA/mm at VDS = 117 V, indicating that the device breakdown is induced by the buffer leakage current. In contrast, IS increases slowly for the HEMT with LG = 3 μm during the whole breakdown measurement. The breakdown for the HEMT with LG = 3 μm is induced by the gate leakage current. The increase of VBR for the HEMT with LG = 3 μm compared with the HEMT with LG = 1 μm is due to the suppression of the buffer leakage current. In order to explain the reason for suppressing the buffer leakage current, Silvaco-ATLAS software is utilized to investigate the depletion region and electric field distributions in the AlGaN/GaN HEMTs.

Figures 6(a) and 6(d) show the electron concentration distributions of the AlGaN/GaN HEMTs with VGS = −6 V and VDS = 0 V. With the increase of the drain terminal, the depletion region under the gate is shrinking, which results from the effect of drain induced barrier lowering (DIBL). For VDS = 100 V, the depletion region in the HEMT with LG = 1 μm is smaller than that in the HEMT with LG = 3 μm as shown in Figs. 6(b) and 6(e). The comparison of these two depletion regions indicates that a larger gate length can suppress DIBL more effectively.

Fig. 6. Off-state logarithmic electron concentration distributions of the AlGaN/GaN HEMTs respectively with (a) LG = 1 μm, VDS = 0 V; (b) LG = 1 μm, VDS = 100 V; (d) LG = 3 μm, VDS = 0 V; (e) LG = 3 μm, VDS = 100 V. Electric field distributions of the AlGaN/GaN HEMTs respectively with (c) LG = 1 μm, VDS = 100 V; (f) LG = 3 μm, VDS = 100 V. All the devices are biased at VGS = −6 V and simulated by Silvaco-ATLAS software.

With the drain biased at a high value, there exists an electric field peak at the drain-side edge of the gate terminal as shown in Figs. 6(c) and 6(f). Correspondingly, the GaN-based materials around the gate are in the high electric field region. As shown in Fig. 6(b), part of the high electric field region in the HEMT with LG = 1 μm is not depleted, and electrons can be injected from the source to this part of the high electric field region, which may initiate impact ionization and the following avalanche breakdown. Hence, the buffer leakage current increases quickly due to the impact ionization at VDS = 82 V, and achieves 1 mA/mm at VDS = 117 V. For the HEMT with LG = 3 μm, the high electric field region is almost completely depleted, and the electrons from the source are difficult to be injected into the high electric field region. The buffer leakage current of the HEMT with LG = 3 μm does not show the abrupt increase during the whole breakdown measurement. Therefore, the gate length of 3 μm can alleviate the buffer-leakage-induced impact ionization compared with the gate length of 1 μm, and the suppression of the impact ionization is the reason for improving the breakdown voltage.

The breakdown for the HEMT with LG = 1 μm belongs to buffer-induced breakdown. The suppression of the buffer leakage changes the breakdown type into gate-induced breakdown for the HEMT with LG = 3 μm. For the gate length larger than 3 μm, the high electric field region is almost completely depleted, which is similar to the case of LG = 3 μm. A similar suppression of the impact ionization exists in each of the HEMTs with LG > 3 μm. For the HEMTs with LG = 5, 8, 13, and 20 μm, the device breakdown types are all gate-induced breakdown. As shown in Table 1, the gate leakage current values are almost equal to each other for the HEMTs with LG = 3, 5, 8, 13, and 20 μm. As a result, there is no obvious difference in breakdown voltages among the HEMTs with LG = 3 μm–20 μm, and they are in a range of 140 V–156 V.

Table 1.

Parameter values of AlGaN/GaN HEMTs. The threshold voltage Vth is defined as the gate voltage with drain leakage current reaching 1 mA/mm. The gate leakage current Igleak values are obtained at VGS = −15 V.

.

In order to find out the accurate gate length in which the buffer-leakage-induced impact ionization can be alleviated, the simulation for the HEMT with LG = 2 μm is carried out, and the results are shown in Fig. 7(a). For the HEMT with LG = 2 μm, part of the high electric field region is not depleted, which may result in impact ionization and avalanche breakdown. Then, the simulation for the HEMT with LG = 2.5 μm is carried out, and the results are shown in Fig. 7(b). For the HEMT with LG = 2.5 μm, the high electric field region is almost completely depleted. Therefore, the gate length value in which buffer-leakage-induced impact ionization begins to be alleviated is in a range between 2 μm and 2.5 μm.

Fig. 7. Off-state logarithmic electron concentration distributions of the AlGaN/GaN HEMTs respectively with (a) LG = 2 μm, VDS = 100 V; (b) LG = 2.5 μm, VDS = 100 V.
4. Conclusions

In this work, the effects of gate length on breakdown voltage are investigated in AlGaN/GaN HEMTs. The value of VBR for the HEMT with LG = 3 μm is 148 V, which is 31 V higher than that for the HEMT with LG = 1 μm. For the HEMT with LG = 3 μm, the high electric field region is almost completely depleted, and it is difficult to inject the electrons from the source into the high electric field region. The gate length of 3 μm can alleviate the buffer-leakage-induced impact ionization compared with the gate length of 1 μm, and the suppression of the impact ionization is the reason for improving the breakdown voltage. A similar suppression of the impact ionization exists in each of the HEMTs with LG > 3 μm, and there is no obvious difference in breakdown voltage among the HEMTs with LG = 3 μm–20 μm. Simulation results indicate that the gate length value in which buffer-leakage-induced impact ionization begins to be alleviated is in a range between 2 μm and 2.5 μm.

Reference
1Lee H SPiedra DSun MGao XGuo SPalacios T 2012 IEEE Electron Dev. Lett. 33 982
2Wu Y FSaxler AMoore MSmith R PSheppard SChavarkar P MWisleder TMishra U KParikh P 2004 IEEE Electron Dev. Lett. 25 117
3Huang XLiu ZLi QLee F C 2014 IEEE Trans. Power Electron. 29 2453
4Xing H LDora YChini AHeikman SKeller SMishra U K 2004 IEEE Electron Dev. Lett. 25 161
5Yang LHu G ZHao YMa X HQuan SYang L YJiang S G 2010 Chin. Phys. B 19 047301
6Zhao S LChen W WYue TWang YLuo JMao WMa X HHao Y 2013 Chin. Phys. B 22 117307
7Zhou Q Huang S Chen H Zhou C Feng Z Cai S Chen K J 2011 IEEE International Electron Devices Meeting December 5–7, 2011 Washington, USA p. 33.4.1 10.1109/IEDM.2011.6131664
8Lee D SLaboutin OCao YJohnson WBeam EKetterson ASchuette MSaunier PPalacios T 2012 IEEE Electron Dev. Lett. 33 976
9Yue YHu ZGuo JSensale-Rodriguez BLi GWang RFaria FFang TSong BGao XGuo SKosel TSnider GFay PJena DXing H 2012 IEEE Electron Dev. Lett. 33 988
10Downey B PMeyer D JKatzer D SRoussos J APan MGao X 2014 IEEE Electron Dev. Lett. 35 527
11Zhao S LHou BChen W WMi M HZheng J XZhang J CMa X HHao Y 2016 IEEE Trans. Power Electron. 31 1517